Amd trinity review: Trinity on the Desktop, Part 2

Trinity on the Desktop, Part 2

by Anand Lal Shimpion October 2, 2012 1:45 AM EST

  • Posted in
  • GPUs
  • CPUs
  • AMD
  • Trinity

178 Comments
|

178 Comments

Introduction & The TestTrinity CPU Performance: The Good and the BadGeneral Performance — SYSMark 2012Content Creation PerformanceVideo Transcoding PerformanceDiscrete GPU Gaming PerformanceOverclocking & Power ConsumptionPower ConsumptionFinal Words


Although AMD’s second-generation mainstream APU platform, codename Trinity, launched months ago in notebooks the official desktop launch is today. Rumor has it that AMD purposefully delayed the desktop Trinity launch to clear out unsold Llano inventories in the channel. Although selling APUs in notebooks is pretty easy, convincing desktop users to forgo the discrete GPU option (and ignore Intel) has been a tough battle for AMD. I keep going back to two slides that show us where AMD wants to go and the cores it’ll take to get there:


The ultimate goal is this beautiful cohesive operation between CPU and GPU on a single die. That future will require a lot of software support, not only at the application level but also at the OS level. And I’m not talking about Windows 8. We’re still far away from this APU dominated future, but AMD is marching in that direction. The second slide shows the x86 cores that we’ll see from AMD along the way. AMD is still playing catch-up in the x86 CPU space and it’s got a lot of lost time to make up for. There’s no hiding the fact this is going to be a multi-year effort to simply get close to Intel’s single-threaded x86 performance. Through pricing, leveraging its GPU technology and throwing more transistors at the problem AMD can still deliver competitive solutions, but it’s not going to be a walk in the park.


Last week we took a look at the GPU side of the desktop Trinity APUs. We looked at the top end 384-core Radeon HD 7660D configuration as well as the slightly slower 256-core Radeon HD 7560D GPU, both of which easily outperformed Intel’s HD 4000 and HD 2500. As far as processor graphics go, Trinity on the desktop maintains a healthy lead over Intel. There’s still a place for discrete GPUs but that’s pretty much at the $100 and above price points.


Today we’re able to talk about pricing and x86 CPU performance among other things. The good news on that front is the most expensive Trinity APU is fully unlocked and is priced at $122:












AMD Socket-FM2 Lineup

 

Modules/Cores

CPU Clock Base/Turbo

L2 Cache

GPU

TDP

Price

A10-5800K

2 / 4

3. 8 / 4.2 GHz

4MB

384 cores @ 800MHz

100W

$122

A10-5700

2 / 4

3.4 / 4.0 GHz

4MB

384 cores @ 760MHz

65W

$122

A8-5600K

2 / 4

3.6 / 3.9 GHz

4MB

256 cores @ 760MHz

100W

$101

A8-5500

2 / 4

3. 2 / 3.7 GHz

4MB

256 cores @ 760MHz

65W

$101

A6-5400K

1 / 2

3.6 / 3.8 GHz

1MB

192 cores @ 760MHz

65W

$67

A4-5300

1 / 2

3.4 / 3.6 GHz

1MB

128 cores @ 724MHz

65W

$53

Athlon X4 750K

2 / 4

3. 4 / 4.0 GHz

4MB

N/A

100W

$81

Athlon X4 740

2 / 4

3.2 / 3.7 GHz

4MB

N/A

65W

$71


Compare this to Llano’s launch where the top end SKU launched at $135 and you’ll see that AMD is somewhat getting with the times. I would still like to see something closer to $100 for the A10-5800K, but I find that I’m usually asking for a better deal than what most CPU makers are willing to give me.


AMD’s competitive target is Intel’s newly released Ivy Bridge Core i3 processors. There are only five Core i3s on the market today, four of which use Intel’s HD 2500 graphics. The cheapest of the lineup is the Core i3 3220 with two cores running at 3.3GHz for $125. Intel disables turbo and other features (there’s effectively no overclocking on these parts), which AMD is attempting to exploit by pitting its Trinity K-series SKUs (fully unlocked) against them. AMD’s TDPs are noticeably higher (100W for the higher end K-series parts compared to 55W for the Core i3s). Intel will easily maintain the power advantage as a result under both CPU and GPU load, although AMD’s GPU does deliver more performance per watt. Power consumption is a major concern of AMD’s at this point. Without a new process node to move to for a while, AMD is hoping to rely on some design tricks to improve things in the future.


At the low end of the stack there are also two Athlon X4s without any active GPU if you just want a traditional Trinity CPU.


This will be our last CPU/APU review on the current test platform/software configuration. The next major CPU review will see a move to a brand new testbed running Windows 8. As always you can get access to far more numbers than what we report here if you use our performance comparison engine: Bench. Of course if you want to see the GPU and GPU Compute performance of AMD’s Trinity APU check out part one of our coverage.









Motherboard:

ASUS P8Z68-V Pro (Intel Z68)

ASUS Crosshair V Formula (AMD 990FX)

Gigabyte GA-F2A85X-UP4 (AMD A85X)

Intel DZ77GA-70K (Intel Z77)

Hard Disk:

Intel X25-M SSD (80GB)

Crucial RealSSD C300

OCZ Agility 3 (240GB)

Memory:

2 x 4GB G. Skill Ripjaws X DDR3-1600 9-9-9-20

Video Card:

ATI Radeon HD 5870 (Windows 7)

AMD Processor Graphics

Intel Processor Graphics

Video Drivers:

AMD Catalyst 12.8

Desktop Resolution:

1920 x 1200

OS:

Windows 7 x64


 

Trinity CPU Performance: The Good and the Bad
Introduction & The TestTrinity CPU Performance: The Good and the BadGeneral Performance — SYSMark 2012Content Creation PerformanceVideo Transcoding PerformanceDiscrete GPU Gaming PerformanceOverclocking & Power ConsumptionPower ConsumptionFinal Words

Tweet

PRINT THIS ARTICLE

The AMD Trinity Review (A10-4600M): A New Hope

by Jarred Waltonon May 15, 2012 12:00 AM EST

  • Posted in
  • Laptops
  • AMD
  • Trinity
  • GPUs
  • 7000M
  • A10

271 Comments
|

271 Comments

Introduction and Piledriver OverviewImproved Turbo, Beefy Interconnects and the Trinity GPUMobile Trinity LineupMeet the AMD Trinity/Comal PrototypeAMD Trinity General PerformanceAMD Trinity Gaming PerformanceAMD’s Heterogeneous Computing with TrinityAMD Trinity: Battery Life Also ImprovedConclusion: What Makes a Trinity?

Introduction and Piledriver Overview

Brazos and Llano were both immensely successful parts for AMD. 2 in Llano).

Trinity Physical Comparison
  Manufacturing Process Die Size Transistor Count
AMD Llano 32nm 228mm2 1.178B
AMD Trinity 32nm 246mm2 1.303B
Intel Sandy Bridge (4C) 32nm 216mm2 1. 16B
Intel Ivy Bridge (4C) 22nm 160mm2 1.4B

Without a change in manufacturing process, AMD is faced with the tough job of increasing performance without ballooning die size. Die size has only gone up by around 7%, but both CPU and GPU performance see double-digit increases over Llano. Power consumption is also improved over Llano, making Trinity a win across the board for AMD compared to its predecessor. If you liked Llano, you’ll love Trinity.

The problem is what happens when you step outside of AMD’s world. Llano had a difficult time competing with Sandy Bridge outside of GPU workloads. AMD’s hope with Trinity is that its hardware improvements combined with more available OpenCL accelerated software will improve its standing vs. Ivy Bridge.

Piledriver: Bulldozer Tuned

While Llano featured as many as four 32nm x86 Stars cores, Trinity features up to two Piledriver modules. Given the not-so-great reception of Bulldozer late last year, we were worried about how a Bulldozer derivative would stack up in Trinity. I’m happy to say that Piledriver is a step forward from the CPU cores used in Llano, largely thanks to a bunch of clean up work from the Bulldozer foundation.

Piledriver picks up where Bulldozer left off. Its fundamental architecture remains completely unchanged, but rather improved in all areas. Piledriver is very much a second pass on the Bulldozer architecture, tidying everything up, capitalizing on low hanging fruit and significantly improving power efficiency. If you were hoping for an architectural reset with Piledriver, you will be disappointed. AMD is committed to Bulldozer and that’s quite obvious if you look at Piledriver’s high level block diagram:

Each Piledriver module is the same 2+1 INT/FP combination that we saw in Bulldozer. You get two integer cores, each with their own schedulers, L1 data caches, and execution units. Between the two is a shared floating point core that can handle instructions from one of two threads at a time. The single FP core shares the data caches of the dual integer cores.

Each module appears to the OS as two cores, however you don’t have as many resources as you would from two traditional AMD cores. This table from our Bulldozer review highlights part of problem when looking at the front end:

Front End Comparison
  AMD Phenom II AMD FX Intel Core i7
Instruction Decode Width 3-wide 4-wide 4-wide
Single Core Peak Decode Rate 3 instructions 4 instructions 4 instructions
Dual Core Peak Decode Rate 6 instructions 4 instructions 8 instructions
Quad Core Peak Decode Rate 12 instructions 8 instructions 16 instructions
Six/Eight Core Peak Decode Rate 18 instructions (6C) 16 instructions 24 instructions (6C)

It’s rare that you get anywhere near peak hardware utilization, so don’t be too put off by these deltas, but it is a tradeoff that AMD made throughout Bulldozer. In general, AMD opted for better utilization of fewer resources (partially through increasing some data structures and other elements that feed execution units) vs. simply throwing more transistors at the problem. AMD also opted to reduce the ratio of integer to FP resources within the x86 portion of its architecture, clearly to support a move to the APU world where the GPU can be a provider of a significant amount of FP support. Piledriver doesn’t fundamentally change any of these balances. The pipeline depth remains unchanged, as does the focus on pursuing higher frequencies.

Fundamental to Piledriver is a significant switch in the type of flip-flops used throughout the design. Flip-flops, or flops as they are commonly called, are simple pieces of logic that store some form of data or state. In a microprocessor they can be found in many places, including the start and end of a pipeline stage. Work is done prior to a flop and committed at the flop or array of flops. The output of these flops becomes the input to the next array of logic. Normally flops are hard edge elements—data is latched at the rising edge of the clock.

In very high frequency designs however, there can be a considerable amount of variability or jitter in the clock. You either have to spend a lot of time ensuring that your design can account for this jitter, or you can incorporate logic that’s more tolerant of jitter. The former requires more effort, while the latter burns more power. Bulldozer opted for the latter.

In order to get Bulldozer to market as quickly as possible, after far too many delays, AMD opted to use soft edge flops quite often in the design. Soft edge flops are the opposite of their harder counterparts; they are designed to allow the clock signal to spill over the clock edge while still functioning. Piledriver on the other hand was the result of a systematic effort to swap in smaller, hard edge flops where there was timing margin in the design. The result is a tangible reduction in power consumption. Across the board there’s a 10% reduction in dynamic power consumption compared to Bulldozer, and some workloads are apparently even pushing a 20% reduction in active power. Given Piledriver’s role in Trinity, as a mostly mobile-focused product, this power reduction was well worth the effort.

At the front end, AMD put in additional work to improve IPC. The schedulers are now more aggressive about freeing up tokens. Similar to the soft vs. hard flip flop debate, it’s always easier to be conservative when you retire an instruction from a queue. It eases verification as you don’t have to be as concerned about conditions where you might accidentally overwrite an instruction too early. With the major effort of getting a brand new architecture off of the ground behind them, Piledriver’s engineers could focus on greater refinement in the schedulers. The structures didn’t get any bigger; AMD just now makes better use of them.

The execution units are also a bit beefier in Piledriver, but not by much. AMD claims significant improvements in floating point and integer divides, calls and returns. For client workloads these gains show minimal (sub 1%) improvements.

Prefetching and branch prediction are both significantly improved with Piledriver. Bulldozer did a simple sequential prefetch, while Piledriver can prefetch variable lengths of data and across page boundaries in the L1 (mainly a server workload benefit). In Bulldozer, if prefetched data wasn’t used (incorrectly prefetched) it would clog up the cache as it would come in as the most recently accessed data. However if prefetched data isn’t immediately used, it’s likely it will never be used. Piledriver now immediately tags unused prefetched data as least-recently-used, allowing the cache controller to quickly evict it if the prefetch was incorrect.

Another change is that Piledriver includes a perceptron branch predictor that supplements the primary branch predictor in Bulldozer. The perceptron algorithm is a history based predictor that’s better suited for predicting certain branches. It works in parallel with the old predictor and simply tags branches that it is known to be good at predicting. If the old predictor and the perceptron predictor disagree on a tagged branch, the perceptron’s path is taken. Improving branch prediction accuracy is a challenge, but it’s necessary in highly pipelined designs. These sorts of secondary predictors are a must as there’s no one-size-fits-all when it comes to branch prediction.

Finally, Piledriver also adds new instructions to better align its ISA with Haswell: FMA3 and F16C.

Improved Turbo, Beefy Interconnects and the Trinity GPU
Introduction and Piledriver OverviewImproved Turbo, Beefy Interconnects and the Trinity GPUMobile Trinity LineupMeet the AMD Trinity/Comal PrototypeAMD Trinity General PerformanceAMD Trinity Gaming PerformanceAMD’s Heterogeneous Computing with TrinityAMD Trinity: Battery Life Also ImprovedConclusion: What Makes a Trinity?

Tweet

PRINT THIS ARTICLE

Review: AMD A10-5800K Trinity APU — CPU

Mobile first, desktop second

In a wide-ranging interview with AMD vice president of desktop marketing, Leslie Sobon, HEXUS learnt that ‘mobility’ is the buzzword permeating the boardrooms of the Austin-based chipmaker. It can be argued that this focus on mobility has been a key driver in AMD pushing out its next-generation accelerated processing unit (APU), codenamed Trinity, into laptop computers first.

The Trinity APU made its notebook debut back on May 15 this year and, despite AMD’s bombastic enthusiasm at that time, we’ve yet to see this new CPU-and-GPU technology gain a steadfast foothold into a market dominated by processors from arch-rival Intel. Underscoring the chasm between technology and marketing, Dell’s UK site doesn’t even enable users to filter laptops by AMD processors.

Now treading on ground more familiar to enthusiasts and technology influencers, almost five months after the initial release, AMD is bringing this same Trinity APU goodness to the desktop. The premise here is to offer mainstream CPU performance and class-leading integrated graphics from a single chip that’s priced at mass-market (sub-£100) levels.

But AMD’s been down this all-in-one chip route before, right? Just over a year ago the firm launched the first-generation desktop APU family, codenamed Llano, and we surmised that while the competent CPU cores were based on old technology, the chip’s graphics performance, taken from AMD’s Radeon division, was in a different, better league to Intel’s. The first question you’ll likely ask is how is Trinity different to Llano? We’re glad you did. The following table illustrates the basic changes on the best APUs from each generation.

Trinity vs. Llano

APU Model

Process

CPU Cores

CPU tech

Max CPU clock

GPU Cores

GPU Tech

Max GPU clock

AMD Turbo Core

Form factor

TDP

Llano

4

32nm

Stars

3.0GHz

400

HD 5000

600MHz

No

FM1

100W

Trinity

4

32nm

Piledriver

4. 2GHz

384

HD 6000

800MHz

Yes, v3.0

FM2

100W

There are at least three important differences between the first- and second-generation desktop APUs. Trinity uses AMD’s Piledriver CPU technology instead of the older Stars architecture of Llano. Equipped with fundamentally higher frequencies — 4.2GHz (Turbo) vs. 3.0GHz — one could assume that Piledriver’s performance is markedly better. This assumption is often incorrect, as the new CPU design, which takes its guts from the much-maligned Bulldozer architecture, isn’t as potent as it may seem at first glance.

Trinity recovers nicely by integrating a GPU that’s based on discrete Radeon HD 6000-series tech and AMD takes the opportunity of clocking it in higher than Llano’s HD 5000-series-derived GPU, too. Perhaps most importantly of all to system builders and do-it-yourself enthusiasts, AMD’s move from Llano to Trinity, while based on the same 32nm manufacturing process, is divergent enough to require a new socket, FM2, and, consequently, a new motherboard.

For the tech aficionados out there, desktop Trinity’s 32nm slab of silicon measures 246mm² and weighs in at 1.30bn transistors, compared to Llano’s 228mm² and 1.18bn. There’s really not much in it from a manufacturing-cum-economics point of view.

This section provides at-a-glance differences between APU generations. We’ve previously covered Trinity’s architecture in far more detail right over here, so do trouble those brain cells if you need further insight into just how it works.

The key takeaways are that while AMD’s Trinity desktop APU appears to be the same as Llano when literally viewed from an entirely superficial level, an overhaul of the CPU and revamp of the GPU portion means they couldn’t be much more different inside.

Tech bores me, which models can I get?

APU Model

CPU Cores

CPU Base Clock

CPU Turbo Clock

GPU Brand

GPU Cores

GPU Base Clock

L2 Cache

Max. DDR3

TDP

Price

A10-5800K

4

3.8GHz

4.2GHz

HD 7660D

384

800MHz

4MB

1,866MHz

100W

£100

A10-5700

4

3.4GHz

4.0GHz

HD 7660D

384

760MHz

4MB

1,866MHz

65W

£100

A8-5600K

4

3.6GHz

3.9GHz

HD 7560D

256

760MHz

4MB

1,866MHz

100W

£80

A8-5500

4

3. 2GHz

3.7GHz

HD 7560D

256

760MHz

4MB

1,866MHz

65W

£80

A6-5400K

2

3.6GHz

3.8GHz

HD 7540D

192

760MHz

1MB

1,866MHz

65W

£55

A4-5300

2

3.4GHz

3.6GHz

HD 7480D

128

723MHz

1MB

1,600MHz

65W

£45

AMD is introducing at least six Trinity APUs today, promising plenty of stock at your favourite retailer. Technology has changed from one generation to the next, clearly, but branding remains much the same. AMD divides desktop Trinity into A10, A8, A6 and A4 families and plays around with speeds, cores and power TDPs in an effort to differentiate models.

All APUs feature on-the-fly overclocking for the CPU assuming there is power budget to do so. K-series chips are unlocked, meaning that the user can increase frequencies/multipliers without impediment. AMD touts bi-directional auto-overclocking, known as Turbo Core 3.0, where either the CPU or GPU’s speed is increased in response to a particular application’s needs, usually achieved by marginally downclocking the portion of the APU that’s not under significant load. However, jumping ahead of ourselves for a moment, our sample’s GPU would not boost past 800MHz in any game.

Carrying on from the previous generation, select discrete Radeon cards can be paired with the on-chip GPU for what AMD terms Dual Graphics. Further tapping into the power of the onboard GPU, AMD enables multi-monitor Eyefinity gaming. Common sense dictates that such Eyefinity-enabled titles will need to run at low-ish resolutions and image-quality settings; an APU’s graphics aren’t nearly as potent as a discrete mid-range card’s.

Our pick of the sextet is the A10-5700; it nearly matches the range-topping A10-5800K in absolute performance yet ships with a significantly lower TDP of 65W. The quad-core APUs carry 4MB of L2 cache, split into 2MB chunks for each Piledriver module — which, remember, consists of two execution cores.

Pricing, too, is keen, with these APU costing between £45 and £100. This is perhaps more of a testament to Intel’s perceived performance hegemony than AMD’s desire to manufacture value-focussed APUs. AMD would dearly love to charge substantially more than £100 for its best-ever desktop APU.

Bah, humbug, a new motherboard?

Yup, FM2-type APUs simply don’t fit, or run, on older FM1 boards used for Llano chips. Investment in a Trinity APU means laying down some more cash for a new motherboard, and here are your choices:

Unashamedly taken from the AMD press deck, this slide details the A55, A75, and A85X chipsets in perfunctory fashion. We find it slightly perplexing that AMD has chosen to keep the same model branding as on Llano boards, and having socket-incompatible A75 ‘FM1’ and A75 ‘FM2’ co-exist for a while will doubtless confuse customers. Whatever the case, ASUS, Gigabyte MSI are to launch a raft of supporting boards.

First-page summary

We’ve learned that AMD is finally bringing its next-generation APU, dubbed Trinity, to the desktop. It carries the same overarching x86 CPU and Radeon GPU architecture of first-generation Llano but supplants both with newer iterations of the technology.

Set against the backdrop of Intel improving its Core processors with the use of Ivy Bridge technology and promising even-greater improvements with next year’s Haswell chips, AMD needs desktop Trinity to sell well. Let’s now subject the range-heading A10-5800K to some stringent competition from both AMD’s previous-generation Trinity and two iterations of Core i3. Who wins out?

AMD Trinity Review | Eurogamer.

net


When it comes to buying a gaming laptop, the accepted wisdom up until this point has been simple; integrated graphics processors are well-suited for playing the classics of yesteryear, but venturing into modern territory — games like Battlefield 3 or Skyrim — requires a dedicated, power-hungry GPU: in other words, a serious card for serious gaming. Nevertheless, Intel and AMD’s integrated graphics cores continue to gain traction in the laptop market for being cheaper, and more energy efficient than dedicated graphics cores. But for many, these integrated processors simply can’t escape the stigma of being the budget choice for gaming, dismissed completely out of hand. The thing is, we’ve now reached the point where even IGPs can put up a strong fight against the current generation HD consoles — and in some cases, even surpass them.


Cue the long-awaited arrival of AMD’s second generation Trinity architecture. This revision sets out to defy common preconceptions of what’s possible with integrated graphics, and perhaps more importantly, to make a stand against Intel’s popular Ivy Bridge processors and HD4000 graphics chips. This is a bold promise indeed, but given AMD’s concurrent position in markets for both CPUs and graphics cards, the potential for it to capitalise here is huge. For the sake of demonstration, we’ve been sent a prototype laptop by the company to showcase the strengths of its hybrid processor in a practical sense. While this particular unit is unlikely to ever see an actual retail release, the Trinity processor is slowly trickling its way into laptops via other brands, with the most notable entry in the UK market so far being Samsung’s NP355V5C notebook.

«Integrated graphics cores are no longer the gameplay jokes they used to be: both AMD Trinity and Intel HD4000 can run Battlefield 3 and Skyrim beyond the console standard.»


So what’s all the fuss about? The Trinity name is more than for show; it refers to a three-way division of the processing unit that should in theory improve its manufacturing costs and power consumption. First up, there’s the A10-4660M CPU, which operates with the updated Piledriver architecture that recently succeeded the Bulldozer series for desktops. This features two sets of two physical cores — essentially represented as a quad core in the Windows task manager, and all cores are clocked at 2695MHz, with the ability to turbo to higher speeds depending on heat and load.


Nearby on the die is the integrated Radeon HD 7660G graphics chip, which has an architectural foundation in the desktop HD 6900 Northern Islands series. This runs at a base core clock of 497MHz that can stretch upwards to 686MHz depending on thermal overhead [Correction: Trinity runs off system RAM and has no attached GDDR5]. Meanwhile, the third part is represented by a new video processing engine, plus multi-display drivers that have their roots in the HD 7900 Southern Islands series. Taken as a whole, this chip is produced at a 32nm manufacturing process with an overall thermal design power (TDP) of just 35 watts. This is AMD’s rebuttal to Intel’s Ivy Bridge range of i5 and i7 CPUs — but is it enough?


Dissecting the Trinity Laptop


Before we jump into the tests, let’s take a look at the kit that surrounds it to get an idea of the nature of laptops it might appear in. The Trinity laptop is notably thinner than our other test unit, the Alienware M14x, which is no doubt buffed outwards by its support of a dedicated NVIDIA GT 650M. Nevertheless, both sport matching sized 14.1 inch LCDs, which appear to be utilising similar twisted nematic (TN) display technology, judging by their unconvincing black levels and poor viewing angles. Resolution for the Trinity laptop is the standard 1366×768 too — potentially a good match for the graphical power it houses.

AMD Trinity Prototype Specs

Few laptops are available at the moment with the Trinity APU, so AMD provided us with a reference 14-inch laptop containing the necessary hardware. Here’s an outline of the kit contained in the non-descript casing.

  • Processor: AMD A10-4660M Piledriver quad core processor at 2695MHz
  • Integrated GPU: AMD Radeon HD 7660G (334MHz Core Clock, 667MHz Memory Clock)
  • Display: 14.1-inch LCD, 1366×768
  • Memory: 4GB DDR3 RAM
  • Storage: Samsung 120GB SSD (830 series)
  • Optical Drive: Philips BD-RW optical drive
  • OS: Windows 7 Premium 64 Bit

Caption

Attribution


The unit is stylised in a black and silver motif, and is sturdily built overall with a matte finish that even extends to the coating on the screen. There is no perceptible graininess to the texture of the display as a result when turned on, however, and response times are expectedly solid for the LCD technology in use here, making it an ideal candidate for gaming. That being accepted, there is some noticeable banding on colour transitions with this screen, which can be especially outstanding when lowering and raising the brightness settings with a vivid picture in the background.

«AMD’s Trinity prototype achieves over nine hours of battery life for standard web browsing and offered us a creditable three hours for StarCraft gameplay — impressive stuff.»


The keyboard offers a satisfying level of give to each tap, and each press results in a light click. Unlike some larger laptops, there’s nothing in the way of flex when the centre is pressured. Below the board is a smoothed trackpad which feels proportionately sized for the device, while above it is a touch-sensitive media bar with glowing LEDs behind each button to show when volume is being raised, lowered, or muted. In terms of raw connectivity, we turn to the edges of the device to find two USB 3.0 slots, one for USB 2.0, outputs for both VGA and HDMI, one eSATA connector, an Ethernet port, and a BD-RW drive. There are also standard 3.5mm headphone and microphone jacks at the front, centimetres besides a toggle for WiFi.


Switching it on solves the mystery of its surrounding specs. In support of the Trinity APU is 4GB of DDR3 RAM clocked at 1333MHz, and a 120GB Samsung SSD. While the latter seems like it’s little on the restrictive side, given that 30GB PC games like Rage exist and are growing in number, it does at least allow for fast streaming game levels and textures. Battery life is also improved by the removal of moving, mechanical parts we’d find in conventional HDDs, which is tantamount to one of AMD’s end goals.


On that note, we’re glad to see these measures paying off; for our tests involving strictly light use, our stopwatch reads nine hours and 40 minutes. This decent turnout is achieved on its stock 6-cell battery supply with both screen brightness and speaker volume set to 50 per cent with WiFi engaged. Otherwise, using the very same settings while playing StarCraft 2 at medium settings gives us just a few minutes over three hours of play. Considering that this RTS only takes advantage of a maximum of two physical cores, this figure may be lower for titles like Battlefield 3 where all of the Trinity cores will be engaged. Fortunately, recharging the laptop back up to 100 per cent capacity takes just one hour and 30 minutes overall.

Alienware M14x R4 Specs


The review unit supplied by Alienware tops out most of the selectable kit — quad-core CPU, top-end GPU for the range and a 1600×900 screen. This haul is approximately twice the cost of laptops using the AMD Trinity solution, but since this is an integrated GPU shoot-out, obviously we are running with the dedicated chip disabled.

  • CPU: Intel Core i7 3720QM CPU at 2. 6GHz
  • Graphics Core: Intel HD4000 Integrated, NVIDIA GeForce GT 650M with 1GB GDDR5 (disabled in these tests)
  • Display: 14.1-inch Alienware WideHD+ WLED, 1600×900
  • Memory: 8GB DDR3 RAM
  • Storage: Samsung 64GB SSD (mSATA) and Seagate 500GB HDD
  • Optical Drive: DVD-RW
  • OS:Windows 7 Premium 64 Bit

Caption

Attribution


Another major consideration for portable hardware is heat and noise. Running the Hardware Monitor program here gives us no readings from any of the sensors linked to the Trinity processor, but in practise during gaming sessions, we find this gets nowhere near the toasty levels that the M14x hits. The fan is certainly audible on the Trinity under load, but ultimately the bigger distraction is the screeching capacitor whine that occurs whenever movements are made on screen. This manifests as a dull whir that emanates from inside the laptop at the slightest movement of the mouse cursor — hopefully a bug exclusive to our unit, and not a characteristic of the final retail units that will use the APU.


On the plus side, the positioning of the fans means there isn’t much in the way of hot air being directed towards your legs. In conjunction with the generally low heat output, this means it’s actually possible to use this as a laptop in the most literal sense.


Performance Analysis: Benchmarks


Moving onto the benchmarks, we pit the Trinity APU directly against the closest rival we have to hand: the fourth revision of Alienware’s M14x gaming laptop. As a 14.1 inch platform based on the Ivy Bridge architecture, it happens to offer an Intel Core i7 with eight threads and four physical cores, plus the integrated HD 4000 graphics processor we need for testing. Naturally, we switch off the installed discrete NVIDIA GT 650M graphics chip, and simply compare their integrated graphics solutions directly for each benchmark.


First up are the synthetic tests. The latest 2011 version of 3DMark offers us a yardstick based around the strengths of DirectX 11, giving us an idea of how games can perform once advanced lighting and tessellation are involved. Four demanding jungle-themed scenarios play out here that guage graphical capability, followed by one for physics, involving towers tumbling like dominoes, and one last sequence which condenses all of these tests into one torturous stress-test.

Futuremark 3DMark 11: Tested on standard Performance and Extreme settings.

«3DMark 11 scores suggest that AMD Trinity offers a substantial graphical advantage over its expensive Intel rival but that neither has the necessary power for taxing 1080p performance on the most challenging titles.»

AMD A10-4660M (Perf) Intel i7 3720QM (Perf) AMD A10-4660M (Extreme) Intel i7 3720QM (Extreme)
Graphics Score 1012 679 340 201
Physics Score 2659 6909 2646 6841
Combined Score 932 802 373 278
3DMark 11 Score P1105 P799 X376 X229


In the standard Performance test we see both integrated graphics processors rendering at 1280×720 resolution, while the Extreme preset ratchets that up to a full 1920×1080. It quickly becomes evident that both AMD’s 7660G and Intel’s HD4000 are out of their league on this higher setting, with the entirety of the test playing out as an hour-long slide-show when we try. In the Performance tests, however, we see much smoother rate of rendering that sees Trinity commanding around a 100 per cent lead in score over the HD4000 in the final Combined test.


That said, Intel’s solution does manage to sail to the horizon with a huge lead in the physics-based scores. The scalability of its quad core processor is accountable for this, where its connection to eight active threads may be advantage enough, as compared to just the four on AMD’s chip. Intel’s advantage over AMD in CPU design is well-known, and this particular processor is one of the best mobile solutions it offers — and it’s pricier too, so this disparity is not unexpected.


The overall lead Trinity has here is significant though, and shows promise for performance leads in games, up to and including those using the prevalent Unreal Engine 3. As usual, we opt for Batman Arkham City’s in-game benchmark option to get an idea of how both processors perform in either DirectX 9 and 11 rendering modes, with each in combination with FXAA High, 4x MSAA and 8x MSAA. Let the tests begin…

Batman: Arkham City Benchmark: DX9, 1366×768, high detail, PhysX off, all other settings enabled.

AMD A10-4660M (FXAA High) Intel i7 3720QM (FXAA High) AMD A10-4660M (4x MSAA) Intel i7 3720QM (4x MSAA) AMD A10-4660M (8x MSAA) Intel i7 3720QM (8x MSAA)
Average FPS 34 36 27 23 25 15
Min FPS 15 14 11 3 7 0
Max FPS 43 44 36 30 34 22


The benchmark is run here on high detail across the board, with ambient occlusion checked too. It flies us through a few of the game’s major locales with a fixed camera, taking us to a goon-filled lair featuring a moving light sources, the entrance to Poison Ivy’s botanical den, and the sky-scraping peaks of Arkham City itself.

«The Batman: Arkham City benchmarks reveal that gameplay performance on some releases between the rival integrated GPUs may not be as far apart as synthetic tests suggest.»


Here we see the gap widens significantly in favour of the Trinity APU for anything involving the older DirectX 9 version of the test, and this distance in results is only accentuated with each bump upwards in the AA settings. Average frame rates are very close for tests involving High quality FXAA here too, and we find this to inevitably be the most judder-free approach to playing the game, short of lowering the global quality settings to medium.


To follow this up, we run through the same gamut of tests with DirectX 11 enabled. We’re initially surprised to find the average frame rates level out at 21FPS for both on FXAA High, where the peaks and troughs are also uncannily close to identical. We notice more screen tearing during these bouts of testing, however, with the bulk of the cuts occurring specifically over light sources

Batman: Arkham City Benchmark: DX11, 1366×768, normal tessellation, high detail, PhysX off, all other settings enabled.

AMD A10-4660M (FXAA High) Intel i7 3720QM (FXAA High) AMD A10-4660M (4x MSAA) Intel i7 3720QM (4x MSAA) AMD A10-4660M (8x MSAA) Intel i7 3720QM (8x MSAA)
Average FPS 21 21 10 15 5 11
Min FPS 3 5 3 2 1 2
Max FPS 29 29 12 21 7 14


Turning on MSAA settings here gives us the complete opposite turnout to that of DX9, and Intel’s CPU takes full charge. Frame rates drop dramatically to under playable levels, with AMD’s Trinity suffering an especially big hit when sampling is ramped up to 8x. This situation would no doubt have benefited from adding dedicated GDDR5 RAM to AMD’s architecture, and extra features such as ambient occlusion and tessellation drag performance down to a choppy 5FPS average.


To double check these issues, we turn to another benchmarking favourite: Metro 2033. In this case, DirectX 11 is selected alongside medium settings, plus with the depth of field check-box ticked. Obviously, PhysX is also disabled here too.


The testing process involves three fly-by passes through an abandoned train station littered with enemies, fire, specular effects and lighting sources. Put into motion, the game at these settings appears ostensibly unplayable for both graphics processors at even these modest settings, regardless of whether we enable 4x MSAA, or 4A Games’ modified AAA alternative.

Metro 2033 Frontline Benchmark: DX11, 1366×768, medium detail, PhysX disabled.

«The Metro 2033 scores show that the most demanding benchmark tests can still bring these modern integrated GPUs to their knees, with neither acquitting itself well.»

AMD A10-4660M (AAA) Intel i7 3720QM (AAA) AMD A10-4660M Intel i7 3720QM (MSAA)
Average FPS 12.33 14.37 8.52 10.38
Min FPS 6.60 10.77 5.16 7.81
Max FPS 16.64 19.42 10.64 16.51


This time the lead is much more pronounced for Intel’s HD4000 processor, and we see it consistently average out at 2FPS or higher than AMD’s 7660G. The addition of 4x MSAA cripples performance for each to the same extent, and we see values dropping below 10 even in the less strenuous areas on AMD’s processor. It’s not a pretty picture, but this benchmarking run is notorious for bringing graphics hardware from all vendors to its knees: max out the settings and even the GTX 680 will struggle to reach 30FPS at 1080p.


Of course, there’s a limit to how much raw numbers can relay when it comes to the overall gaming experience. So, as with our Face-Off features, we losslessly capture video of the Trinity laptop playing the most demanding games, and feed that through our frame-rate analysis tools to get a better view of where the laptop falls short and where it excels. As always, Battlefield 3, Crysis 2 and The Elder Scrolls V: Skyrim form our standard gamut of tests, and in each case we match footage with that of the M14x running on the IGP, along with the Xbox 360 and PS3 versions in order to give a console performance perspective. Perhaps we’re asking a bit too much here, but we also engage v-sync on the PC comparisons, as screen-tear is so distracting on console.


Gameplay Analysis: Battlefield 3


Jumping in at the deep end, we find running Battlefield 3 at the Trinity laptop screen’s native 1366×768 resolution to be a write-off for sustaining 30FPS on higher-end settings levels. Similarly, the medium graphical preset hinders us in hitting that targeted level of performance. Particularly in the opening sequence of the Operation Swordbreaker level, during the stroll around the military camp with your squad, the only way to achieve this is by running the game at 1280×720 at the Low preset. The only tweak we make over and above this is the raising of the FXAA to medium quality, which greatly improves the look of the game at such a low resolution, and typically only at the expense of 1FPS less on average — worth the sacrifice, in other words.


In effect, these settings are remarkably close to those on 360 and PS3, with the noticeable proximity culling for objects and shadow detail. Bearing in mind that this is the lowest end of the graphical scale possible on PC, we’re surprised to find that some textures are much improved over their console counterparts in places, and filtering is less intrusive — the dedicated 512MB of graphics RAM clearly has its advantages. One sore spot is the lack of filtering on shadow aliasing, though, which crops up in most globally illuminated areas and remains immune to our post processing. It’s with these matching low settings that we have our M14x test the game as well, and the difference in performance is very encouraging.

«Battlefield 3 on low settings with FXAA applied looks better than the console versions in some respects and matches frame-rate is most situations — with v-sync engaged to boot.»

Battlefield 3 shows us how well a DirectX 11 game optimised for multiple cores runs on the Trinity APU. The good news is that the laptop hits the same 30FPS rung as the 360, and generally sustains a lead over the M14x’s Intel processor. The bad news is that settings need to be reduced to low — though asset quality is largely comparable to consoles on this preset.

Additional Video Analyses:

  • Battlefield 3: AMD Trinity vs. Xbox 360
  • Battlefield 3: AMD Trinity vs. PS3


Be it in cut-scene or battle, we see the Trinity laptop take a relatively steady lead over the M14x in Battlefield 3. As mentioned before, we shut off access to the Alienware machine’s NVIDIA GPU, and the game is running solely off the laptop’s Intel HD4000 graphics. Where the frame rate isn’t neck-and-neck, we see a 4-6FPS difference to the AMD chip’s advantage. It’s less clear-cut during battle however, but overall 30FPS is easily sustainable for both platforms at these settings, though the Trinity laptop is more likely to hold out at 34FPS for long stretches.


By comparison to the console versions in these terms, we’re spared the constant tearing, but frame rate is slightly worse during that initial section where your squad gathers. Outside of this, like-for-like moments are identical, and we even see gains on the Trinity laptop side during shoot-outs. This is another encouraging result.


Gameplay Analysis: Skyrim


Swapping Frostbite 2 out for Bethesda’s Creation Engine, we move on to The Elder Scrolls V: Skyim to see how the Trinity APU handles large environments and long draw distances in this massive fantasy world. In keeping with the standard set by Battlefield 3, we set the resolution to 1280×720 and find the setting which best suits our 30FPS target.


For this case, as with all our previous laptop tests, we find the high preset remains a snug fit. The only cause for concern here is the 8x MSAA that’s selected by default, which we swap out for the less processor-intensive FXAA. As is standard, we run our analysis over the very first cut-scene, where your unnamed prisoner character is being transported downhill by horse-cart to Helgen village. This stresses most platforms to a greater degree than the general, free-roaming run of play can, owing largely to the number of NPCs condensed into one area.

«The Trinity APU eclipses both Intel HD4000 and PlayStation 3 in terms of frame-rates, at Skyrim’s high graphical preset — though MSAA is swapped out for the cheaper FXAA.»

Skyrim’s opening sequence shows us the biggest gains for the hybrid AMD architecture. In this case, we can play the game at high settings with FXAA enabled, and it still takes a lead over the M14x and PS3 playthroughs.

Additional Video Analyses:

  • Skyrim: AMD Trinity vs. Xbox 360
  • Skyrim: AMD Trinity vs. PS3


The results show an even more consistent lead for the Trinity APU over the M14x’s Intel i7 processor than in Battlefield 3. The 30FPS target-line wavers in and out of reach for both laptops, depending on the segment of the journey, but overall it’s AMD’s chipset which comes out on top for most of the trip. This is can be by a matter of 4FPS, though they both average out at the target frame rate once settled in the town.


Unfortunately, performance simply doesn’t quite holds its own against the 360 version, which locks itself to 30FPS for much of the route downwards, albeit with tearing. However, Trinity does command a marginal lead over PS3 for the most part.


In all, the results are to the credit of the AMD’s laptop, particularly when considering how much further afield draw distances for objects, geometry and NPCs are on the selected high preset. Running off the SSD also allows textures to stream in more immediately, where we notice the rocks on the walls for the PS3 version popping in when we’re mere inches away.


Gameplay Analysis: Crysis 2


Last but not least, we put Crysis 2 through its paces across all formats. In keeping with our previous tests, and to better match the resolutions of the console versions, we set the resolution to 1280×720 and roll the settings all the way down to the lowest preset, bewilderingly named ‘high’.


As our previous benchmarks have already highlighted, the Trinity APU’s difficulty in handling rendering techniques such as tessellation remains an issue here. The game can’t be enjoyed at anything close to a consistent 30FPS during the prologue mission, and so we simply find ourselves turning this off to preserve playability. Similarly, the high resolution texture pack is disabled, although not by choice — it’s simply blanked out by default due to the lack of dedicated RAM.

«The integrated GPUs struggle with Crysis 2’s effects work and we had to drop down to 1024×600 in order to achieve a consistent 30FPS in all gameplay scenarios.»

Crysis 2 on Trinity laptop, versus the M14x, PS3 and 360. On the high preset, performance on AMD’s integrated graphics processor is comparable to the PS3’s throughout the opening section, but falls short of the 360 in most cases. For our tests, we run without DirectX 11 features or high resolution textures enabled.

Additional Video Analyses:

  • Crysis 2: AMD Trinity vs. Xbox 360
  • Crysis 2: AMD Trinity vs. PS3


To best test CryEngine3, we run through the opening submarine sequence, which gives us a faceful of particle and water effects coupled with the presence of multiple allies up-close. This inevitably proves to be the weakest point during analysis of the Trinity laptop, which only affords itself a refresh of 20FPS for much of the swimming sections. This proves to be quite comparable to the PS3’s performance here overall, though the 360 has a much more convincing stab at holding the full 30 here — though once again, Microsoft’s platform is liable to tear frames occasionally near the top of the screen.


The situation turns around for the better once we hit the streets of New York, where we see the AMD machine ably hitting that same target, and matching consoles platforms. In comparison to the M14x, results are mixed: the myriad of effects during the submarine collision gets the better of both laptops, but Alienware’s 14-incher clears around 2FPS on average over its rival. In later sections the roles are reversed, meaning that, by and large, it’s close enough to call it a tie.


Of the three games subjected to the spotlight, Crysis 2 shows the least flattering results. As proved to be the case with the game on other mobile platforms, we find the best way to improve the standing of its performance is to lower the resolution. Sure enough, dropping the values down to the netbook-standard of 1024×600 has a huge impact, and brings the frame rate back into comfortable 30FPS territory for almost every scenario.


Overall, it’s a shame that we have to make practically every concession possible to get Crytek’s game into a playable state on the Trinity laptop, but on the plus side, the visuals are still impressively detailed even at this ground-level preset.


AMD Trinity: The Digital Foundry Verdict


All in all, the improvements made to overall performance in AMD’s latest revision of the Trinity architecture are impressive, but not quite as drastic as we’d hoped. Yes, it has the edge in our synthetic tests with 3DMark 11, but in our gameplay tests, the results are much closer to Intel’s HD4000 than we would have expected. Now, the quad core i7 is considerably more expensive than Trinity — AMD typically excels in price vs. performance contests, so we also ran all of our gameplay analyses with two cores disabled on the Intel platform, in order to approximate a dual core chip running the HD4000 GPU — something a bit more comparable price-wise to Trinity. It made little difference to the Skyrim and Crysis 2 performance levels, but Battlefield 3 really struggled, dipping below 20FPs. Generally speaking, games prefer stronger GPUs and most titles still target dual core systems, but Battlefield 3 is clearly an exception.

«Trinity proves that you can enjoy top-tier games on an integrated graphics core but our feeling is we’re a year or so away from IGP performance that comprehensively bests current-gen consoles.»


Regardless, the fact that it was so close surprised us. AMD’s big response to Intel’s current generation of processors succeeds in immediate terms, but its R&D departments will need to dig much deeper in order to compete with the upcoming Ivy Bridge successor, dubbed Haswell — rumoured to bring a 2x-3x performance boost over the current hardware. Its handling of DirectX 11 titles also needs addressing, as the current revision came across as somewhat under-optimised in many of our benchmark tests.


The silver lining is this though; the Trinity APU is capable of convincingly holding its own against the console standard. For games like Battlefield 3 and Skyrim at least, achieving a 720p30 output at graphical presets in a very similar ballpark to these versions is definitely a reality for this small processing unit — not a result to be scoffed at given its excellent battery life, low rated TDP and heat output. Crysis 2 is the only real odd-one-out here, where the laptop struggled to maintain a 30FPS refresh, and lacked the resources to run the higher resolution textures.


So what of Trinity’s future? Given that the scant selection of laptops currently using the architecture cost just shy of the £600 mark, such as the Samsung NP355V5C, there’s no doubt that it’s a cheaper option than most laptops equipped with the high-end Intel Core i7 CPU we use for testing. It’s also encouraging to see the laptop fulfil its promise of improved energy efficiency; it tallies up a reasonable three hours of battery life while gaming, and almost ten during light web browsing.


The only complaint here isn’t usability, but its cost in the grand scheme of the market. Put into perspective, the Acer Timeline Ultra M3 we reviewed recently can be purchased for a similar price, and features a dedicated NVIDIA GT 640M GPU for vastly improved gaming performance. Alas, dedicated cards continue to be the most cost-efficient conduit for modern gaming on a laptop, and for AMD, its hybrid processor needs to be either cheaper, more powerful — or preferably, both — to be truly compelling.

AMD Trinity (A10-4600M) Review | Trusted Reviews

Verdict

AMD has long played second fiddle to Intel when it comes to providing the chips that power our laptops. Plain and simply, Intel has nearly always offered a better balance of battery life and performance. Last year AMD unveiled its codenamed-Llano chips, which packed in class leading graphics performance but still trailed when it came to real world battery life and CPU performance. Now the company is back with its latest codenamed-Trinity chips, and the claims as to its abilities are bold indeed. Not only should its Radeon HD 7000 graphics comfortably beat both Intel’s existing ‘Sandy Bridge’ based chips and latest ‘Ivy Bridge’ mobile range but also it should have great battery life and CPU performance too.

We’ve spent some time with a test laptop based on the new AMD A10-4600M chip to see just how it stacks up in real world usage. We’ll be putting it to the test later on but first a few more details about Trinity.

AMD Trinity Architecture
AMD Trinity is the codename for the company’s new range of what it refers to as its APUs (Accelerated Processing Units). AMD uses this terminology because unlike CPUs of old, these new chips don’t just house the main processor but the graphics processor (which itself can be used for computing tasks other than graphics), memory controller and a number of dedicated units for speeding up things like video decoding. 2 with 1.4billion transistors.

Trinity is built using AMD’s existing 32nm manufacturing process, just as with its Llano chips but despite this it has managed to double performance-per-watt while both CPU and GPU performance increase is in the double digits. This 32nm is roughly an equivalent generation to Intel’s Sandy Bridge but behind its just-about-arriving-now Ivy Bridge chips which use a smaller and more power efficient 22nm process.

AMD Piledriver
The beating heart of Trinity is the Piledriver CPU core, which replaces the Stars based design of Llano. Direct comparison is a little tricky as each Piledriver ‘module’ contains what is roughly equivalent to two normal cores, though not quite. The chips will come with either one or two modules making for what are essentially either dual or quad core chips.

We won’t dive too deep into the architecture here (Anandtech has done a great deep dive of the tech if you’d like to read more) but the key is that it’s an almost completely new design, which offers both performance and power saving benefits. However, AMD readily admits it still can’t compete with Intel for raw CPU performance for any given Thermal Design Power (TDP – the maximum heat/power the chip is designed to output), and is relying on its chips delivering ‘enough’ performance in this area while backing it up with excellent graphics and media acceleration performance, and good battery life (low power usage).

It’s a sound logic and one that certainly reflects the usage pattern of our times. After all, most people spend the majority of their laptop time performing relatively idle tasks like web browsing, writing and watching video, not performing intense calculations or doing heavy multitasking. It’s often only gaming that really calls upon a CPU/graphics chip to pull its weight. We’ll talk more on this point a bit later though.

Buried within the silicon there are a whole host of other tweaks and improvements that should result in lower power usage and better performance but one of the more prominent is the new AMD Turbo Core 3. 0 technology.

Turbo Core existed on Llano where it dynamically adjusted the clock speed of the CPU depending on workload, to get the most performance when needed without overheating the system. But it only worked on the CPU, not the GPU. With Turbo Clock 3.0 the chip can increase or reduce the clock speed of both as and when required. The algorithm for doing this is also much improved so that the chip really does eek out every last morsal of performance where it can.

On the flip side, in idle moments the chip can also power down almost completely to extend battery life, indeed AMD is claiming superior idle battery life to Intel.

AMD Radeon HD 7000 Graphics
Key to the appeal of this new AMD trinity range is its graphics capabilities. Each model in the range has a slightly different configuration, unlike Intels’ Sandy Bridge (Intel HD 3000) and Ivy Bridge (Intel HD 4000) chips which largely use identical configurations within each range. Confusing though this makes it, one thing should hold true, which is that we expect to see most of these AMD configurations beat both Intel Sandy and Ivy Bridge equivalents. We shall have to wait until the lower power products come to market before we can test this though. Also, whether this results in truly playable framerates in demanding games, we shall see, but at least less stressful titles should perform better.

What makes these parts potentially really exciting, though, is that they can be partnered with extra graphics chips from AMD’s 7000 series mobile graphics range. If configured in the same system the two can work together, just like Crossfire/SLI in a desktop system. We haven’t had a chance to test this here but figures provided by AMD show a near doubling of performance over the APU alone. Meanwhile, when the system is idle the extra GPU should consume almost no power so should have minimal impact on performance. Again, all this we’ve yet to test, though, and there’s no guarantee we’ll see genuinely thin and light laptops with this sort of power.

As to the GPU itself, it’s very much similar to AMD’s latest desktop graphics cards in terms of fundamental design, just with less of everything. At most it will contain 384 cores, with the lower end parts dropping to 192 cores. These will be accompanied by up to 24 texture units and 8 ROPs, making the high-end part very roughly equivalent to ¼ of an AMD Radeon HD 6970 desktop card.

HD Media Accelerator
Rounding out the key features of the new design is the HD Media Accelerator. This is a block of the chip dedicated to speeding up certain tasks like video decoding – something which is supported in all the major browsers, Windows Media Player and VLC – and encoding.

AMD is also touting its OpenCL support, which sees applications such as the above, Photoshop CS6, receive increased performance from intensive tasks being taken on by the GPU rather than the CPU.

There’s also AMD’s Quick Stream technology. This prioritises media-related internet traffic for smooth streaming video playback. Clearly this is largely reliant on your network connection still but if you’ve other apps accessing the internet, this technology will ensure they don’t disrupt your viewing.

AMD A-Series APUs
The new Trinity APUs will be coming to market as the A-series, with models ranging from the top performing A10 to the A6. This is a little confusing as the previous generation chips were also called the A-series, so you’ll have to check out full model numbers to differentiate – Llano used Axx-3000 style numbering, Trinity will use Axx-4000 numbering.

Within the numbering differences there are also some key feature differences. The chips will be available in two main configurations based on the type of mounting package they use: PGA or BGA. The standard set of APUs use the classic Pin Grid Array (PGA) arrangement where the chip is housed on a mounting that has hundreds of pins on its underside to connect to the motherboard. Meanwhile Ball Grid Array (BGA) uses tiny balls of metal instead of pins to make for an even smaller overall package.

Within each type of package there were also be differentiation based on the TDP of the chip, with 35W chips available for the larger PGA package and 25W and 17W for the BGA package. As you might expect, the former will find their way into larger more powerful laptops while the latter will be used in slimmer, even Ultrabook-style form factors.

Each configuration also has slightly different graphics capabilities. The full list of available versions is below.

AMD Trinity Test Platform
So onto the crux of the matter – how do these new chips perform when in an actual laptop?

We were provided with a test laptop running the range topping A10-4600M APU. This had been crammed into a 14in chassis, but far from being a sleek and portable slice of desire, it’s a fairly utilitarian looking largely plastic slab that’s well over an inch thick. Still, it’s only meant for testing the underlying hardware and what’s more within that frame it has a Blu-ray drive, VGA and HDMI video outputs, two USB 3.0 ports, one USB 2.0 port, an ExpressCard slot, a memory card reader, Ethernet and both headphone and microphone jacks. It also has 4GB of RAM and a 128GB SSD. Consider the space you could save taking out the Blu-ray drive, and several of those ports and you could conceivably make a relatively slim machine with this core hardware configuration.

The combination of the high-end chip and SSD means this machine should outperform a great many equivalently sized laptops, and is perhaps a little unreflective of the sort of system you’re likely to encounter but it still gives us a strong indicator of real world performance. Literally powering the system is a 4400mAh battery, which is about typical for this size of machine.

Now, finding a system to compare this one to was a bit tricky as few systems pack quite the same balance of components but we’ve picked what we think is a reasonably representative set. We’ve got a lower power Sandy Bridge chip being used by the archetypal Ultrabook the Asus ZenBook UX31, the a more direct competitor with a faster Sandy Bridge in the Lenovo ThinkPad X220 and finally we’ve chosen a superfast, quad core Sandy Bridge equipped HP Pavilion dv7-6b51ea Beats Edition desktop replacement laptop that also includes a dedicated AMD Radeon HD 6490M graphics card.

We’ll fess up that we simply haven’t been able to source an Ivy Bridge test system for comparison in time for this review but we’ll of course revisit the topic when we get our hands on an Ivy Bridge system. Having perused various other sources, we also have a good grasp on what we’re likely to see for CPU, GPU and to a certain extent battery life from such systems too.

Here’s the full comparison list:

So without further ado, onto the testing!

CPU Performance
First of our tests was the classic Cinebench. It’s a reliable indicator of the pure grunt of a CPU, which fully taxes both single and multi-core processors.

It’s immediately obvious just where the AMD A10-4600M’s deficiencies lie. With a Cinebench CPU score of 2.03pts, it just about matches the low power Sandy Bridge chip of the Zenbook but is utterly trounced by the more powerful HP. As for Ivy Bridge, it should come as no surprise that scores we’ve seen
suggest it pulls out even more of a lead over the HP here, with a scores of 6. 5 being bandied around.

We also ran the general system performance test, PCMark 07, which showed AMD trailing both the Lenovo and Zenbook – the HP’s score is low as it uses a hard drive rather than SSD. Ivy Bridge scores for PCMark07 we’ve seen are in the region of 6500, again showing just how fast that chip can be.

However, the key here is that AMD has indeed judged things correctly in terms of day to day use. Subjectively the A10-4600M feels plenty fast enough, providing nippy web browsing, smooth video playback and ample productivity performance. After all, Ultrabooks like the Asus Zenbook are considered more than usable and the A10 is on par.

GPU Performance
Turning to gaming, we used our classic notebook gaming benchmarks on this and our other comparison systems and the results are striking to say the least. In the relatively undemanding TrackMania Nations Forever test the A10 is neck and neck with the dedicated AMD graphics card on the HP, and holds a comfortable lead over both the Lenovo X220 and Zenbook. But TrackMania, at the settings we run, is so undemanding that it actually becomes limited by the CPU speed and it’s only in Stalker: Call of Pripyat we see the true picture. Here the A10 has a 60 per cent performance advantage over the HP’s AMD Radeon HD6490M graphics card, while the two Sandy Bridge systems are left far behind.

Does this result in a genuinely enjoyable real world gaming experience? Well, it’s a close one. While 40fps in Stalker is impressive, it’s only being run at medium detail settings and a resolution that’s lower than the number of pixels on your average laptop. Performance probably needs to double again before we’re really getting to a gaming-grade machine for even the latest games.

Nonetheless, less demanding favourites such as Counter-Strike: Source, World of Warcraft and Call Of Duty should all prove playable.

Battery Life
Perhaps the biggest shock, at least according to our tests is just how little power the A10-A4600M consumes, or more specifically how long this test platform’s battery life is. Despite housing the most powerful chip in AMD’s new range it comfortably beat both the Zenbook and the HP gaming machine, providing nearly an extra hours use, despite having the smallest battery on test. Only the X220 beat it, but that laptop has a 60 per cent bigger battery.

Our test uses the industry standard MobileMark Productivity test, which simulates a reasonably typical usage pattern of the user editing some documents, browsing the web, creating a powerpoint and watching some video, all interspersed with pauses to simulate the user sitting and having a think. We’ve found it to be a very good indicator of battery life in average use but there are some other extreme use cases where it is less indicative, in particular if you’re watching video – say on a long haul flight – or playing games.

With regards the latter, it simply isn’t sensible to test battery usage as the GPU on any machine will drain power very quickly. However, watching video is something that other machines can do for hours on end, and here Trinity trails slightly, with the test platform providing around 3. 5 hours of h.264 video playback compared to around 5.5 on the Zenbook. But, of course, here the argument comes back to this being a higher power chip, and the lower power Trinitys may still improve on this.

What we can say for certain is that AMD is definitely within touching distance of Intel on the battery life front.

Verdict
There are still key questions to be answered about the AMD Trinity platform, such as how the lower performance parts will fare and what sort of systems we’ll see sporting the new chips. Not to mention how much those systems will cost. But, on the evidence here it would appear AMD has done most of what it can. CPU performance does still trail both Intel Ivy Bridge and Sandy Bridge but crucially is ample for general day to day computing. Meanwhile GPU/gaming performance is class leading and battery life is at least on par. Is this the dawn of a new AMD-dominated era for laptops? Perhaps not, but no longer should we be confined to a choice of one.

Unlike other sites, we test every laptop we review thoroughly over an extended period of time. We use industry standard tests to compare features properly. We’ll always tell you what we find. We never, ever, accept money to review a product.

Find out more about how we test in our ethics policy.

Used as our main laptop for the review period

Tested for at least a week

Used consistent benchmarks for fair comparisons with other laptops

Reviewed using respected industry benchmarks and real world use

AMD Trinity On The Desktop: A10, A8, And A6 Get Benchmarked!

This story, a preview of AMD’s Trinity-based A10, A8, and A6 families, was originally published on June 14, 2012. It was not condoned, supported or sponsored in any way by AMD. The piece appears here, unchanged, with the same information presented nearly four months ago.

About a month ago, AMD took the wraps off of its Trinity-based APU. Hotly-anticipated, all eyes turned to see how the second-generation amalgamation of x86 cores and graphics processing resources performed. Why were so many enthusiasts interested in a decidedly mainstream piece of hardware?

Let’s just say Trinity’s composition is…unique.

AMD’s new APU is the first component sporting its Piledriver x86 architecture. After the disappointment that was Bulldozer, which we first evaluated in FX-8150, hopes had to be pinned on a follow-up. And Trinity has it. Back when we were first briefed on Bulldozer, AMD showed us roadmaps with a new architectural revision pushing 10-15%-higher performance each year. Now, power users want to know if Trinity’s Piledriver-based cores deliver on the company’s promise.

Moreover, Trinity employs a newer graphics architecture than Llano. Instead of the VLIW5 arrangement, which also sat at the heart of Radeon HD 6800 and older GPUs, it utilizes the VLIW4 design that went into AMD’s Radeon HD 6900-series cards. Everything after the 6900s swapped over to Graphics Core Next, so VLIW4 isn’t a very prolific implementation. But it’s supposed to be more efficient. Naturally, then, we all want to see how Trinity’s on-die GPU compares to what came before.

We’re Going Mobile, Mav

There was just one problem with last month’s introduction: it only covered the mobile implementation of Trinity.

That was the right move for AMD, no question. It doesn’t take a page of analysis to figure out how putting a CPU and GPU on the same piece of silicon can help address the physical, thermal, and power-oriented issues that laptop manufacturers have to overcome as they design new products.

But enthusiasts were left with questions. Most obviously, how might Piledriver be expected to behave in an FX-branded device? Does it ameliorate Bulldozer’s weaknesses? Given a similar 100 W ceiling on the desktop and the same 32 nm manufacturing process, does Piledriver/VLIW4 deliver an appreciable benefit compared to Stars/VLIW5?

Answering those questions requires the freedom to tweak and tune around in a motherboard BIOS. So, we got our hands on a trio of Trinity-based desktop APUs and set out to preview their performance.

I say preview because hardware based on the Trinity design isn’t going to be available in the channel until later this year. It has been reported that there are still a lot of Llano-based APUs out there, which AMD needs to sell off. So, it’s making Trinity available to OEMs designing notebooks and desktops in time for back-to-school. But you won’t be able to buy these chips for a while. Moreover, the motherboards supporting them with Socket FM2 interfaces aren’t fully-baked, either.

Meet The Desktop Trinity Line-Up

Radeon HD GPU (MHz) Shaders TDP Cores Base CPU (GHz) Turbo Core (GHz) L2 Cache Unlocked
A10-5800K 7660D 800 384 100 W 4 3.8 4.2 4 MB Yes
A10-5700 7660D 760 384 65 W 4 3.4 4.0 4 MB No
A8-5600K 7560D 760 256 100 W 4 3. 6 3.9 4 MB Yes
A8-5500 7560D 760 256 65 W 4 3.2 3.7 4 MB No
A6-5400K 7540D 192 65 W 2 3.6 3.8 1 MB Yes
A4-5300 7480D 128 65 W 2 1 MB No

Of the six models purportedly planned, we have three of them: the A10-5800K, the A8-5600K, and the A6-5400K.

The A10-5800K will be AMD’s flagship. A pair of Piledriver modules technically makes this a quad-core APU, though, as we know, each module shares certain resources. The top-end A10 operates at a 3.8 GHz base clock that scales up to 4.2 GHz via Turbo Core, though our sample spent most of its time at 4 GHz (an intermediate P-state). Each of the -5800K’s Piledriver modules has its own 2 MB shared L2 cache, adding up to 4 MB across the chip. And AMD arms its two A10 APUs with Radeon HD 7660D graphics—a 384-shader engine operating at 800 MHz on the -5800K (and, reportedly, 760 MHz on the -5700).   

A small step down, the A8-5600K also leverages two Piledriver modules and 4 MB of total L2 cache (none of the Trinity-based APUs have L3). Its base frequency is 3.6 GHz with a 3.9 GHz Turbo Core ceiling. Both of the A8s come with Radeon HD 7560D graphics (256 shaders running at 760 MHz).

AMD’s A6-5400K represents a more significant departure from the other K-series SKUs. To begin, it bears a 65 W TDP, instead of 100 W. It’s also a single-module APU armed with two integer cores and a single floating-point unit. And instead of a shared 2 MB L2 cache, the -5400K is trimmed down to 1 MB of shared space. Radeon HD 7540D graphics are composed of 192 shader cores operating at an undisclosed frequency.

  • 1

Current page:
Trinity: Coming Soon To A Desktop Near You

Next Page Piledriver: Half Of The Trinity Story

Chris Angelini is an Editor Emeritus at Tom’s Hardware US. He edits hardware reviews and covers high-profile CPU and GPU launches.

AMD Trinity Mobile Platform Review

A10-4600M A10-4600M Mobile Platform First Look

  • Trinity Platform
  • New APU Graphics Core
  • New A-Series Mobility Lineup
  • AMD005 Prototype Notebook Performance 9005 synthetic benchmarks
  • Software performance
  • Gaming performance
  • Video playback
  • Battery life
  • Conclusions

AMD launched its VISION brand back in September 2009, and in 2011 the company introduced the so-called APU (Accelerated Processing Unit), launching AMD C and E series chips based on the Brazos platform. They combine the power of GPU and CPU in one chip, becoming one of the most energy efficient ultra-mobile solutions. And later last year, AMD announced the A-series, a hybrid solution platform codenamed Llano that was targeted at the mainstream PC market.

AMD’s previous series of APUs called Brazos and Llano were very well received by the market and proved to be quite successful for the company. They lacked stars from the sky in terms of maximum performance (especially in the CPU part), but they offered a good balance: universal computing cores powerful enough for most users and very good graphics performance for integrated solutions. Together with the low power consumption, this resulted in the outstanding energy efficiency of the first APUs.

And quite recently — on May 15, 2012 — AMD introduced an updated series A of its hybrid solutions, previously known under the code name Trinity, which has improved consumer characteristics compared to Llano. The new chips combine two or four «Piledriver» processor cores, as well as a «Northern Islands» series video core with 384 VLIW4 computing cores.

The main advantage of the APU is its high performance in 3D games. And the Trinity platform offers arguably the best features in its price range and within known power consumption limitations. The graphics core has been updated in the new APUs, now using a newer architecture, known to us from the AMD Radeon HD 7600 (codename «Thames»). The new video core provides very high performance compared to other hybrid (CPU + GPU) solutions.

The updated A-series of chips includes up to four x86-compatible computing cores, which are based on an improved architecture that first appeared in the Bulldozer processors. And the advent of support for the third generation of AMD Turbo Core technology ensures the highest possible performance of the CPU and GPU in a variety of workloads and stringent power requirements. With TDP limits set, the new A-series chips are great for ultraportable and thin laptops, as well as for home desktop PCs, although such APUs will be released a little later. Let’s compare the main characteristics of Llano and Trinity:

The new chip is manufactured using the same 32nm process and has 1.3 billion transistors, slightly more than Llano. The crystal area is 246 mm 2 , which is also slightly larger than the Llano area. By comparison, Intel’s quad-core Sandy Bridge is also made using the 32nm process technology and has almost the same number of transistors and die area as Llano (1. 2 billion transistors and 216mm 2 , respectively). But in the production of Ivy Bridge, a more advanced 22 nm process technology is already used, and with a complexity almost like that of Trinity (1.4 billion transistors), this processor from Intel has a much smaller area of ​​​​160 mm 2 .

Intel’s advantage in process speed is undeniable, and without switching to a new process technology, AMD had to curtail its appetite for more complex APUs. Compared to Llano, the size and complexity of the die have increased slightly, and the performance of the CPU and GPU parts, as well as their energy efficiency, although they have increased, but not as significantly as they could with 28 nm production, for example. But due to the improved architecture of both the CPU and GPU, Trinity has been able to increase the power, and this APU is a logical development of its predecessor, and a very good solution overall.

Trinity platform

So, the new APU series from AMD is based on a chip consisting of 1. 3 billion transistors, made on the basis of 32 nm HKMG process technology, and having an area of ​​246 mm 2 . The chip has two versions: FS1r2 722-pin uPGA and FP2 827-pin uBGA. The mobile version of Trinity has a typical power consumption (TDP) of 17 to 35 watts, depending on the model, while for desktop APUs this parameter reaches 100 watts.

The new A-series chips have up to four x86 cores, up to 128 KB L1 cache (64 KB instructions, 64 KB data) and up to 4 MB L2 cache. The clock frequency of «notebook» models reaches 3.2 GHz in turbo mode. The following RAM types are supported: DDR3-1600 (1.5V), LVDDR3-1600 (1.35V), ULVDDR3-1333 (1.25V) in dual channel mode.

The graphics core contains up to 384 processing cores and supports DirectX 11 API, the chip includes hardware units for encoding and decoding video data: UVD 3 and VCE. The integrated GPU in Trinity operates at frequencies from 424 to 800 MHz. Up to four video receivers can be used to display an image, all types of outputs are supported: Display Port, HDMI, DVI for three displays, and the fourth one can be connected via DisplayPort 1. 2 using a special hub. The analog connection uses the DAC built into the chipset.

Speaking of the chipset used. The new platform uses the already known chipset (Fusion Controller Hub) model A70M (Hudson M3), which is familiar from Llano. The chipset, although not new, is manufactured on a 65nm process, but it provides Trinity with everything it needs, supporting six SATA-III ports (with the ability to organize RAID 0/1 arrays), four USB 3.0 and 10 USB 2.0 ports (plus two USB 1.1 additional). Everything else of the current is also there, and as for the support of the chipset «only» PCI Express 2.0, in the case of mobile systems, the third version of PCIe is simply not needed, since it is still not easy to notice the sense of it even on desktop systems. The power consumption of the FCH chip is low — from 2.7 to 4.7 W under typical conditions.

Piledriver Compute Cores

As you probably remember, the Llano APU had four x86 Stars cores, and the Trinity included two Piledriver modules. These are improved cores compared to Bulldozer and are clearly better than the CPU cores used in Llano. The Piledriver has tweaked some of the weaknesses of the Bulldozer, although the overall architecture has remained the same.

Each Piledriver module contains the combination of two integer and one floating point processing core already known since Bulldozer. Each of the integer cores has its own schedulers, L1 cache for data, and execution units. The module also contains a common FP core that processes floating point instructions and uses a shared cache memory.

AMD engineers modified the compute core to increase the number of instructions per clock (IPC) executed by the microprocessor. The execution units themselves have not changed much and have become only marginally more productive than the Bulldozer in some operations (such as INT and FP division). More important changes have been made to the integer and floating point schedulers, as well as significantly improved branch prediction and prefetching.

L2 cache efficiency has also increased, and L1 TLB has become larger. And another expected change in Piledriver was the update of the instruction set architecture (ISA) with new instructions: FMA3 and F16C, in addition to AVX, AVX 1.1 and AES.

Turbo Core 3.0 Technology

Technologies that automatically overclock one or more CPU cores, as well as the integrated GPU, have become widespread in recent years and are now almost everywhere. Llano already had support for Turbo Core technology, but Trinity has greatly improved it.

Turbo Core 3.0 supports overclocking for both CPU cores and GPU parts of the chip, and in Llano only the former could be accelerated (if there was “free” power consumption, of course), and the graphics core in the previous APU always worked at the base frequency. In Trinity, if the CPU cores do not use the entire possible power reserve (when it does not exceed the TDP value), and the GPU is loaded with work, then the frequency of the latter increases. The same thing works for CPU cores — if the main load goes to one of the x86 cores, then its frequency increases to the maximum mark, if the power consumption does not exceed the set TDP value — see the diagram:

The on-chip control circuit monitors the power consumption of all units, and has been made more complex in Trinity. In Llano, the Turbo Core scheme simply monitors only the activity of the CPU and GPU, and increases the frequency of the CPU if the GPU is not loaded with work, and in Trinity the consumption of each block is calculated based on their load, and then the temperature regime for them, and the accuracy of these calculations high enough. As a result, the Turbo Core 3.0 control scheme allows faster and more efficient control of frequency changes, and with it, the overall energy efficiency of the solution also increases.

By the way, Trinity’s numerous efficiency and power management improvements have resulted in improved battery life. According to AMD, such devices can work up to 11 hours in idle mode. The total average power consumption of the system, including both the APU and the chipset (more precisely, the Fusion Controller Hub) is only 1-2 W in idle mode and only 6 W in video data playback mode. What happens in practice, we checked in one of the following sections of the material.

Memory interface and other connections

The main theoretical advantage of APUs is their heterogeneous Heterogeneous System Architecture (HSA), where a single chip contains CPU and GPU cores that perform their specialized tasks using the same system memory, and communication between them can be very fast.

So far, not all of this is implemented in current chips, but in the near future it will become an important advantage of hybrid solutions — only a wide chip bus between the CPU and GPU will make many tasks easier. Here’s how AMD sees the development path for its APUs — if the GPU already has access to RAM, then future models should have shared memory addressing, as well as context switching for GPU computing:0003

Like previous APUs, Trinity chips contain two 64-bit DDR3 memory controllers that support standards up to DDR3-1866 (provided throughput up to 29. 8 GB / s). The maximum amount of supported memory for mobile Trinity chips is 32 GB, and for desktop — 64 GB. Of the innovations, one can only note the added support for memory chips operating at a voltage reduced to 1.25 V.

The former Hyper Transport for external connections has been replaced with PCI Express. A 128-bit bi-directional Fusion Control Link (FCL) provides memory access for external devices. So, the GPU with its help gets access to the cache memory and RAM, and the CPU gets access to the dedicated framebuffer. Trinity also has support for a 256-bit bi-directional Radeon Memory Bus (RMB) for direct access to DRAM memory controllers, as well as communication between the CPU and GPU. RMB allows the graphics core to quickly access system memory.

And for accessing discrete GPUs used in tandem with Trinity, directly to the virtual memory of the CPU, IOMMU v2 is used. Compared to the scheme in Llano, data transfer to the GPU has been simplified, now there is no need to copy them from the CPU address space to the RAM area that the graphics core has access to, now data is directly sent from RAM to video memory, bypassing unnecessary copying from one area of ​​RAM to another.

The graphics core of the new

GPUs in Trinity is based on the Cayman architecture we first saw in the Northern Islands family. The video core built into the APU uses the VLIW4 design and contains 6 SIMD engines, each of which has 16 VLIW4 blocks, that is, in total we get 384 computing cores. This number is valid only for A10 models, which have 384 cores each, while chips marked A8 and A6 have 256 and 192 active stream processors, respectively.

«Northern Islands» can be called the previous generation of AMD graphics architecture, although only video cards for the upper price range, the Radeon HD 6900 series, were released based on it. Inexpensive options with VLIW4 never came out. Interestingly, although Trinity has fewer processing cores in the GPU compared to Llano, the transition from VLIW5 to VLIW4 increased the efficiency of their use, since the fifth block of VLIW5 was busy working in an extremely limited range of tasks — the same transcendental functions use only 3- 4 blocks available. The use of VLIW4 simplified both the tasks of the scheduler and the management of registers, which led to an additional increase in efficiency.

In addition to the stream processors, the GPU includes 24 texture units (4 TMUs per SIMD) and 8 ROP units, which is about a quarter of the Radeon HD 6970, if you do not take into account the lower frequency. However, the Trinity graphics core’s turbo clock for the top models is 686 MHz, which is not that far from the 880 MHz for the Radeon HD 6970. Islands, as well as support for all known types of full-screen anti-aliasing, including SSAA, EQAA and MLAA. Naturally, the graphics core supports DirectX 11 and OpenCL 1.1 — these are some of AMD’s advantages over Sandy Bridge (but not Ivy Bridge). Read more about the Northern Islands graphics architecture in the Radeon HD 69 baseline review.70.

The well-known AMD Eyefinity technology is used to display images, the new APUs support up to four monitors and independent audio streams, as well as DisplayPort 1. 2 outputs with data transfer rates up to 5.4 Gb / s and support for multi-stream output. It should be noted that the new APU also includes the HD Media Accelerator, which improves video quality (post-processing) and includes UVD 3 video decoding and VCE video coding units.

Although the GPU in Trinity is VLIW4, the video encoding unit was borrowed from the later Graphics Core Next architecture. The third generation UVD features support for MPEG-4/DivX hardware processing, as well as the ability to decode two channels of FullHD video, which is also used when decoding video data in stereo format.

The technology for transcoding video data was named AMD Accelerated Video Converter . Multi-threaded H.264 hardware video encoder supports resolutions up to FullHD, 4:2:0 color sampling, variable compression quality, and specialized optimizations for different types of images. Provides quick access to framebuffer data for video transcoding, video conferencing tasks, and wireless image transfer to an external display. The VCE hardware block provides energy-efficient, faster-than-real-time video encoding with low latency.

In addition, it is worth noting the technology for improving the quality of playing streaming video — AMD Quick Stream Technology , as well as the technology for real-time video image stabilization AMD Steady Video. Quick Stream is interesting in that video streaming traffic on compatible AMD platforms is given the highest priority over other tasks that use the network channel. This achieves smooth playback of streaming video data without waiting for their loading.

AMD Steady Video Technology improves poor quality handheld footage without the use of a tripod or other similar image stabilization tools. GPU-assisted video stabilization technology has been supported in AMD solutions for some time, but its second version has appeared in the Radeon HD 7000 series of video cards.

The algorithm of the software stabilizer is quite simple: on the basis of the video stream, statistics about the camera movement (shift, rotation, zoom) are collected and this movement is compensated in the current frame, relative to the previous ones — the image is shifted, rotated and scaled so that the image does not jump much and remains stable.

Although the task is simple, it is very resource-intensive, because there are two million pixels in a frame, and 30-60 frames per second. And to keep track of all possible frame offsets, you need to do a lot of calculations. Graphics cores supporting Steady Video 2.0 are capable of handling random shifts up to 32 pixels in any direction, and this requires support for specialized commands, which is now included in the latest generation of APUs.

A Series of New Mobility Solutions

The Trinity platform comes to market in two forms, as does Llano. Desktop solutions are based on Virgo chips, but they will enter the market later — closer to autumn. In the meantime, APU models for laptops have been released, codenamed Comal. AMD’s mobile solutions are favored for many reasons, especially since Trinity has the advantage of power efficiency, which is especially important for laptops.

This can also be seen from the established figures for typical energy consumption. While Llano had only two variants with TDPs of 35W and 45W, mobile Trinities have models with consumption: 17W, 25W and 35W (for desktop PCs there will be levels of 65 and 100W). In addition, according to AMD, the new generation of APUs is almost twice as energy efficient as Llano. In total, Trinity mobile chips came out five different models, aimed at different markets, and they all differ in their consumer characteristics: 497 (655) 35 A6-4400M HD 7520G 2 2,7 (3,2) 1 192 497 (686) 35 A10-4655M HD 7620G 4 2,0 (2,8) 4 384 360 (497) 25 A6-4455M HD 7500G 2 2.1 (2.6) 2 256 327 (424) 17

As we noted above, Trinity uses modules containing two Piledriver cores with one common FP unit (FP / SSE). Therefore, we can say that Trinity chips are quad-core or dual-core processors. And although if you count by the number of FP blocks, then a “real” quad-core does not work, but the number of certain executive devices in itself is not as important as the overall computing performance.

And compared to Llano solutions based on old cores, the CPU frequencies of the Trinity part are much higher, this applies to both the base frequency and the turbo frequency. The top model A10-4600M has a base frequency more than half as high as the A8-3500M from the Llano family, and its turbo frequency is a third higher. On the other hand, the pipeline of the Piledriver kernel is longer than in the modified K10, which will affect some applications, and the performance difference will not be so impressive.

The GPU part of Trinity is also very different from what we saw in Llano. We have already noted that the old APUs used the VLIW5 architecture graphics core, known from the Radeon HD 5000 series models, and different APU models had 400, 320 or 240 cores. Trinity uses the VLIW4 architecture seen in desktop models of the Radeon HD 69 series.00, and the number of active streaming cores in the new chip models is: 384, 256 and 192.

operating frequencies for the GPU in Trinity, the graphics performance of the new APUs has grown even more seriously than the performance of general-purpose x86 cores.

AMD pits its new solutions against corresponding Intel models based on the estimated retail value of the end devices. So, the A10 model is positioned between Intel Core i5 and Core i7, A8 — between Core i5 and i3, A6 — slightly lower than Core i3, and the younger A4 should be slightly more expensive than laptops with Pentium, but cheaper than all Intel Core.

Interestingly, AMD is using the A10 designation for its top Trinity-based models, after all there used to be only lower-performance models named A8 and A6. This is understandable, because according to the company, the A10-4600M model provides about 56% more GPU performance and 29% faster general-purpose computing compared to the A8-3500M. True, with the second digit it is not clear whether we are talking about CPU performance or, nevertheless, including universal computing, in which the GPU also helps.

The A10-4600M is the most powerful APU to date, designed for performance mid-range notebooks that are well-suited for light gaming and other typical applications. The A8-4500M is more than a third slower in terms of graphics performance, and the general-purpose computing cores have lost a little in frequency, but this APU can be used in cheaper laptops, although it will already be noticeably heavier in games. Well, the simplest A6-4400M contains only two universal CPU cores, and the GPU has about half the performance of the top solution. All models support DDR3 memory types up to DDR3-1600.

The remaining two models from the new APU line are lower power consumption and designed for use in thin laptops like the HP Sleekbook — that is, in fact, analogues of ultrabooks based on Intel processors. And matching Trinity desktop processors, when they hit the market, could be the basis for new form factors in compact PCs.

The more powerful A10-4655M has only ten percent less CPU performance than the A10-4600M, and a third lower graphics processing speed. At the same time, such power is content with the consumption of only 25 watts of energy! For the younger ULV model A6-4455M, the TDP is even lower — only 17 W, which is completely the same as similar models from Intel. Naturally, the speed of the CPU and GPU in this model is greatly reduced — it has only two Piledriver cores and 256 processors in the GPU, and the frequencies are noticeably reduced. It should also be noted that low-power models have lost support for DDR3-1600 memory, providing memory standards up to DDR3-1333 inclusive.

Approximate estimates of the performance of new APUs can be made according to data from AMD, which compares Trinity with Llano in terms of energy efficiency in graphics and other applications separately:

Probably, this column also takes into account the speed in applications with support for OpenCL acceleration. Much more interesting comparative tests with a competing Intel Core i7-2720QM in DirectX 9and 10 games:

True, there are no specific figures here, but only the advantage of the AMD solution, indicated in percentage. And it is quite natural that it is quite large, because the competitor’s processor has an outdated GPU. Intel processors up to Ivy Bridge (whose mobile versions have not yet been released) have an integrated graphics core without DirectX 11 support, and to achieve acceptable performance in modern games, Intel processors can only be helped by installing a discrete accelerator from NVIDIA, which increases the price of the final solution. Especially when compared to AMD’s APU-based laptops, they provide similar speed in 3D games without the use of additional chips.

AMD Trinity Laptop Prototype

The Trinity Prototype Mobile Solution that was presented to us by AMD at the Austin press event already looks more like the final solution than it did with Zacate, for example. Although the design of the laptop was developed by one of the well-known manufacturers, it is definitely not intended to enter the market, although it fulfills its purpose well — it is quite possible to draw conclusions about the platform using its example.

Such a decision is almost the only opportunity for journalists to get acquainted with the novelty even before laptops based on it hit retail stores. At the same time, the prototype is quite functional, and all the usual tests on it pass perfectly. Interestingly, there are AMD logos on the laptop: on the lid, under the screen and above the keyboard. Since the laptop will not hit the market in this form, it makes no sense to disassemble the design solutions used in it — the models that have already gone to retail are completely different. Yes, this is for the best, since the prototype looks too simple and inelegant, unlike the suitcase in which it was given to us:

Of the technical parameters that are worth mentioning, we note that the APU model is the A10-4600M with the standard parameters listed above. AMD’s prototype laptop has a decent 4 GB of memory and an SSD, a decent battery life, and even an optical Blu-ray combo drive. Of course, it is far from being as thin as ultrabooks, but this is understandable — the prototype simply did not have such a goal. Let’s look at the technical characteristics of the model we are considering today:

As you can see, the A10-4600M operates at 2.3 GHz and has the ability to automatically overclock to 3.2 GHz (when only one of the available computing cores is loaded) using Turbo Core 3.0 technology, as well as cache 2 MB L2 memory per dual-core module. Let’s see what interesting things the CPU-Z diagnostic utility can tell us about the CPU and system used:

We didn’t notice anything particularly interesting — the utility is already able to determine the characteristics of the Trinity platform chips. Information about the cache memory and supported extensions, the number of physical and logical processors are correct. The x86 core clock is shown in idle state, and the chipset was identified as A55/A60M.

The APU has a relatively high frequency and the four (or two, depending on how you count) available cores should be enough for most common tasks like office applications and browsers, except for the most demanding computing like professional applications in 3D modeling or video editing. And in most modern gaming applications, the CPU speed should be enough.

The prototype laptop was equipped with 4 GB of DDR3 memory, which is quite common for laptops of this class. For data storage, the AMD notebook is equipped with a fast, if not very capacious, SSD from Samsung. So you don’t need to worry about the speed of loading and system operation — an SSD will provide quick access to data and will not become a performance limiter.

Another important hardware feature of the prototype is the integrated video subsystem, available in the A10-4600M processor. Although this is an integrated solution, it is very powerful and energy efficient, and should provide 3D performance on par with some discrete graphics cards, especially when compared to past generations. And it’s completely wrong to compare with the integrated video from the same Intel, since in the same Sandy Bridge games, if they run without problems and artifacts, then the integrated GPUs are unable to provide acceptable FPS in them even at low settings.

Let’s see what the GPU-Z test utility can tell about the characteristics of the prototype graphics core based on Trinity:

Radeon HD 7660G

/or invalid data. So it happened in our case — a lot of things are not defined at all, and what is is not always indicated correctly. So the readings of the utility in this case are practically useless, because the utility could not even show support for DirectX 11 and OpenCL.

Everything else in the provided prototype laptop worries us to a much lesser extent. Its communication capabilities are not too impressive, but there is a necessary set of interfaces: a Gigabit Ethernet network adapter, Wi-Fi 802.11b/g/n and Bluetooth 2.1 wireless interfaces (not even 3. 0, oddly enough). That’s why he’s a prototype. Let’s move on to examining the performance of the new APU.

Performance in synthetic benchmarks

As always, we start looking at performance from synthetic benchmarks, which show speed in artificial conditions, allowing you to quite clearly limit the influence of various subsystems on each other: CPU from GPU and vice versa. In this section of the article, we will look at the results of synthetic system performance tests in the following test suites: PCMark Vantage, Cinebench, 3DMark’06 and ’11, and Heaven 3.0.

First, let’s take a look at the Windows 7 performance ratings. This is the simplest synthetic performance method available on every system with Windows 7 or Vista installed. For comparison, we took mobile systems from Acer and ASUS previously tested using this method, as well as an engineering sample from AMD Zacate.

9024 9024 9024 9024 9024 9024 902 shows that the performance of the x86 cores of the new Trinity platform is quite good and roughly corresponds to that of the quad-core Core i7, although not new, but still, except for the speed of accessing data from memory, which depends on the size and speed of the cache memory. Interestingly, the A10-4600M turned out to be faster than the “ultrabook” Core i5-2467M. Well, in the drive test, two test systems from AMD are in the lead, which is explained by the use of full-fledged SSD drives in them, unlike HDD and hybrid systems from other test participants.

We are most interested in graphics performance, and here the new APU showed itself exceptionally well. In the «gaming» 3D graphics mode, they showed a result that almost matched the speed of such fast solutions as the previous generation AMD Radeon HD 5850 and the latest NVIDIA GeForce GT 640M. And in the Aero graphics subtest, there is almost no lag behind the indicated Radeon, and the less productive integrated Intel video core.

However, we didn’t expect anything special from the built-in test in Windows, because it is far from ideal, especially in determining 3D performance, which we will return to more than once. And now let’s look at the results of the PCMark Vantage system-wide test. Let’s take into account both the final result and individual results by subsystems. Detailed numbers will help us evaluate the performance of the various components of the laptop and how they contribute to the overall score.

Windows 7 rating AMD
Trinity
(A10-4600M
graphics Aero 6. 7 5.7 6.9 5.1 4.0
6. 5.9 5.5
Hard disc 7.6 5.9 5.9 5.8
PCMark Vantage AMD
Trinity
(A10-4600M
HD7660G)
Acer M3
(i5-2467M
GT640M)
Acer
5943G
(i7-720QM
HD5850)
ASUS
K52Jr
(i3-350
HD5470)
AMD
Zacate
(E-350
HD6310)
PCMark Score 10056 6106 5632 4445 3680
Memories Score 5834 4624 4134 2916 2240
TV and Movies Score 4004 2639 4029 3242 1595
Gaming Score 7272 8316 5788 3648 3722
Music Score 11570 8489 4599 4659 4916
Communications Score 9973 8181 4017 3717 3024
Productivity Score 12354 8434 4391 4087 4582
HDD Score 22013 15381 3072 2760 13809

There is no benefit and practical meaning in such a comparison, but detailed results are interesting, since they immediately indicate the strengths and weaknesses of the tested solutions.

So, in the RAM subtest, the new platform from AMD surprisingly became the fastest, overtaking all other test systems. The rather fast DDR3 and good caching are probably to blame for this. The result in «TV and Movies» is normal, at the level of a quad-core Acer laptop, and the huge difference in the rest of the system tests in favor of today’s prototype is due to its use of an SSD as the only drive — which is why many tests showed such strong results. However, without a sufficiently powerful central processor, they would not exist either.

The most interesting «gaming» test, in which the result was obtained between the Radeon HD 5850 and the GeForce GT 640M, and closer to the latter. Unfortunately, this assessment cannot be objective, since the comparison is spoiled by the presence of an SSD in some configurations, and the Gaming Score considers an average rating that measures the speed of loading data from the drive in games as well. And PCMark Vantage in general depends too much on the speed of the installed drive.

The next test we’ll look at is Cinebench of the old version of R10 that we’ve been running since 2010. This is not exactly «pure» synthetics, but rather a performance test based on the code of the widely used Cinema 4D application, a professional package for creating and rendering three-dimensional images and animations.

Cinebench contains three subtests: rendering using one CPU core, all CPU cores (in this case, four threads running on two cores) and the most interesting OpenGL subtest for us, which uses real-time rendering of a complex 3D scene. The last test allows you to evaluate the performance of the graphics subsystem when working in similar professional packages that use OpenGL.

Cinebench R10 AMD
Trinity

2226
Opengl 5597 5061 6860 4114 1960

Consider the processor tests CENCH The APU we are reviewing today has four integer cores and two FP cores, the performance gain from «multi-core» in this test turned out to be almost three times, even despite the influence of Turbo Core, which spoils a direct comparison. In the case of Intel processors, they were helped by Hyper Threading, which allows four threads to run on a dual-core processor and eight on a quad-core one.

The comparison with the Core i5-2467M is quite interesting. If in the single-threaded test the Intel solution wins, which has a more productive x86 core, then in the multi-threaded test, the new AMD A10-4600M, which has a larger number of cores, is already breaking ahead. That is, by itself, each core in Trinity is slower, but due to their number, a gain is obtained.

The OpenGL subtest is also interesting, the results of which show that although the Radeon HD 7660G is inferior to the mobile version of the Radeon HD 5850, the new GeForce GT 640M falls behind in this test, since this test is not a strong point of NVIDIA video cards. In general, the top model of A-series chips performed quite well in Cinebench. 943G
(i7-720QM
HD5850) ASUS
K52Jr
(i3-350
HD5470) AMD
Zacate
(E-350
HD6310) Score 7955 10504 9210 4047 2011 GT1 23,9 40,6 27,1 10,7 5,4 GT2 24,9 36,8 31. 5 12.2 6,2 HDR1 34,8 48,3 38,9 15,9 8,1 HDR2 36,8 51,5 42.8 17.8 9.0

You can clearly see the difference in speed between older laptops and more modern ones, whose average frame rates in the tests of this package are already quite acceptable. Even such a powerful once mobile Radeon HD 5850 is only slightly ahead of the recently introduced novelty — the hybrid solution of the Trinity platform. And for other graphics cores built into the CPU, this test is completely too difficult, which can be seen on the example of the AMD Zacate video core, and the GPU in Sandy Bridge is even weaker.

The Radeon HD 7660G does the job very well, delivering frame rates of around 25-35 FPS. Of course, this is less than that of the same GeForce GT 640M, well, that’s why it is discrete graphics, which together with the CPU consume much more than the A10-4600M alone. In general, the overall 3DMark’06 score is usually a good reflection of the performance of different GPUs. The GT 640M is clearly the best in the test, followed by the Radeon HD 5850, and our today’s hero took the honorable third place, and this is an excellent result for a hybrid processor!

These were all old synthetic tests, the results of which we have included in order to compare them with previously tested laptop models. A lot of time has passed since then, new test packages have been released that are more relevant for evaluating the performance of modern video cards. The first modern test will be 3DMark’11 of the same Futuremark. Acer M3
(i5-2467M
GT640M) Score 1153 1773 Graphics 1062 1697 Physics 2642 2724 Combined 960 1494 GT1 5. 28 8.27 GT2 5.07 7.94 GT3

5,93 10.26 10.26

9015EAL0137 GT4 3.22 5.06

We will only compare the results of the AMD A10-4600M in this package with the numbers of the recently tested gaming ultrabook Acer Timeline Ultra M3, which has a discrete GeForce GT 640M graphics card from NVIDIA. Because these are the first mobile solutions that we tested in the 3DMark’11 test suite.

Trinity APU-based system’s 3DMark’11 score with a total score of 1153 at default settings is about the same as a desktop GeForce GT 430 and one and a half times worse than AMD Radeon HD 6670. Although this is not such a high performance in terms of desktop solutions, but an excellent level for an integrated mobile solution.

The performance of the Radeon HD 7660G is enough for many modern games, especially multi-platform ones and with not the highest settings. But what will happen in gaming applications that actively use DirectX 11 features such as tessellation, compute shaders, etc.? To find out, we tested the prototype on Trinity in one of the most difficult 3D tests — Unigine Heaven 3.0.

In addition to tessellation tests in three modes, we also tested various levels of full-screen anti-aliasing using the MSAA method and determined the performance drop when anisotropic filtering is enabled. For convenience, all results are presented in the form of a diagram:

Even with reduced shader complexity, the Heaven test is quite difficult for laptops, and even more so for integrated graphics. But the Radeon HD 7660G doesn’t do too badly, delivering almost 30 FPS with anti-aliasing, anisotropic texture filtering, and tessellation off, and turning on anisotropic filtering drops the average fps by 5%.

Let’s see how much performance drops when full-screen multisampling anti-aliasing (MSAA) is enabled. Of course, in this case, the rendering speed decreases even more, and in the case of 8x MSAA, the FPS drop is especially large, but the level of 2x is not so difficult for the Trinity graphics core and, most likely, this solution will be able to provide a playable frame rate in undemanding games even with multisampling enabled.

Tessellation reduces the performance of the integrated video core in the A10-4600M even more, so it is unlikely that you will be able to play DirectX 11 games on a laptop with integrated graphics. But almost the same is observed in the case of much more powerful solutions, even the minimum level of tessellation significantly reduces the rendering speed. Well, nothing new — extreme settings are clearly not for such mobile solutions.

And we are moving from ambiguous system-wide and synthetic tests, sometimes showing rather strange results, to testing the new mobile APU from AMD, in a set of real gaming applications, both modern and long used in our performance studies.

Performance in various software

In previous articles on AMD hybrid systems, we often wondered when GPU computing will start to be used in software that is often used by ordinary users, well, at least part of them? In games, after all, GPU calculations are already used, both in the form of PhysX and in the form of post-processing on DirectCompute. For a long time, nothing but games, in fact, was not.

For scientific computing and some other tasks, computing on the GPU has long been of great importance, but not for the average user. Few people are engaged in video encoding on their own, transcoding from format to format, too. Well, editing with encoding their own videos there — not everyone spends their time on this.

In general, then we concluded that although GPU computing looks like a very promising direction, but at that time there was almost no sense in computing on the GPU. But the emergence of APUs and other hybrid chips gave an additional impetus to the development and emergence of such software. The possibility of parallel computing appeared in a large part of the systems, not only focused on games and having discrete graphics cards. And the development of the open standard OpenCL also helped in increasing the number of applications for computing on graphics cores. Well, let’s see what we are now offered to calculate on GPU

We have known for a long time that one of the first tasks implemented by the GPGPU is the processing and encoding of video data. But the development of video encoders does not stand still, in future versions of the well-known x264 codec, which is considered the most popular among H.264 encoders and is used in many applications, OpenCL acceleration is expected to appear. In the meantime, let’s consider the software where such acceleration has already been implemented in one form or another.

For example, ArcSoft MediaConverter 7.5 is a powerful yet easy-to-use media converter. With it, you can easily convert video files for use in phones, players and other devices. The latest versions of this package use the capabilities of the hardware VCE encoder of Radeon video cards (including the one in Trinity) when transcoding video — when transcoding to the format of devices that support H.264.

Another application from the same company is Link+ 3. This is an application for convenient access to multimedia data (photos, music, videos) from any device on the local network. Link+ 3 automatically combines the capabilities of network devices and allows you to view media files from them. We are more interested in AMD technology support: UVD for viewing, VCE for transcoding, HD Media Accelerator for smooth and high-quality playback. ArcSoft’s SimHD technology uses the power of general-purpose GPU computing to scale video, while watching video is stabilized with Steady Video.

There are other similar applications like CyberLink MediaEspresso. Version 6.5 supports the capabilities of hardware video conversion — AMD Accelerated Video Converter, using the VCE block when transcoding. And CyberLink PowerDirector 10 is even more advanced, its main component is the TrueVelocity 2 video engine, which is optimized to use the capabilities of modern AMD GPUs.

This application also uses Accelerated Video Converter for transcoding (hardware UVD decoding and VCE encoding) and OpenCL acceleration for additional effects such as: Zoom In, Gaussian Blur, Color Focus, etc.

In addition to video processing applications, advanced GPU capabilities are used in media players like ArcSoft TotalMedia Theater 5. The fifth version supports OpenCL acceleration of the already mentioned ArcSoft SimHD technology, which includes scaling, denoising, dynamic contrast and frame rate conversion. In addition, the capabilities of the UVD 3 hardware video decoding unit and AMD HD3D technology are used to view video in stereo format.

Almost all of this video conversion and viewing software was previously known. Much more interesting are the applications of those companies that have not previously been accelerated using graphics chips. So, among Adobe’s applications, Flash can be noted, where the power of the GPU is used in three-dimensional applications, and modern versions of Flash (starting from 11.2) support very extensive capabilities for hardware acceleration of 3D graphics.

But much more interesting is the very recent version of the Adobe Photoshop CS 6 graphics package, which offers GPU hardware acceleration for some of its features using OpenCL and OpenGL. And if we have been familiar with OpenGL acceleration for some time, then the use of OpenCL appeared in CS6 for the first time. In total, more than 30 functions are accelerated in the new version of the graphics package, including Liquify, Transform and Warping.

The new Mercury Graphics Engine displays results immediately, almost in real time. And the power of OpenCL is used to accelerate the computationally intensive «Blur» effects. The GPU acceleration setting «Use Graphics Processor to Accelerate Computations» is enabled by default. Among the other GPU-accelerated tools of the new version of Photoshop CS6, we note the Oil Paint filter, adaptive perspective correction (for wide-angle lenses), a gallery of lighting effects, as well as transformation and warping tools.

The Liquify filter is accelerated with OpenGL and completely redesigned in CS6 to use the Mercury Graphics Engine for loading, previewing and final rendering. When programmatically processing large images with a filter in Photoshop CS5.5, the work was noticeably less comfortable, and now applying the filter practically does not slow down. If we talk about specific numbers, then the AMD A10-4600M, with the inclusion of GPU acceleration, is more than twice as fast at this job, and faster than competing solutions from Intel.

The new «Blur» effect gallery provides the ability to quickly apply complex effects like Field Blur, Iris Blur and Tilt-Shift — mimic the appropriate lens type by setting the focus area and blurring the rest of the image. This is a new feature introduced in Photoshop CS6 and uses OpenCL for final rendering. As a result, the same A10-4600M provides a 7-fold increase in speed with GPU acceleration enabled, and in general, it is noticeably faster than competing platforms that do not have OpenCL support.

It was all theory with only a few numbers, but what happens in practice? How much does the graphics core in Trinity chips speed up calculations? Let’s take a look at a few GPU applications. The first is MotionDSP’s vReveal 3.3, a simple yet powerful tool for organizing, editing and enhancing your videos with ease.

One of the most interesting features is the «One-Click Fix» function, which automatically improves the quality of the video, correcting imperfections such as low contrast, incorrect colors (white balance), and also stabilizes the video. Hardware video encoding is supported with Accelerated Video Converter and HD Media Accelerator, and OpenCL is used in other operations.

We tested the «render» time of a short high-resolution video in vReveal, applying the same automatic quality enhancement to it. At the output, the video really became smoother and more stable, as well as improved contrast and color saturation. But what about the speed, what gives the use of the GPU in this task? GPGPU Operation time 5:35 0:56

As you can see, the difference in video processing performance turned out to be very large — with the help of the GPU, the system managed to process the video 6 times faster than when using only x86 cores. The result is very good, as video processing is highly parallelized and suitable for acceleration on hybrid systems. Let’s see what happens next — in software for a different purpose.

We have already mentioned above that when playing video data, GPU capabilities can be used, this applies to both banal DXVA acceleration and more advanced post-processing and video stabilization methods. One of the most common media players is the open source VLC Media Player.

In the latest versions, this player supports features of AMD’s new APUs such as Steady Video 2.0 real-time video stabilization, and also uses OpenCL to enhance playback through post-processing such as noise reduction.

Video stabilization really works well, although not without «childhood illnesses» so far — it does not turn on on all videos, it does not work well in battery mode, etc., but these are all software problems that will be fixed in the near future. More interesting is the possibility of GPU acceleration during video decoding and post-processing, and we tested it:

If far from all users are engaged in complex video and image processing, then almost everyone is familiar with archivers to one degree or another. We already wrote in the reviews of the new series of AMD Radeon HD 7000 video cards about the support of new GPUs by the WinZip 16.5 archiver. WinZip is one of the most popular file compression, encoding, and backup utilities. And even with the fact that its popularity has fallen in recent years, WinZip remains one of the most common archivers.

More interestingly, WinZip 16.5 supports not only multi-threaded file compression on multi-core CPUs, but also OpenCL-accelerated compression. For more efficient GPU compression, file processing had to be parallelized — with OpenCL enabled, the archiver processes several files at once.

AMD partner press releases claim support for OpenCL acceleration on all compatible AMD products, from APUs to AMD Radeon discrete graphics, and up to 2.5x faster compression than WinZip 16. The same applies to encryption using the AES algorithm, which requires a lot of computing resources and is well parallelized, and therefore also accelerated using OpenCL.

The 2.5 times speedup figure seems too high to us, and the comparison with the old version of the archiver is not so interesting, so we checked the compression speed on two sets of files. The first such set was the game Lost Planet, consisting of more than 200 files with a total volume of 7.5 GB. ZIPX format was used for compression, with and without AES encryption:

WinZIP Software OpenCL
ZIPX method 27:25

26:16
ZIPX+AES128 method 27:16 25:09

It doesn’t smell like 2.5 times! The speed difference we got was only 4% and 8% for compression in normal mode and using AES encryption. This is clearly not enough to consider the problem suitable for GPU computing. It is very likely that data compression in the ZIP format is simply poorly parallelized, and when transferred to the GPU, the acceleration is very weak.

But maybe a small increase in performance is due to a small number of files that are poorly parallelized and compressed? We checked the second set of files, consisting of executable files and data files with various drivers (in total, more than 7000 files of various sizes, the total size is 1. 3 GB).

After a not so good example, let’s go back to image processing — but this time to still images and to Adobe Photoshop’s competitor, if you can call it that — GNU Image Manipulation Program (GIMP) version 2.8. It is the most popular open source image editor that is widely used around the world.

This version introduced support for OpenCL acceleration, designed to improve the performance of rendering, filters, and other computing tasks. The current version already supports OpenCL acceleration for 19filters — the so-called GEGL operations. A future major update to GIMP will bring the GEGL library into the main processing pipeline, while the current OpenCL acceleration works with GEGL filters, but not the GIMP pipeline as a whole. So in full-fledged releases of the next versions, the benefits of OpenCL should become even greater.

GPU acceleration works best for 4-channel images with 8-bits per color — and this is the most requested format. Moreover, it is desirable that the horizontal and vertical resolutions of the images are evenly divisible by 512. To get the maximum difference, we tested the processing of an image with a size of 40

3.983 snn-mean 0.156 6.721

Well, now we see a big difference again. Moreover, the speed of execution of OpenCL filters on the CPU and GPU differs not by 2.5 or even 10 times, but by up to 100! We got the advantage of the GPU over the CPU from 15 to 108 times, depending on the applied filter. It is clear that image processing is most suitable for using the power of the graphics core, and for the CPU, the task may simply not be optimized enough, since the OpenCL code on the CPU is not always executed efficiently. In any case, those who edit images in GIMP and use similar filters will be happy.

Gaming performance

This is one of the more interesting sections of the material. If in terms of performance in office tasks and video data acceleration integrated graphics cores have long caught up with discrete solutions, and the difference between dedicated and integrated video cores in these tasks is not so great, then in terms of 3D performance the lag is still quite noticeable, even taking into account the significant increase in the performance of integrated graphics cores in recent years.

It will be all the more interesting to see what the new AMD platform can give in these conditions. After all, all APUs had an advantage in games, and Trinity is likely to become the best hybrid chip with integrated graphics for maximum performance. Although it is unlikely that anyone will choose a laptop for gaming, considering models with integrated video cores, but such powerful integrated solutions may well give undemanding users the opportunity to play many of today’s 3D games. Even if the user has to lower a couple of rendering quality settings.

Since this is one of the most important sections of the review, there will be a lot of game tests in our material. First, we’ll look at a few older games at relatively low gaming quality settings to compare the results of a prototype laptop based on the A10-4600M hybrid chip with previously tested mobile systems with AMD graphics solutions.

And we will start with projects that are not too demanding by modern standards. The first game in the review will be the game of the famous series Call of Duty — the first part of Modern Warfare. Newer games in the Call of Duty series are technically not much different from MW, and they have almost the same engine. For tests, a demo recording of a multiplayer battle was used.

In the case of the old game CoD: Modern Warfare, in addition to the minimum quality mode, we also used the maximum settings using full-screen anti-aliasing MSAA 4x. In both modes, the new APU model from AMD showed excellent results. In simple mode, the speed is limited to 90 FPS, and in this mode the tested prototype of the laptop was not inferior to the near-top Acer 5943G.

Well, in the maximum quality mode with multisampling, the speed is already limited by the capabilities of the graphics cores, and here the test laptop on Trinity is behind the top solution of not so long ago. And the main conclusion is that in outdated games, the A10-4600M is quite capable of providing a playable frame rate in difficult conditions at maximum settings even with anti-aliasing turned on, while other integrated solutions can be played normally only at medium quality settings.

Not all games are demanding on GPU power, there are a large number of games from the recent past that work well even on weak systems. Usually these are multi-platform projects, designed, among other things, to work on game consoles, the hardware of which was also released quite a long time ago and lags far behind modern PC hardware. One such game is Resident Evil 5:

This is another game that was released immediately on both consoles and PC. Although Resident Evil 5 is a multi-platform game, it is quite demanding on the power of the system, including the GPU. For example, a low-power GPU in the AMD Zacate platform cannot provide the required 25-30 FPS even at medium quality settings, and the weakest discrete graphics card from AMD somehow shows a level of 30-40 FPS.

But the Radeon HD 7660G model as part of the top-end Trinity chip, on which the prototype under consideration is based, showed a very good comparative result, but only in the medium quality mode. Rendering in Resident Evil 5 at low settings is limited by the speed of the CPU and in it the Acer Aspire 5943G laptop, which has a powerful quad-core Core i7, significantly outperformed other comparison participants.

But at medium settings, the influence of CPU power is leveled, and GPU power becomes the main frame rate limiter. And then the new Trinity platform bounced back, showing more than 50 average FPS and almost reaching the result of a powerful discrete graphics card Radeon HD 5850. This game at medium quality settings on the A10-4600M works quite quickly, so it will even set the maximum quality.

Street Fighter IV is another multi-platform game based on the same engine. It belongs to the genre of fighting games, which differs from most others in that it requires at least 60 frames per second for a comfortable gameplay. But the game is old and graphically uncomplicated, so in all the test settings we chose a couple of years ago for the then laptops, such FPS is provided.

In this case, at minimum settings, almost all video cards provided acceptable performance, except for Zacate, and in the medium mode, even the weakest Radeon HD 5470M could not provide a comfortable frame rate change. But the AMD A10-4600M hybrid model turned out to be very fast again, although it lost to the system with the Mobility Radeon HD 5850 — after all, this is a discrete video card with much higher power consumption, albeit outdated. At 100 frames per second, this game will obviously be able to increase the quality settings on systems based on the Trinity APU.

Another old multi-platform game, but more demanding and even having DirectX 10 support is Lost Planet. In this performance test, AMD’s new solution once again performed very well, losing not so much to the much more powerful laptop from Acer. In Lost Planet, we only compared all solutions at low settings, since even they do not always provide high rendering speed on mid-range laptops.

In the Cave subtest, performance is limited by the speed of the CPU, and therefore an old laptop with a quad-core CPU wins much more in it than in the Snow subtest, which shows the speed of the graphics core. In the latest test, AMD’s new product is only 20% slower than the old discrete solution, and for the Trinity hybrid processor, this can be considered a decent success. On such a system, it will even be possible to set the settings for higher rendering quality while maintaining an acceptable FPS.

Temporarily finishing with multi-platform games, and moving on to exclusive PC games from the most common genres: RTS and FPS. The first on the list is the old real-time strategy World in Conflict:

And again we see a situation where at low settings the old solution with a quad-core processor outperforms our novelty more than at medium rendering quality settings. This is explained in exactly the same way as in previous tests — in the medium quality settings, the systems do not rest on the power of the central processors, and therefore the Radeon HD 7660G shows a good result between the mobile versions of the Radeon HD 5470 and HD 5850.

World in Conflict is quite CPU dependent, and only at medium settings do the tests show GPU speed. Tests have shown that on the A10-4600M hybrid-chip prototype we are reviewing today, it will be quite enough to increase the gaming settings above average in order to achieve a better picture while maintaining an acceptable frame rate. Moreover, even 30 FPS is enough for a real-time strategy. Let’s see what happens in first-person shooters, which are the most demanding on GPU power.

STALKER: Call of Pripyat is an example of a rather «heavy» game for GPUs, despite the fact that it is far from new. The maximum settings in it can bring even the most powerful video cards of desktop computers to their knees, to say nothing of mobile ones. It saves that the game’s graphics engine is perfectly scalable and customizable, and the lowest quality mode («static lighting») allows even integrated video cores to show a frame rate sufficient for a comfortable game.

In light mode, the rendering speed is again limited by the system CPU, so the prototype on Trinity is quite seriously inferior to a laptop with a very powerful Intel Core i7 processor. On average, in terms of the severity of the settings in the «full dynamic lighting» mode, the speed of all laptops is noticeably lower, and the Radeon HD 7660G in this mode lags behind it not so much, although the difference is still large. And in the case of a graphics-heavy game like Call of Pripyat, on systems with the new mobile APU, it will not be possible to seriously increase the graphic settings above average.

Far Cry 2 is a multi-platform project, but it features advanced graphics at the time of its release, significantly improved in the PC version. As we found out in previous times, integrated Intel graphics solutions and even the weakest discrete mobile video cards can hardly pull it — they do not provide playable FPS even at medium quality settings, not to mention high quality settings using DirectX 10.

And here is a powerful hybrid APU A10-4600M models are a completely different matter! A prototype mobile system based on this chip, which has a Radeon HD 7660G, showed quite good speed even at high settings with DirectX 10 enabled. Just think, modern integrated graphics will give a comfortable FPS in this game at these settings, providing more than 40 frames per second! In such conditions, the speed of the weakest solutions, including integrated graphics from Intel (up to Ivy Bridge), will not provide even 25-30 FPS.

And on a laptop with AMD’s new solution, you can even raise a few quality settings to even higher ones, getting a better picture and quite sufficient rendering speed. Or even turn on full-screen anti-aliasing, which until recently was not available even on low-end mobile discrete graphics cards.

Unfortunately, due to the dampness of the platform and drivers for it, Crysis Warhead, a very heavy game for video cards, did not start on the AMD Trinity prototype. So we’re jumping straight to another legacy game from our mobile graphics tests, DiRT 2, a racing game from Codemasters. This game supports DirectX 11 features such as tessellation and DirectCompute and includes a decent benchmark. Unfortunately, we did not test the ASUS K52Jr and the Zacate-based system in this game, so their results are not on the diagram.

But the AMD A10-4600M APU handles the task very well, at medium settings providing a more than acceptable rendering speed of 45 FPS. Although the gap to the system with the mobile Radeon HD 5850 is quite large — in our opinion, the APU lacks the most memory bandwidth, which limits the rendering speed in this game.

However, for an integrated video card, the result is still very good, and it makes it possible to try high settings, which we will try to do later when testing in the next part of this game — DiRT 3.

Let’s take a look at the last game from the outdated test set — another multi-platform project with a special improved PC version — Just Cause 2. An ASUS laptop with a Radeon HD 5470M, as well as a test system based on AMD Zacate did not participate in this comparison again.

Judging by the FPS numbers shown, Just Cause 2 is one of the most difficult gaming tests for not very powerful mobile video cards. Even at the lowest settings, a very powerful video card from the last generation gives only 60 FPS, and at high (not maximum!) Quality, it barely reaches the minimum performance level necessary to achieve playability.

But the Mobility Radeon HD 5850, which is part of the Acer Aspire 5943G configuration, still managed to show an acceptable frame rate with a high-quality picture, which our today’s hero, the A10-4600M chip with the Radeon HD 7660G, failed to do. In this game, systems with Trinity will have to be set to medium settings, since at high picture quality settings only 25 frames per second are provided, which is not enough for a normal game.

Although it is already possible to draw conclusions about the level of 3D performance of AMD’s new solution based on outdated gaming tests, these are still rather old projects that were released several years ago. And our testing would be incomplete without including the latest applications. And not in low and medium settings, but in more complex ones. To do this, we took a set of several modern games, tested them in high quality modes, and sometimes with the inclusion of DirectX 11 effects, MSAA full-screen anti-aliasing and even PhysX effects (in this case, implemented in software, of course):

So let’s look at the games one by one. Mafia 2 also didn’t work on the AMD Trinity laptop due to some incompatibility, and it would be interesting to see how the new APU handles this game when physical effects are enabled, because even when they are executed on NVIDIA mobile discrete graphics cards, the speed sometimes drops below comfortable minimum.

But we have another project with support for GPU PhysX hardware physics calculations — Batman Arkham City. At high settings, the average frame rate in the test on Trinity reached 45 FPS, which is very good for a mobile chip with integrated graphics, and when extreme quality settings were turned on, including tessellation and other DirectX 11 effects, the speed dropped to 22 FPS, which, although not playable, but an amazing result for such a chip (the latest discrete graphics card GeForce GT 640M had only a little more).

The inclusion of numerous PhysX effects in this game affects the speed even more, since the «physics» is processed by the central processor. And FPS in this case drops to 16, which is already noticeably lower than playability. But this is just a mobile solution with an integrated GPU and software PhysX, so even this performance for Trinity is an outstanding achievement.

Let’s move on to the game DiRT 3, the second part of which we have already considered a little higher. The third one differs little from its predecessor technologically, but we checked the high and “ultra-high” quality settings. The new AMD A10-4600M mobile APU, which has a Radeon HD 7660G video core, coped very well with high settings, providing more than 40 FPS, but the Ultra mode was not given to the new chip — 22 FPS cannot be considered playable performance.

Similar results were shown in the F1 2011 project based on the same game engine. This game is dedicated to the last season of Formula 1 and the new APU model from AMD at high settings can make it possible to play relatively comfortably, with an average FPS above 30. But in the «ultra» version, we again see only a little more than 20 FPS, which is clearly unplayable , but do not forget that this is integrated graphics!

Hard Reset has good graphics, but it is not too demanding on GPU power. And our today’s hero — a prototype laptop based on Trinity — showed good speed in this game: at medium settings, more than 30 FPS, at ultra-high settings — about 25 FPS, which is close to playability.

The second part of Lost Planet is even more GPU intensive and uses DirectX 11 features such as tessellation and DirectCompute. Therefore, in the high settings mode, including tessellation and other demanding effects, the performance of the AMD A10-4600M was clearly not enough, and the speed “dipped” to 12 FPS. And even at medium settings, the frame rate did not exceed 25 FPS, which suggests that Lost Planet 2 is one of the toughest 3D performance tests for the GPU.

Aliens vs Predator also uses new DirectX 11 features such as tessellation and compute shaders in post-processing and is quite GPU heavy, though not as heavy as the previous one. At low settings in the game on the test system with Trinity, the frame rate was obtained above 35 FPS, and at high settings, with SSAO and tessellation enabled, the rendering speed was again below the playability limit — just above 20 frames per second. However, here the discrete GeForce GT 640M scored only 30 FPS, so the result is excellent for the integrated video core.

The last modern game included in our tests was the popular project Crysis 2. The second part did not raise the bar of GPU power requirements too much, compared to the first one, and the built-in benchmark, although it uses tessellation and advanced DX11 effects, but even on the mobile chart shows quite good results. At Very high and Extreme settings, we got 22-29 FPS, which is again an excellent result for a laptop with an APU.

The performance figures obtained in modern games at «heavy» settings are impressive, especially against the background of other processors with integrated graphics and discrete video cards of past generations. In our tests, the hybrid AMD A10-4600M performed quite well — its performance level is noticeably higher than in the previous generation and will clearly be better than the next generation of mobile Ivy Bridge from Intel.

It’s not even about comparing average frame rates, but that AMD’s mobile hybrid chip, combining CPU and GPU, is for the first time able to provide playability at high quality settings in a large number of modern games. While the competitor’s integrated graphics are often unable to provide minimum playability even at low settings, not to mention medium and maximum.

And if the performance of the video core in the APU is still not enough, then soon mobile PCs will be offered that have both an APU and a discrete Radeon HD 7000 mobile graphics card, which will be able to work together on rendering, which will give even greater performance, as well as improve applicability of laptops in solving various problems.

Video data playback

In addition to the high frame rate in modern games, it is important for laptops that hardware acceleration of decoding of all formats is supported by the graphic video core, including the integrated one. Although even the simplest processors now handle this work in software, hardware decoding using specialized blocks in the GPU is much more energy efficient and can increase battery life, which is important for mobile solutions.

Our previous tests have shown that there are no problems with hardware acceleration of video data decoding on any GPU, even Intel integrated solutions do a good job, although video cores built into Intel processors still have some problems.

But we are not interested in Intel, but in the new APU from AMD. Let’s check what the A10-4600M does with video decoding in practice. For tests, we took one MPEG-2 file with interlaced Full HD, one high-resolution VC-1 file, and a set of clips of the most common H.264 (MPEG-4 AVC) format with different resolutions and bitrates.

H.264 480p 5% 5% 5% 6% H.264 720p 5% 13% 10% 10% H.264 1080p (20 Mbit/s) 5% 5% 5% 6% H.264 1080P (40 Mbps) 6% 6%

6% 5% 7%

Modern GPUs have long been able to cope with MPEG2 acceleration, except when post-processing is required to eliminate interlacing (deinterlacing — deinterlacing). It is this clip that is included in our test set, and some lag of laptops with Radeon graphics cores (including those with the new APU) in the case of an MPEG2 file is explained by a better deinterlacing algorithm. However, the test file played perfectly on all systems, including our today’s hero — a prototype system based on Trinity from AMD.

When decoding a VC-1 video, the AMD A10-4600M is also fine, which cannot be said about an Acer laptop that uses a video core built into an Intel Core processor with Sandy Bridge architecture, which cannot decode video in VC-1 format in hardware (according to at least in the MPC-HC player). And in general, the new APU did a great job with all the videos. The H.264 format in any of its manifestations succumbed to the A10-4600M very easily, the GPU copes with videos perfectly, with approximately the same CPU load.

When playing all videos, DXVA acceleration works efficiently, and now almost any integrated mobile video core can handle HD video decoding even in the case of the heaviest videos with maximum quality and bitrate. But how efficient is video decoding on the Trinity APU? Let’s check it out by measuring the battery life in different modes.

Battery life

Before looking at the capabilities of AMD’s prototype laptop, we need to remember that its configuration includes a rather large screen and an optical drive, and a lithium-polymer battery has six cells with a capacity of about 56 Wh — this is the average level. The manufacturer claims a maximum battery life for Trinity-based laptops of over 11 hours, but this number is explicitly stated for idle mode.

Let’s take AMD’s word for it, because we didn’t check the idle mode when the maximum power saving profile was enabled, because we simply don’t see any point in it, because you need to work on a laptop, and not just leave it to devour the battery. And if it is not needed, then let it go to sleep.

The first test mode is the active reading (or Internet surfing) mode with the MP3 audio file player turned on in the background, and the second is the rather popular H.264 movie viewing mode with DXVA acceleration enabled. The power saving profile in these two modes was «balanced» — which is the default and installed by most laptops. 2:29 1:43

Recall that the Acer Aspire 5943G model has a much larger battery (83 Wh versus 56 Wh for our today’s hero), the Acer M3 has almost the same the same capacity, while the ASUS laptop has a smaller one (48 Wh). You can clearly see the difference in the time of release of laptops. Even the most capacious battery did not help the old top model Aspire 5943G, and in read mode it worked for very little time.

AMD’s A10-4600M prototype laptop showed a very good read time of more than 7 hours, coming close to the very good result of Acer’s gaming ultrabook, which used an Intel Core i5-2467M APU with a much lower TDP. That is, low-power models of the Trinity platform, like the A6 and A4, will show the result even better. AMD’s power reduction technologies have proven to be very effective.

When viewing hardware-decoded H.264 video, the systems did not last as long, but the difference between the solutions is about the same. Although almost all laptops allow you to watch two hours of video on battery power (except for ASUS with a weak battery), only the Acer Aspire Timeline Ultra M3 and the prototype on AMD A10-4600M were able to provide about 5 hours of video viewing in such conditions.

Let’s see what happens in the maximum gaming load. As a «load» 3D application, we previously used the benchmark built into the game Lost Planet, which is quite heavy on both the CPU and GPU, and its playback is looped, which is great for our task. We checked not only the battery life in performance mode, but also the resulting rendering speed: 75.1 24.8

And when the discrete video core came into operation in the Acer gaming ultrabook, we saw another advantage of our today’s hero — the Trinity platform. In this case, the A10-4600M delivers maximum uptime at slightly lower performance than a clearly more powerful solution.

And outdated laptops are the best indicator of progress. The Aspire 5943G, even with a noticeably larger battery, did not last that long, and the performance in the game Lost Planet and the top model of the new APU turned out to be quite sufficient, and in terms of battery life, the prototype from AMD completely became the winner of the comparison — an excellent result for Trinity!

Although even cost-effective solutions like the AMD A10-4600M won’t let you play offline on your mobile PC for even a couple of hours, so demanding 3D games on laptops without a power outlet nearby still won’t last long.

Conclusions

With the release of Trinity, AMD has continued its «hybrid» strategy started in Llano and Zacate. While huge jumps in performance were not expected due to the lack of progress in the process technology used, the CPU and GPU parts in the new APUs received a decent increase in performance and efficiency compared to the previous generation. Although in terms of universal computing on the CPU, the AMD solution may lag behind the modern solutions of its competitor (we are talking about future mobile Ivy Bridge), but the speed of the graphics core in Trinity will clearly remain the highest in the class.

With the new Trinity series, AMD continues to take a different approach to balancing CPU and GPU speed compared to Intel. And even the release of 22 nm competitor solutions with the latest video core of the HD 4000 model will not be able to help them get ahead of the corresponding Trinity models in terms of consumption. AMD hybrid chips will continue to win in graphics tasks, although the competitor has clearly come closer due to the release of chips based on a more advanced technical process, with which we will compare Trinity in future materials.

Of particular note is the increase in the number and quality of applications that use the capabilities of graphics cores in general-purpose computing. If at the time of the release of Zacate and Llano we noted that there were no such applications at all, now they have already appeared. Moreover, it concerns not only and not so much the usual applications for processing video data, but also archivers, graphics packages, etc. Although the ideal has not yet been achieved, it will be interesting to see how the situation develops further. In any case, we note the clear progress of AMD solutions in supporting GPGPU calculations already in real applications — here they also have a clear advantage over their competitor. And further expansion of the use of OpenCL in software will only strengthen the position of the company.

As for the architectural changes in the composition of the Trinity blocks, here we note that the improvements in the Piledriver cores clearly benefited the new APU. In the case of the desktop solutions of the AMD FX line, it was very difficult for them to compete, and in Piledriver, the computational efficiency was clearly improved. And while AMD couldn’t improve Trinity’s performance as much as it could by switching the chips to a «thinner» process technology, the use of modified x86-compatible computing cores definitely gave them an increase in speed.

Switching to a more advanced process technology would have given an even greater increase in performance, but even in this form, Trinity is a very well-designed platform that squeezes all the juice out of the available 32 nm. In addition to improvements in CPU cores, which led to an increase in computational speed, it should be noted the use of a more efficient VLIW4 graphics architecture, which allowed a significant increase in speed in 3D tasks with a similar complexity and chip size, compared to Llano.

Even if Trinity doesn’t break the speed records for general-purpose computing on x86-cores, in the released APUs it is quite enough for most applications. Much more important is efficiency and battery life, and another strong point of the released Trinity mobile hybrid chips is very good energy efficiency. The battery life of the tested prototype was very good, and in a 3D game it was outstanding. At the same time, we tested not the most economical option from the line of new APUs. And we can say for sure that, compared to Llano, it turned out to be a clear step forward, and in terms of energy efficiency, AMD solutions will be competitive even compared to the latest 22 nm Intel processors.

In general, in comparison of two giants: AMD and Intel, the result remains the same. If Intel has some advantage in terms of CPU performance, which also uses the fact that it has its own chip factories that quickly switch to newer technical processes, then AMD has an advantage in terms of power and functionality of graphic solutions — their APUs have clearly the best opportunities in gaming applications. The new hybrid chip from AMD proved to be able to provide acceptable performance in a large number of modern games at high quality settings.

Yes, Intel has a partnership with NVIDIA, and the use of discrete graphics in addition to the integrated graphics in the CPU solves some of the problems. But the advantages of AMD are not only the high speed of the integrated GPUs, they are also able to simultaneously use the power of the integrated and discrete graphics of the new generation, getting even greater speed — AMD Radeon Dual Graphics technology is responsible for this.

As part of the material, it remains for us to consider the price issue. And so far, not everything is clear. Simply because the real entry of solutions to the retail market can change a lot — after all, the cost of the final product depends on the price of many of its components, and although APU is one of the most important, it is only one. The Trinity seems to be the best fit for laptops like the prototype we got for testing — its 14-inch chassis packs enough power for most tasks, even gaming. Moreover, we are talking about most of the demanding modern games.

At the same time, this laptop is small in size, relatively light and has a decent battery life. And the price of such solutions promises to be not too high — lower than that of the same ultrabooks, for example. Which, although smaller, are less powerful. On the other hand, there are more powerful solutions, like the gaming ultrabook we recently tested with a discrete NVIDIA GeForce GT 640M graphics card — they are faster, but also more expensive, and consume more energy. Yes, and we are promised the release of hybrid systems with integrated and discrete graphics from AMD, which will use advanced switching between GPUs, similar to NVIDIA Optimus.

We don’t have enough retail pricing information for Trinity-based laptops and competing Intel solutions to draw any final conclusions. Indeed, from the point of view of a potential buyer, it is the price that is the most important characteristic of any product. We are confident that AMD and its end-to-end partners will be able to offer competitive prices for mobile PCs based on the very good Trinity platform chips. Notebooks based on the AMD A10 platform are expected to sell for around $700, lower than Intel Ivy Bridge-based ultrabooks expected around the same time. And at the time of release, the new APUs will provide a great combination of features and performance for the money.

AMD A10-5800K processor review and testing GECID.com. Page 1

::>Processors
>2012
> AMD APU A10-5800K

15-10-2012

Page 1
Page 2
One page

October of this year can rightly be called “AMD month”. After all, it was during this period of time that the world-famous AMD company planned to introduce two new series of processors. And if the chips code-named Vishera for the Socket AM3+ platform are still waiting for their «high point», then the new generation APU Trinity for desktop systems has already seen the light and appeared before the public.

Let’s take a look at the second generation Trinity APU for PC using the most powerful processor in the series, the AMD A10-5800K . But let’s start, perhaps, not with him, but with the features of the architecture of the new processors.

In general, the concept of APU is not new and has already firmly settled in the lexicon of many users. For those who do not follow the state of the IT market, we recall that the APU (Accelerated Processing Unit) is a processor with a powerful integrated graphics core. First of all, this concept was attractive to manufacturers of laptops and netbooks, allowing you to increase the possibilities of mobile solutions. APUs for desktop computers were also appreciated. These APUs are known to us under the name Llano, operating on the Socket FM1 platform. The next step in the development of hybrid processors was the Trinity APU generation and the Socket FM2 platform, the capabilities of which we will consider today.

In fact, the APU structure assumes the presence of three main components: a processor computing unit, an integrated graphics video core, and an integrated northbridge. This concept was also used on AMD’s first generation of desktop APUs, the Llano APUs, and the new Trinity APUs are virtually no different in this regard.

In the maximum configuration, Trinity processors can include up to four x86 cores. Moreover, the computing unit itself is based on the new Piledriver architecture, which is a further development of the Bulldozer architecture. Here, as in the Llano APU, the “old” 32-nanometer process technology is also used. But compared to the previous generation of APUs, the Trinity APU’s die area has increased to 246 square meters. mm (in APU Llano it was 228 sq. mm). This, in turn, made it possible to increase the number of transistors to 1.3 billion. According to AMD, the video core integrated into the Trinity chip corresponds to the generation of AMD Radeon HD 7000 video cards.

Above is a schematic diagram of a quad-core Trinity processor. As you can see, almost half of the chip is occupied by the integrated video core, which has the same architecture as the AMD Nothern Islands family of graphics processors. A separate block is the combined northbridge, which is a link for other components of the chip. As a computing unit, the same so-called «dual-core modules» are used, each of which contains two computing devices capable of processing two data streams simultaneously. At the same time, dual-core APU Trinity models will be equipped with one such module, and quad-core models, respectively, with two. 24 PCI Express lanes are used to communicate with external devices.

Among the innovations, we note the presence of a new dual-channel memory controller and an AMD HD Media Accelerator video encoding unit. It is also worth mentioning the support for HDMI, DisplayPort 1.2 and DVI interfaces.

But, of course, the main «highlight» in the Trinity APU is the use of the new Piledriver architecture, which is an upgraded version of the Bulldozer. Let’s see what changes are present in the new architecture.

First, the branch prediction block, a device that determines the direction of branches (predicting whether a conditional branch will be taken) in an executable program, has been improved. Secondly, the schedulers of integer (Int Scheduler) and real (FPU Scheduler) execution units have been improved. The main task of these schedulers is to distribute commands to execution units as they are ready.

Other features worth noting are improved L2 cache efficiency, increased L1 TLB buffer size, support for new F16C and FMA3 instructions, as well as an increase in the speed of execution of some basic processor instructions, such as INT / FP divide, SYSCALL / SYSRET .

As you can see, the difference between the Piledriver and Bulldozer architectures is quite noticeable, but how much it will affect performance, let’s see from the test results.

Naturally, we could not ignore the built-in graphics core, which is codenamed Devastator. Moreover, the main bet AMD made on its hybrid processors lies precisely in the capabilities of the integrated video core.

The Devastator graphics core is based on the VLIW4 architecture, although the earlier Llano APU used a VLIW5 architecture core. To what extent such a move is justified, we will find out a little later during testing. For now, let’s just give some comparative numbers for quad-core APUs of different generations:

§ the number of computing units in the video core of APU Trinity 384 versus 400 in APU Llano;

§ 800 MHz APU Trinity graphics core operating frequency versus 600 MHz in Llano APU.

The Devastator graphics core fully supports DirectX 11, OpenCL 1.1 and DirectCompute 11. In addition, thanks to Eyefinity technology, it is possible to connect four image output devices. Also, the APU Trinity implements the Dual Graphics function, which allows you to combine the power of integrated and discrete video.

Well, we will probably not torment our readers with a description of the technical side of the new Trinity APUs, but will offer a direct look at their results, as they say «in action». To do this, we will use the presentation materials provided by AMD.

Naturally, AMD’s competitor in the processor market is Intel. Therefore, it is not surprising that the models of this particular company were taken for comparison. Here you see the use of OpenCL features in practice. The difference is especially noticeable when working with graphics.

From this slide, you can see that the use of the integrated video core on the VLIW4 architecture was a completely justified move from a technological point of view. An increase of 37% in the program for testing video accelerators 3DMark 11 speaks for itself.

It is clear that many consumers will be primarily interested in performance not in synthetic tests, but in real gaming applications. This collection of slides showcases performance improvements in some popular games. Particularly interesting are the results of comparing APU Trinity with a bunch of Intel Core i5-3450 + NVIDIA GeForce GT 630 2 GB DDR3 (discrete graphics card). The last slide gives a clear idea of ​​the advantage of working in Dual Graphics mode.

The above is a demonstration of using Eyefinity technology. With the release of the Windows 8 operating system, which supports the new Metro interface, Eyefinity technology will become even more relevant.

Model range

At the moment, the Trinity processor series is represented by six models so far. As in the case of the Llano APU, the manufacturer has divided the new hybrid processors into classes depending on their performance. The Trinity processor family for personal computers, as well as some of their characteristics, are as follows:

Model APU

A10-5800K

A10-5700

A8-5600K

A8-5500

A6-5400K

A4-5300

Brand AMD Radeon

HD 7660D

HD 7660D

HD 7560D

HD 7560D

HD 7540D

HD 7480D

Thermal package (TDP), W

100

65

100

65

65

65

Number of processor cores

4

4

4

4

2

2

Processor clock speed (max/base), GHz

4. 2 / 3.8

4.0 / 3.4

3.9 / 3.6

3.7 / 3.2

3.8 / 3.6

3.6 / 3.4

Number of cores AMD Radeon

384

384

256

256

192

128

GPU clock frequency, MHz

800

760

760

760

760

724

Level 2 cache (L2), MB

4

4

4

4

1

1

Maximum supported DDR3 memory speed, MHz

1866

1866

1866

1866

1866

1600

Recommended retail price, USD

122

122

101

101

67

53

The flagship A10 series is represented by two models A10-5800K and A10-5700, there are no architectural differences between them, both contain 4 cores and 384 stream processors. The only difference is that the first processor has an increased clock frequency of both the computing module and the video core, and also has an unlocked multiplier. This, in turn, led to a different thermal package for these two models.

As regards the A8 class processors, which include the A8-5600K and A8-5500 models, here the changes affected mainly the graphics part. The processors retained all the same 4 cores, only the frequency of their work was slightly reduced compared to the APU A10. Just like in the flagship A10 series, here one processor is focused on overclocking and has an unlocked multiplier, and the second is designed to work in quiet multimedia systems with reduced heat dissipation of 65W.

The A6-5400K and A4-5300 processors close the lineup. Both APUs contain 2 cores each, while the total amount of L2 cache has dropped to 1 MB. The number of GPU shader processors has also been reduced.

Appearance and packaging

Well, now it’s time to go directly to testing one of the new generation Trinity processors — AMD A10-5800K . A tray version of the processor came to our test lab, so it is difficult to describe the packaging and the cooling system. Let’s move on to the appearance of the APU itself.

Externally, the processor is absolutely no different from its predecessor, made for the Socket FM1 platform. On the heat-distributing cover of the processor there is a marking and the name of the manufacturing country, in this case Malaysia.

Back side of the Llano series processor (Socket FM1)

Back side of the Trinity series processor (Socket FM2)

And if a different number of processor «legs» is not visually noticeable (904 in the Trinity APU versus 905 in the Llano APU), then a different arrangement of the «keys» is still striking. Therefore, processors for the Socket FM1 and Socket FM2 platforms are not compatible.

That is, the transition to a new generation of hybrid processors actually forces you to change the motherboard, and it does not have to be based on the new AMD A85X chipset, but the well-known less expensive AMD A75 and A55 can also be used. Anticipating the outrage of some users about this, AMD announced that the next generation of APUs will be based on the same Socket FM2 processor socket.

Specification

Model

AMD A10-5800K

Marking

AD580KW0A44HJ

Processor socket

Socket FM2

Clock frequency (nominal), MHz

3800

Maximum clock speed with Turbo Core 3.0, MHz

4200

Multiplier

38

Bus frequency, MHz

100

L1 cache size, KB

2×64 (instruction memory)

4×16 (data memory)

L2 cache size, KB

2×2048

L3 cache size, KB

Core

Trinity

Number of cores/threads

4/4

Instruction support

MMX, SSE, SSE2, SSE3, SSE4, SSE4A, x86-64, AMD-V, AES, AVX, XOP

Supply voltage, V

0. 825 — 1.475

Power dissipation, W

100

Critical temperature, °C

74

Process

32 nm

Technology support

Dual Graphics

UVD3

Turbo Core 3.0

PowerNow!

Eyefinity

Built-in memory controller

Maximum memory, GB

64

Memory types

DDR3 (up to 1866 MHz)

Number of memory channels

2

Radeon HD 7660D integrated graphics

Stream Processors

384

SIMD

6

Texture blocks

24

Rasterization modules

8

GPU clock frequency, MHz

800

Instruction support

DirectX 11 (Tessellation, ShaderModel 5. 0)

DirectCompute 11

OpenCL 1.1

The CPU-Z program confirms the information written above in the table. As you can see, in normal operation, the AMD A10-5800K clock speed is 3800 MHz, while the nominal voltage is 1.464 V.

In the dynamic overclocking mode or simply «auto overclocking» using Turbo Core 3.0 technology, the processor speed increases to 4200 MHz.

The AMD A10-5800K cache is distributed as follows. L1 cache: 16 KB for each of 4 cores is allocated for data with 4 channels of associativity, while there are 64 KB for instructions for each dual-core module (recall, there are 2 in a quad-core processor) with 2 channels associativity. L2 cache: 2 MB per dual-core processor module with 16 channels of associativity. There is no L3 cache.

The DDR3 memory controller operates in dual-channel mode and is capable of supporting RAM up to DDR3-1866 MHz.

Graphics core specification is fully confirmed by the GPU-Z utility. As we said above, the Radeon HD 7660D has 384 shader units and runs at 800 MHz.

AMD Trinity (A10-4600M) Overview | Reliable Reviews

AMD has long played a second role to Intel when it comes to providing the chips that power our laptops. Simply put, Intel has almost always offered the best balance of battery life and performance. AMD introduced its chips last year, codenamed Llano, which had best-in-class graphics performance but still lagged behind when it came to real-time battery life and CPU performance. Now the company is back with its latest chips, code-named Trinity, and the claims about its capabilities are really bold. The Radeon HD 7000 graphics should not only comfortably beat both Intel’s existing Sandy Bridge-based chips and the latest Ivy Bridge mobile lines, but also have excellent battery life and processor performance.

We spent some time on a test laptop based on AMD’s new A10-4600M chip to see how it stacks up in the real world. We’ll check it out later, but first, a little more about Trinity.

AMD Trinity Architecture
AMD Trinity is the codename for the company’s new line of things it calls APUs (accelerated processors). AMD uses this terminology because unlike older processors, these new chips contain not only the main processor, but also a GPU (which can itself be used for computing tasks other than graphics), a memory controller, and a number of dedicated modules to speed things up. like video decoding. It is technically still the central processing unit (CPU) of the entire system, but it also does a lot more.

Intel has done a very similar job with its chips, which is why we can see laptops as thin, light, and powerful as they are today — it just doesn’t treat them as APUs so readily.

Trinity chips will combine two or four CPU cores with up to 384 graphics cores on a single piece of silicon, resulting in 1.330 billion transistor chips. This is just a fraction more than Llano (1.170 billion), while the die size has also increased from 228mm^2 to 246mm^2. 2 with 1.4 billion transistors.

Trinity is built using AMD’s existing 32nm manufacturing process, just like with its Llano chips, but despite this, it managed to double the performance per watt while both CPU and GPU performance increased to double digits digits. That 32nm is roughly the same as Intel’s Sandy Bridge generation, but behind the now just-released Ivy Bridge chips that use the smaller and more power-efficient 22nm process.

AMD Piledriver
Trinity’s beating heart is the Piledriver CPU core that replaces the star-based Llano design. A direct comparison is a little trickier, as each Piledriver «module» contains what is roughly the equivalent of two normal cores, though not quite. The chips will ship with one or two modules, which are either dual-core or quad-core processors.

We won’t dive too deep into the architecture here (Anandtech I did a great tech deep dive if you’d like to read more), but the key is that it’s an almost brand new design that offers both performance and power saving benefits. However, AMD readily admits that it still can’t compete with Intel for raw CPU performance for any given design power (TDP is the maximum power/power the chip expects to deliver), and relies on its chips to provide «enough» performance at that areas, supporting it with excellent graphics and multimedia, as well as good battery life (low power consumption).

This is a reasonable logic that certainly reflects the way we use our time. After all, most people spend most of their time on a laptop doing relatively idle tasks like web browsing, recording and watching videos without doing intensive computing or heavy multitasking. Often these are only games that require a CPU/GPU to add weight. We’ll talk about this a little later.

Hidden in silicon are a number of other tweaks and improvements that should result in lower power consumption and better performance, but one of the most notable is the new AMD Turbo Core 3.0 technology.

Turbo Core existed on Llano where it dynamically adjusted the CPU clock speed based on the workload to get the best performance when needed without overheating the system. But it only ran on the CPU, not the GPU. With Turbo Clock 3.0, the chip can increase or decrease the clock speed as needed. The algorithm for this has also been vastly improved, so that the chip really gets the full performance hit where possible.

On the other hand, during idle moments, the chip can also turn off almost completely to extend battery life, as AMD claims excellent standby battery life from Intel.

AMD Radeon HD 7000 Graphics
The key to the appeal of this new AMD Trinity line is its graphics capabilities. Each model in the line has a slightly different configuration, unlike Intel’s Sandy Bridge (Intel HD 3000) and Ivy Bridge (Intel HD 4000) chips, which basically use identical configurations in each range. While this does make for some confusion, one thing has to be true: we expect most of these AMD configurations to outperform both Intel’s Sandy and Ivy Bridge. We’ll have to wait for low power products to hit the market before we can test that out. Also, we’ll see if this results in really playable frames in demanding games, but at least less stressful games should run better.

What makes these parts potentially really exciting is that they can be paired with additional graphics chips from the AMD 7000 series of mobile GPUs. If they are configured on the same system, they can work together like Crossfire / SLI on desktop. We haven’t had a chance to test it here, but the figures provided by AMD show almost doubling the performance compared to the APU. Meanwhile, when the system is idle, the additional GPU consumes almost no power, so should have minimal impact on performance. Again, we have yet to test all of this, and there is no guarantee that we will see really thin and light laptops with this kind of power.

As for the GPU itself, it’s very similar to the latest AMD desktop graphics cards in terms of fundamental design, but with less of everything. At the most, it will contain 384 cores, with the lower end pieces dropping to 192 cores. These will be accompanied by up to 24 texture units and 8 ROPs, making the high-end part roughly equivalent to ¼ of an AMD Radeon HD 6970 desktop card. This is a chip unit designed to speed up certain tasks such as video decoding — something that is supported in all major browsers, Windows Media Player and VLC — and encoding.

AMD is also touting its support for OpenCL, whereby applications such as the above, Photoshop CS6, get increased performance due to intensive tasks performed by the GPU rather than the CPU.

Also includes AMD Quick Stream technology. This prioritizes media traffic for smooth video streaming. Obviously, this depends a lot on your network connection, but if you have other applications accessing the internet, this technology ensures that they don’t interfere with your browsing.

AMD A-Series APUs
The new Trinity APUs will be available as the A-Series, with models ranging from the most powerful A10 to A6. This is a little confusing, as the previous generation chips were also called A-series, so you’ll have to check the full model numbers to distinguish — Llano used Axx-3000 style numbering, Trinity will use Axx-4000 style numbering.

Within the numbering differences there are also some key differences. Chips will be available in two main configurations depending on the type of mounting package used: PGA or BGA. The standard APU set uses the classic Prid Grid Array (PGA) scheme, in which the chip is placed on a mount that has hundreds of pins on the underside for connecting to the motherboard. At the same time, the Ball Grid Array (BGA) uses tiny metal balls instead of pins to make the package even smaller.

There was also a differentiation based on chip TDP in each package type, with 35W chips available for the larger PGA package and 25W and 17W for the BGA package. As you might expect, the former will find their way into larger, more powerful laptops, while the latter will be used in thinner, even Ultrabook-style, form factors.

Each configuration also has slightly different graphics capabilities. The full list of available versions is below.

AMD Trinity Test Platform
So, really, how do these new chips work in a real laptop?

We were given a test laptop running the top-end APU A10-4600M. It’s been crammed into a 14″ chassis, but far from being a sleek and portable piece of desire, it’s a rather utilitarian kind of largely plastic slab that’s over an inch thick. However, it is only intended for testing the base hardware and, in addition, it has a Blu-ray drive, VGA and HDMI video outputs, two USB 3.0 ports, one USB 2.0 port, an ExpressCard slot, a memory card reader, Ethernet, and connectors. for headphones and microphone. It also has 4GB RAM and 128GB SSD. Think about how much space you could save by removing the Blu-ray disc and a few of those ports, and you could possibly make a relatively thin machine with this basic hardware configuration.

The combination of a high-end chip and an SSD means this machine should outperform a lot of equivalent-sized laptops and may not be a bit of a reflection of a system you’re likely to encounter, but still gives us a strong real-world indicator. performance. It literally uses a 4400 mAh battery to power the system, which is typical for this size machine.

Now finding a system to compare this system to was a little tricky as few systems contain the same balance of components, but we chose what we consider to be a reasonably representative set. We have the low-power Sandy Bridge chip used as a typical Asus ZenBook UX31 ultrabook, a more direct competitor with the faster Sandy Bridge in the Lenovo ThinkPad X220, and finally we opted for the ultra-fast quad-core Sandy Bridge powered HP Pavilion dv7-desktop laptop. 6b51ea Beats Edition, which also features a dedicated AMD Radeon HD 649 graphics card0M.

We admit that we just couldn’t find an Ivy Bridge test system to compare with during this review, but we will of course return to this topic when we get our hands on an Ivy Bridge system. Having studied various other sources, we are also well versed in what we are most likely to see for the CPU, GPU and, to a certain extent, for the battery of such systems.

Here’s the full comparison list:

So without further ado, for testing!

Processor performance
The first of our tests was the classic Cinebench. This is a reliable measure of net CPU utilization that fully taxes both single-core and multi-core processors.

It’s immediately obvious what the shortcomings of the AMD A10-4600M are. With a Cinebench CPU score of 2.03, it’s almost on par with the low-power Zenbook Sandy Bridge chip, but is extremely concerned about the more powerful HP. As for Ivy Bridge, it’s not surprising that the results we’ve seen are

suggest he pulled out even more HP lead here with 6.5 points that were centered around.

We also ran a general system performance test, PCMark 07, which showed AMD trailing both Lenovo and Zenbook — HP has a low score because it uses a hard drive rather than an SSD. The Ivy Bridge scores for PCMark07 we’ve seen are in the 6500 region, again showing just how fast this chip can be.

However, the key point here is that AMD has really got it right in terms of everyday use. Subjectively, the A10-4600M feels fast enough, delivering smooth web browsing, smooth video playback, and sufficient performance. After all, ultrabooks like the Asus Zenbook are considered more than usable, and the A10 is on par.

GPU performance
Turning to gaming, we ran our classic laptop benchmarks on this and our other benchmarks, and the results are astounding, to say the least. In the comparatively undemanding TrackMania Nations Forever A10 test, it’s a great solution with AMD’s dedicated graphics card on HP, and it’s well ahead of both the Lenovo X220 and the Zenbook. But TrackMania, with the settings that we run, is so undemanding that it actually becomes limited due to the speed of the processor, and it is only in Stalker: Call of Pripyat that we see the true picture. Here the A10 has a 60 percent advantage over the HP Radeon HD649 graphics card.0M, and the two Sandy Bridge systems are far behind.

Does this result in a really enjoyable real world gaming experience? Well, it’s close. Despite the Stalker’s impressive 40 fps, it only runs at medium detail settings and a resolution that’s lower than the pixel count on your average laptop. Performance should probably double again before we actually get to a gaming machine for even the most recent games.

However, less demanding favorites like Counter-Strike: Source, World of Warcraft, and Call Of Duty should all prove playable.

Battery life
Perhaps the biggest shock, at least according to our tests, is how little power the A10-A4600M consumes, or more specifically, how long the battery on this test platform lasts. Despite housing the most powerful chip in AMD’s new lineup, it comfortably beats both the Zenbook and the HP gaming machine, delivering almost extra usage time despite testing the smallest battery. Only the X220 surpassed it, but this laptop has 60 percent more battery.

Our test uses the industry standard MobileMark Productivity test, which simulates a fairly typical usage pattern by a user who edits some documents, surfs the web, creates PowerPoint, and watches videos, all punctuated with pauses to simulate a sitting and thinking user. We found this to be a very good indicator of battery life for average usage, but there are other extreme cases where it’s less indicative, particularly if you’re watching videos — like on a long-haul flight — or playing games.0003

As for the latter, it just doesn’t make sense to check battery usage, since the GPU on any machine drains power very quickly. That said, watching videos is something other machines can do for hours on end, and this is where the Trinity lags a bit, with the test platform delivering around 3.5 hours of h.264 video playback, compared to 5.5 on the Zenbook. But of course, the argument here is that this is a higher power chip, and lower power triplets can still improve that.

What we can say for sure is that AMD is definitely not far behind Intel in terms of battery life.

judgment
Key questions remain to be answered about the AMD Trinity platform, such as how lower performance components will evolve and what systems we will see with new chips. Not to mention how much these systems will cost. But from the evidence here, it looks like AMD has done most of what it can. Processor performance still lags behind both Intel’s Ivy Bridge and Sandy Bridge, but most importantly, it’s perfectly adequate for everyday computing. Meanwhile, GPU and gaming performance is class-leading and battery life is at least on par. Is this the dawn of a new era of AMD dominance for laptops? Perhaps not, but we should no longer be limited to choosing one.

Unlike other sites, we thoroughly test every product we review. We use standard industry benchmarks to properly compare features. We will always tell you what we find. We never, ever accept money for a product review.
Tell us what you think — send your letters to the editor.

AMD Trinity Features and Benefits

Mobile AMD Trinity processors are designed to compete with Intel’s Ivy Bridge chips. CHIP will talk about the key features and test results of the new CPUs. AMD prefers to refer to its processors with integrated graphics as APU (Accelerated Processing Unit). Unlike Intel, AMD’s strategy is to increase the performance of the integrated graphics architecture in the first place, and not the processor cores. AMD provided us with a Trinity-based laptop with a new A10-4655M processor, which during testing showed amazing graphics power.

Features and Benefits

Like previous LIano chips, AMD Trinity is manufactured using the 32nm process technology. Nevertheless, compared to its predecessors, the performance of the new CPUs has increased, while the level of power dissipation has decreased. At the time of the release of the new platform, AMD introduced two dual-core mobile processors for ultra-compact laptops — the A10-4655M and A6-4455M models. For multimedia notebooks, the company introduced three new solutions: two quad-core A10-4600M and A8-4500M, and one dual-core A6-4400M. The unified media interface (Unified Media Interface) of these processors connects the processor cores, graphics chip and cache memory. It also contains a dual-channel memory controller. In addition, all new AMD Trinity processors support Turbo Core 3.0 auto-overclocking technology, which, like Intel’s Turbo Boost technology, is designed to provide end devices with the best balance of power consumption and performance.

AMD Trinity processor specifications. AMD Trinity chips contain up to four CPU cores and 384 Radeon 7000 series GPU compute cores.

Powerful graphics, Eyefinity, OpenCL

Each A-series processor has its own integrated video core model (see table above). What unites them is that they all belong to the Radeon 7000 series of graphics solutions. The integrated graphics chip supports DirectX 11, Shader Model 5.0 and OpenCL 1.1. In addition, there is support for Hybrid CrossFireX technology. At the same time, the integrated graphics core and a discrete video card are able to dynamically distribute video processing tasks among themselves. If you use multiple displays at the same time, then you will surely be pleased with the presence of Eyefinity technology, which allows you to connect multiple monitors to a PC or laptop. All A-series processors also have a built-in UVD3 (Universal Video Decoder 3) video decoder by default. This component reduces the load on the CPU during video playback, providing a lower
power consumption. Another significant advantage is that the new chips support the OpenCL 1.1 programming language, which allows you to significantly increase the speed of photo and video editing.

Graphics performance in games. The integrated Radeon HD 7660G graphics core of the new AMD A10-4600M processor handles even the H.A.W.X game with ease. 2(1). But in the demanding Metro 2033 (2), to get an acceptable level of FPS, the quality of the graphics had to be reduced.

Benchmark: AMD Trinity vs. Intel Ivy Bridge

AMD places its A10-4600M processor on par with Intel’s Core i7 mobile chips in terms of performance. However, the test results tell a different story: today, in the fight with Intel, AMD still has little chance. The Intel Core i7-3610QM processor demonstrates in benchmarks two to four times faster than the A10-4600M. For example, in 3DMark Vantage, a laptop based on this solution earned 19,968 points versus 10,116 points for a similar laptop based on the AMD A10-4600M CPU. However, in terms of graphics power, the benchmark results show that the Radeon HD 7660G video chip in the A10-4600M processor is noticeably ahead of the integrated Intel HD Graphics 4000 graphics core in the Core i7-3610 QM. For comparison, in the 3DMark 06 benchmark, a laptop based on the new AMD chip scored 22,626 points, while a laptop based on an Intel processor scored only 21,058 points. And in the system-demanding game H.A.W.X. 2 with support for DirectX 11 technology, the AMD A10-4600M graphics chip outperformed its rival by almost 30%. This means that if you own a compact laptop and are satisfied with the average level of graphics quality in modern games, the AMD A10-4655M processor will allow you to save on a discrete video card.

Benchmark Results Comparison Graphics Test The Bottom Line: Two Sides of the Same Coin

The video subsystem of the new AMD Trinity processors delivers unmatched performance for integrated graphics. If we talk about the power of processor cores, then AMD Trinity is significantly behind Intel’s solutions based on the Ivy Bridge microarchitecture. However, the «A» series chips clearly demonstrated that AMD is moving in the right direction. The engineers of this company are trying to succeed in the segment where they have the most chances, namely in the field of graphics processing. A direct comparison of the two integrated solutions showed that the Intel HD Graphics 4000 video core is inferior in performance to the top version of the integrated Radeon graphics chip of AMD Trinity processors. In addition, the presence of UVD3 and a significant acceleration of programs based on OpenCL are additional advantages in the AMD piggy bank. It remains to be hoped that software developers will make better use of the available functions in the future. But AMD treated the optimization of the processor unit lightly. Therefore, even the most powerful chip in the A10-4600M line is significantly inferior to the modern third-generation Core i7 from Intel. At best, it can be compared to the old second-generation Intel Core i5 processor on the Sandy Bridge microarchitecture.

Test system

The test involved a laptop that is used by AMD as a prototype and will never go on sale. It is equipped with an AMD A10-4655M processor, 4 GB of RAM and a 128 GB SSD. The resolution of the 15.6-inch display is 1366×768 pixels. To compare the performance of the processors, a similar laptop based on the Intel Core i7-3610QM was used.

Processor A10-4600M [in 16 benchmarks]

AMD
A10-4600M

  • Interface
  • Core frequency
  • Video memory size
  • Memory type
  • Memory frequency
  • Maximum resolution

Description

AMD started AMD A10-4600M sales 15 May 2012. This is Trinity architecture notebook processor primarily aimed at office systems. It has 4 cores and 4 threads and is manufactured using 32nm process technology, the maximum frequency is 3200MHz, the multiplier is locked.

In terms of compatibility, this is a socket processor
AMD Socket FS1r2
with TDP 35W. It supports DDR3 memory.

It provides poor benchmark performance at

1.91%

from the leader, which is AMD EPYC 7h22.


A10
4600M

or

AMD-V +

RAM support

Types, maximum size and number of channels of RAM supported by A10-4600M. Higher memory frequency may be supported depending on the motherboard.

Memory types DDR3 of 5200 (Ryzen 5 7600X)

Integrated video — specifications

General parameters of the integrated video card in A10-4600M.

GPU AMD Radeon HD 7660G

Benchmarks

These are the results of the A10-4600M performance tests in non-gaming benchmarks. The overall score is set from 0 to 100, where 100 corresponds to the fastest processor at the moment.


Overall performance in tests

This is our overall performance rating. We regularly improve our algorithms, but if you find any inconsistencies, feel free to speak up in the comments section, we usually fix problems quickly.

A10-4600M
1.91

  • Passmark
  • GeekBench 5 Single-Core
  • GeekBench 5 Multi-Core
  • Cinebench 10 32-bit single-core
  • Cinebench 10 32-bit multi-core
  • 3DMark06 CPU
  • Cinebench 11.5 64-bit multi-core
  • Cinebench 15 64-bit multi-core
  • Cinebench 15 64-bit single-core
  • Cinebench 11.5 64-bit single-core
  • TrueCrypt AES
  • x264 encoding pass 2
  • x264 encoding pass 1
  • WinRAR 4.0
  • Geekbench 2
Passmark

Passmark CPU Mark is a widely used benchmark that consists of 8 different tests, including integer and floating point calculations, extended instruction tests, compression, encryption, and game physics calculations. Also includes a separate single-threaded test.

Benchmark coverage: 68%

A10-4600M
1925

GeekBench 5 Single-Core

GeekBench 5 Single-Core is a cross-platform application designed as CPU benchmarks that independently recreate certain real world tasks that can accurately measure performance. This version uses only one processor core.

Benchmark coverage: 37%

A10-4600M
372

GeekBench 5 Multi-Core

GeekBench 5 Multi-Core is a cross-platform application designed as CPU benchmarks that independently recreate certain real world tasks that can accurately measure performance. This version uses all available processor cores.

Benchmark coverage: 37%

A10-4600M
974

Cinebench 10 32-bit single-core

Cinebench R10 is a very outdated ray tracing benchmark for processors developed by the authors of Cinema 4D — Maxon. The Single-Core version uses a single CPU thread to render a futuristic motorcycle model.

Benchmark coverage: 20%

A10-4600M
2078

Cinebench 10 32-bit multi-core

Cinebench Release 10 Multi Core is a variant of Cinebench R10 that uses all processor threads. The possible number of threads in this version is limited to 16.

Benchmark coverage: 19%

A10-4600M
5696

3DMark06 CPU

3DMark06 is an outdated set of benchmarks based on DirectX 9 by Futuremark. Its processor part contains two tests, one of which calculates the pathfinding of game AI, the other emulates game physics using the PhysX package.

Benchmark coverage: 19%

A10-4600M
2866

Cinebench 11.5 64-bit multi-core

Cinebench Release 11.5 Multi Core is a variant of Cinebench R11.5 that uses all processor threads. This version supports a maximum of 64 threads.

Benchmark coverage: 17%

A10-4600M
2

Cinebench 15 64-bit multi-core

Cinebench Release 15 Multi Core (sometimes referred to as Multi-Thread) is a variant of Cinebench R15 that uses all of the processor threads.

Benchmark coverage: 14%

A10-4600M
187

Cinebench 15 64-bit single-core

Cinebench R15 (Release 15) is a benchmark created by Maxon, the creator of the popular Cinema 4D 3D modeling package. It was superseded by later versions of Cinebench using more modern variants of the Cinema 4D engine. The Single Core version (sometimes referred to as Single-Thread) uses only one CPU thread to render a room full of mirror balls and complexly shaped lights.

Benchmark coverage: 13%

A10-4600M
63

Cinebench 11.5 64-bit single-core

Cinebench R11.5 is an old benchmark developed by Maxon. authors of Cinema 4D. It has been superseded by later versions of Cinebench which use more modern variants of the Cinema 4D engine. The Single Core version loads one CPU thread with ray tracing, rendering a glossy room full of crystal spheres and lights.

Benchmark coverage: 13%

A10-4600M
0.7

TrueCrypt AES

TrueCrypt is a deprecated program that was widely used to encrypt disk partitions on the fly. It contains several built-in benchmarks, one of which is TrueCrypt AES. It measures the speed of data encryption using the AES algorithm. The result of the test is the encryption speed in gigabytes per second.

Benchmark coverage: 13%

A10-4600M
1

x264 encoding pass 2

x264 Pass 2 is a slower MPEG4 x264 video compression benchmark, resulting in a variable bit rate output file. This results in a better quality of the resulting video file, as a higher bit rate is used when it is needed more. The benchmark result is still measured in frames per second.

Benchmark coverage: 12%

A10-4600M
13

x264 encoding pass 1

The x264 benchmark uses the MPEG 4 x264 compression method to encode the sample video in HD (720p). Pass 1 is a faster option that produces an output file at a constant bit rate. Its result is measured in frames per second, that is, how many frames of the source video file were encoded in one second on average.

Benchmark coverage: 12%

A10-4600M
58

WinRAR 4.0

WinRAR 4.0 is an outdated version of the popular archiver. It contains an internal speed test using maximum compression by the RAR algorithm on large amounts of randomly generated data. Results are measured in kilobytes per second.

Benchmark coverage: 12%

A10-4600M
1892

Geekbench 2

Benchmark coverage: 5%

A10-4600M
4729


Relative capacity

Overall performance of the A10-4600M compared to its closest competitor in notebook processors.


Intel Core i5-580M
101.05

Intel Core i7-840QM
100.52

Intel Core i3-5015U
100.52

AMD A10-4600M
100

Intel Core M-5Y10c
98.95

Intel Core M-5Y10a
98.95

Intel Core i5-3439Y
98.95

Competitor from Intel

We believe that the nearest equivalent to A10-4600M from Intel is Core i3-5015U, which is 1% faster on average and higher by 8 positions in our rating.


Core i3
5015U

Compare

Here are some of Intel’s closest competitors to the A10-4600M:

Intel Core i3-4120U
101.05

Intel Core i3-5015U
100. 52

Intel Core i7-840QM
100.52

AMD A10-4600M
100

Intel Core i5-3439Y
98.95

Intel Core M-5Y10c
98.95

Intel Core M-5Y10a
98.95

Other processors

Here we recommend several processors that are more or less similar in performance to the reviewed one.


Core i7
840QM

Compare


Core i3
5015U

Compare


A10 Pro
7350B

Compare


Core M
5Y10c

Compare


Core i5
2410M

Compare


Core i5
560M

Compare

Recommended video cards

According to our statistics, these video cards are most often used with A10-4600M:


Radeon HD
7660G

33. 6%


Radeon HD
7660G + HD 7670M Dual Graphics

22.7%


Radeon HD
7670M

13.9%


Radeon HD
7600G

5.9%


Radeon
Graphics

1.8%


Radeon HD
8670M

1.3%


Radeon HD
7970M

1.3%


Radeon HD
7770

0.9%


GeForce GTX
1050 Ti

0.8%


HD
Graphics 4000

0.8%

User rating

Here you can see the evaluation of the processor by users, as well as put your own rating.


Tips and comments

Here you can ask a question about the A10-4600M processor, agree or disagree with our judgements, or report errors or inaccuracies on the site.


Please enable JavaScript to view the comments powered by Disqus.

AMD A8-5500 Processor Review: Specifications, Benchmarks

The A8-5500 processor was released by AMD, release date: October 2012. The processor is designed for desktop computers and is built on the Trinity architecture.

Processor locked for overclocking. The total number of cores is 4. The maximum clock frequency of the processor is 3.2 GHz. Technological process — 32 nm. Cache size: L1 — 128 KB (per core), L2 — 1024 KB (per core).

Supported memory type: DDR3.

Supported socket type: FM2. The maximum number of processors in the configuration is 1. Power consumption (TDP): 65 Watt.

Benchmarks

PassMark
Single thread mark
Top1 CPU
This CPU
PassMark
CPU mark
Top1 CPU
This CPU
118142
Geekbench 4
Single Core
Top1 CPU
This CPU
Geekbench 4
Multi-Core
Top1 CPU
This CPU
CompuBench 1. 5 Desktop
Face Detection
Top1 CPU
This CPU
56.680 mPixels/s
4.063 mPixels/s
CompuBench 1.5 Desktop
Ocean Surface Simulation
Top1 CPU
This CPU
741.453 Frames/s
8.685 Frames/s
CompuBench 1.5 Desktop
T-Rex
Top1 CPU
This CPU
3. 237 Frames/s
0.185 Frames/s
CompuBench 1.5 Desktop
Video Composition
Top1 CPU
This CPU
49.002 Frames/s
0.504 Frames/s
CompuBench 1.5 Desktop
Bitcoin Mining
Top1 CPU
This CPU
218.231 mHash/s
3. 038 mHash/s
GFXBench 4.0
Car Chase Offscreen
Top1 CPU
This CPU
9047 Frames
789 Frames
GFXBench 4.0
Manhattan
Top1 CPU
This CPU
7128 Frames
2225 Frames
GFXBench 4. 0
T-Rex
Top1 CPU
This CPU
12887 Frames
4729 Frames
GFXBench 4.0
Car Chase Offscreen
Top1 CPU
This CPU
9047.000 Fps
789.000 Fps
GFXBench 4.0
Manhattan
Top1 CPU
This CPU
7128. 000 Fps
2225.000 Fps
GFXBench 4.0
T-Rex
Top1 CPU
This CPU
12887.000 Fps
4729.000 Fps
Name Meaning
PassMark — Single thread mark 1333
PassMark — CPU mark 2587
Geekbench 4 — Single Core 455
Geekbench 4 — Multi-Core 1243
CompuBench 1. 5 Desktop — Face Detection 4.063 mPixels/s
CompuBench 1.5 Desktop — Ocean Surface Simulation 8.685 Frames/s
CompuBench 1.5 Desktop — T-Rex 0.185 Frames/s
CompuBench 1.5 Desktop — Video Composition 0.504 Frames/s
CompuBench 1.5 Desktop — Bitcoin Mining 3.038 mHash/s
GFXBench 4.0 — Car Chase Offscreen 789 Frames
GFXBench 4.0 — Manhattan 2225 Frames
GFXBench 4.0 — T-Rex 4729 Frames
GFXBench 4. 0 — Car Chase Offscreen 789.000 Fps
GFXBench 4.0 — Manhattan 2225.000 Fps
GFXBench 4.0 — T-Rex 4729.000 Fps

Features

Architecture name Trinity
Production date October 2012
Place in the ranking 1479
Applicability Desktop
Support 64 bit
Crystal area 246mm
Level 1 cache 128 KB (per core)
Level 2 cache 1024 KB (per core)
Process 32 nm
Maximum case temperature (TCase) 71 °C
Maximum frequency 3.

2022 © All rights reserved