Intel extreme graphics specs: Intel Extreme Graphics Specs | TechPowerUp GPU Database

Navigate this review:

• Intel NUC 12 Extreme (Dragon Canyon) pricing and availability
• Intel NUC 12 Extreme (Dragon Canyon) specs
• Design: The Intel NUC 12 Extreme leaves no extra space in the chassis
• Performance: It comes with a locked Core i9-12900
• Who should buy the Intel NUC 12 Extreme (Dragon Canyon)?

Intel NUC 12 Extreme (Dragon Canyon) pricing and availability

• The Intel NUC 12 Extreme will ship in Q2 2022, and costs $1,149 and $1,449 for the Core i7 and Core i9 models, respectively.

The Intel NUC 12 Extreme was announced in February, and it’s set to be made available in the second quarter of this year. Now that Intel’s full range of desktop processors are shipping, it should be available soon.

There are two configurations. One of them comes with a Core i7-12700 for $1,149, and the other comes with a Core i9-12900 for $1,449. Other than the CPUs, the products are mostly the same with some very minor differences noted in the spec sheet.

It’s only sold as a DIY kit, meaning that you’ll have to bring at least your own RAM and storage. The CPUs do come with integrated graphics, but you’ll probably want to add on your own graphics card. Of course, the larger issue there is that the cheapest way to get a new graphics card right now is to get it in a new, pre-built PC.

Intel NUC 12 Extreme (Dragon Canyon) specs

Processor	• 12th Generation Core i9-12900 processor Up to 5.1GHz Turbo Boost Max, 8P+8E cores, 24 threads, 30MB L3 Cache • 12th Generation Core i7-12700 processor Up to 4. 9GHz Turbo Boost Max, 8P+4E cores, 20 threads, 25MB L3 Cache
Graphics	PCIe x16 slot for add-in discrete graphics (up to 12’’), Intel® UHD Graphics 770
Storage	3 x M.2 key M slots: 1 x 80 CPU-attached PCIe x4 Gen4 NVMe, 2 x 42/80 PCH-attached PCIe x4 Gen4 NVMe or SATA3 SSD RAID-0 and RAID-1 capable Intel Optane SSD and Intel Optane Memory M10, h20, and h30 ready
Memory	Dual-channel DDR4-3200 SODIMMs, 1.2V, 64GB max
Ports	• HDMI 2.0b port • Mini DisplayPort 1.4 port • Front and rear Thunderbolt 4 ports with 5V/9V fast phone charging profiles • Six USB 3.1 Gen2 ports • 10GbE port standard • Intel 2.5Gb (i225-LM) Ethernet port available on Intel Core i9 processor SKUs • SDXC slot with UHS-II support
Connectivity	Intel Wireless-AX 201, IEEE 802.11ax 2×2/ Wi-Fi 6E AX211 (select SKUs) + Bluetooth 5, with internal antennas
System BIOS	• 256Mb Flash EEPROM with Intel Platform Innovation Framework for EFI Plug and Play • Advanced configuration and power interface V5. 0b, SMBIOS2.5 • Intel BIOS • Intel Express BIOS update support
Hardware management features	• Voltage and temperature sensing • ACPI-compliant power management control
Expansion capabilities	• Two internal USB 2.0 headers
Audio	• Up to 7.1 multi-channel digital audio via HDMI or DisplayPort signals
OS compatibility	• Windows 10 • Various Linux distros
Environment operating temperature	• 0 C to +35 C
Storage temperature	• -20 C to +60 C
Price	$1,150 / $1,450

As you can see, there are two models, one of which comes with a Core i7 and the other of which comes with a Core i9. Intel sent me the latter for review, although it did not have any input on the content of the review.

• It comes with fans on all sides, and there’s no extra room inside the body.
• RGB lighting is included in a skull on the front and on the bottom.
• Disassembly is a bit complicated.

The look and feel

The Intel NUC 12 Extreme is one of the most interestingly designed PCs I’ve ever seen, because there isn’t an inch of this thing that goes to waste. Once you add a graphics card, which isn’t actually mandatory, there’s no room left in the chassis. Luckily, there are fans on all sides, and surprisingly, I didn’t have any issues with overheating.

It is necessary to clean the sides every week or two though, as it gets pretty obvious when dust starts to collect.

With an innovative design, Intel packs a ton of performance into an impossibly small footprint.

This unit is much larger than a regular NUC mini PC, of course. It’s meant to be as small as it gets while still having removable parts. There’s a big light-up skull on the front, which is clearly supposed to appeal to gamers, and there’s matching RGB lighting on the bottom. Of course, if any of this isn’t for you, you can just turn the lighting off in Intel NUC Software Studio. When the skull isn’t lit up, it just looks like a black panel.

Right in the middle of the front if the power button, and below that, there are several front-facing ports, including USB 3.2 Gen 2 Type-A, Thunderbolt 4, a 3.5mm audio jack, and an SD card reader. On the back, there are an additional six USB 3.2 Gen 2 Type-A ports, two Thunderbolt 4 ports, Ethernet, and there’s even an HDMI port in case you don’t bring your own graphics.

Assembly and disassembly

I found that taking it apart and putting it back together was a bit complicated, at least when compared to a full-size tower that would just require the removal of a couple of thumb screws. These are the steps you need to take to get this open:

1. Remove the back panel with four Phillips-head screws. Bonus points for the engineers here as the screws stay in the panel, so you won’t lose them.
2. Slide off both side panels.
3. Lift the top panel.
4. Unscrew the flap on the back panel and fold it down.

OK, now the chassis is open. If you’re installing a graphics card, it’s pretty obvious how to do that. You just pop it into the PCIe slot and connect the power cables, while will be readily exposed for you. However, that’s going to block your access to the compute unit. In other words, make sure you install the RAM and SSD first.

Behind a piece you’ll have to unscrew that’s there for support, you’ll find the compute unit. It’s just a rectangular black box. You can unscrew them and pop open the compute unit’s lid. You could remove the compute unit if you want to, but I really wouldn’t recommend it. The wires that it’s connected to seem fragile.

Upon opening the compute unit, that’s where you’ll be able to install the RAM and SSD. As noted in the specs, you’ll need to make sure you get DDR4 SODIMM memory. That means that if you were thinking of taking some DDR4 memory out of a desktop you had lying around, or reusing this memory in a future build, you’re out of luck. SODIMM memory is smaller than standard size.

Once you pop in the SSD and memory, you can close up the compute unit lid, hook up your graphics card if you have one, screw in the flap on the back, reattach the top panel, slide on the side panels, and screw on the rear panel. Plug it in and you’re good to go.

Obviously, the additional steps represent more complication in an era when a lot of desktop-makers are emphasizing simplicity and tool-less repair. But if you’re going for anything called NUC, you’re looking for something with a smaller footprint. The engineering here is phenomenal, and I love how much performance I get for the volume of the chassis. My main desktop is the complete opposite. It’s a tower that’s as big as it gets, so it’s always nice to try something a little different.

Performance: It comes with a locked Core i9-12900

• CPU performance with the Core i9-12900 is really fast.
• It’s disappointing that it only supports DDR4-3200 SODIMM.

The Intel NUC 12 Extreme is a brilliant PC for photo and video editing.

You’ve got two options in the CPU with the Intel NUC 12 Extreme: a Core i7-12700 or a Core i9-12900. Intel sent me the latter, and I filled up the rest with 64GB RAM and a 2TB SSD, both supplied by Corsair, and an NVIDIA GeForce RTX 2080 Ti. I actually could have tried putting an RTX 3090 in there, but let’s be clear about something: no one should ever pair a Core i9-12900 with an RTX 3090. If you’re shelling out for an RTX 3090, also shell out for a Core i9-12900K.

The ‘K’ in that SKU means that it’s unlocked for overclocking, and it’s important to consider if you want it when deciding what kind of system you need. It’s not just for the Core i9 either, as you can get a Core i5-12600K and a Core i7-12700K. If you’re looking for gaming performance, that’s probably the better direction.

I think of the non-K chips as more mainstream, and more for creators. This has been a brilliant PC for video editing, photo editing in Photoshop and Lightroom, and more. Intel’s new 12th-gen processors have big and little cores, so they have the power when you need it, and can use less power when you don’t need it. This processor has a total of 16 cores (8 P-cores and 8 E-cores) and 24 threads.

The Core i7-12700 comes with 12 cores and 20 thread, and while it has the same amount of performance cores, there are only four efficient cores. Personally, I think that when deciding between the Core i7-12700 and the Core i9-12900 for the NUC 12 Extreme, think of your use case. The Core i7 is a high-end productivity chip that has the power to do more. The Core i9 is for extreme power.

Performance benchmarks

For performance benchmarks, I tried to focus on CPU performance rather than overall performance. When you build a PC, there are a lot of things that contribute to performance. Some things are obvious like the CPU and GPU, but you could still be bottlenecking yourself with slow storage, slow memory, or even just a cheap CPU cooler. In other words, since you’re bringing your own GPU, RAM, and SSD, the main thing to rate here is the CPU.

Out of the following tests, Geekbench and Cinebench are the ones that test CPU performance. 3DMark can also be broken down into CPU and GPU scores (12,920 and 13,914 for Time Spy, respectively, and 6,264 and 6,554 for Time Spy Extreme).

I still haven’t seen anything beat Intel in single-core scores, although the Core i9-12900 doesn’t quite match up to the Core i9-12900K, which has a higher TDP. Ultimately, the CPU performance is almost as good as it gets, right behind the K-series variant.

RAM and SSD

Like I said earlier, Corsair sent over some DDR4 memory and an SSD, both of which fit the specs of this machine. Let’s start with the DDR4-3200 SODIMM memory.

While DDR4 is still common, DDR5 would be nice here.

When it comes to memory, the market is a bit complicated right now. Intel’s 12th-gen processors add support for DDR5 memory; however, DDR5 is super expensive still thanks to shortages around the world. Because of that, you’re still seeing lots of products ship with DDR4.

On a regular day, there isn’t much of a trade-off. With DDR4 SODIMM memory, however, you can see that there’s more of a difference than when I reviewed the HP OMEN 45L, which has DDR4. Being that Intel made this rig, I’d have hoped for DDR5 support.

As for the PCIe Gen 4 SSD, you can see that speeds are blazing fast. Due to the small space in the Intel NUC 12 Extreme, I actually had to remove the built-in heatsink that Corsair ships it with, so I’m looking forward to doing some further testing with it in my main rig.

Who should buy the Intel NUC 12 Extreme (Dragon Canyon)?

Obviously, the Intel NUC 12 Extreme is not a product that’s for everyone. After all, it’s not even a fully built computer, nor is it as flexible as a completely custom build.

Who should buy an Intel NUC 12 Extreme:

• People who want the flexibility of adding some of their own parts without doing a fully custom build
• Users who want a powerful PC that will take up as little space as possible

Who should not buy the Intel NUC 12 Extreme:

• People looking for a pre-built computer
• Customers looking for the least inexpensive way to get a new computer with a new graphics card

I absolutely love the Intel NUC 12 Extreme. It’s a phenomenal PC with fast CPU performance. It all fits in the most compact footprint I could imagine for something that will take a full-size graphics card.

But you have to remember that it isn’t cheap. You have to want this. And of course, I think it helps if you already have some of the parts handy.

Graphics: fast, slow and integrated

Part 18: Intel HD Graphics 4000 in different environments and the impact of the latter on the performance of the first

Processors based on the Ivy Bridge microarchitecture appeared a year ago, so everyone who follows this topic even a little , know: what is the name of the older video core built into the desktop Core i7. That’s right — Intel HD Graphics 4000. And if we go down a little lower in the rankings somewhere up to the level of Core i3, then what will we find there? Most Intel HD Graphics 2500 models, but i3-3225 and the recently announced 3245 have the same HDG 4000. Notebook models also have it, and in all of them without exception (with the exception of Celeron and Pentium, which are considered separate from the Core categories) : from extreme i7-3940XM (four cores up to 3. 9GHz, TDP 55W), to i3-3229Y Tablet (two cores @ 1.4GHz, TDP 13W). But is it the same video core? In the case of discrete video cards, the question would be meaningless: one can be installed in a computer with any processor (at least theoretically). With an integrated solution, everything is more complicated. Firstly, even with a cursory glance, the difference in the maximum frequency of the GPU is noticeable, and the range is extremely wide — from 850 MHz (just i3-3229Y) to 1.35 GHz (i7-3940XM), i.e., it differs by more than one and a half times. Secondly, we are not talking about some fixed frequencies — even in the first generation of Core GPUs of mobile processors, Turbo Boost technology began to be used, and it is also used for processor cores. What does this lead to? The frequency of both of them changes dynamically, and it depends both on the load on the CPU and GPU, and on what heat package you ultimately need to “keep within”. In general, everything is unpredictable in advance, but there is an assumption that mobile graphics, even though they are called the same as desktop graphics, are slower.

The discrepancy in end systems is not limited to the GPU frequency alone. Even in the entry-level discrete video card market, their final characteristics are at the mercy of manufacturers, and the developer of the video processor itself is not controlled in any way. The discrepancy with the official performance characteristics can be significant, which we have recently observed: four (!) Of the five Palit video cards were somewhat (to put it mildly) different from what NVIDIA intended. And it’s easy to see that the main differences concerned not even the frequencies of the chip, but the memory system. However, this is quite possible in the case of integrated graphics, especially since in this case the memory is rarely soldered on the board. Accordingly, options are possible. For example, the «official» DDR3-1600 or the slower DDR-1333 — which modules the manufacturer (or user) decides to use, they will be. But this, at least, is somehow amenable to manual configuration, but if the manufacturer decides to install only one SO-DIMM slot (most often inexpensive ultrabook models “sin” like this, but not only them), we will get a completely different level of performance of the graphics core , despite the fact that the specifications of the computer will still say «Intel HD Graphics 4000».

Is it possible to test all options and give an unambiguous answer: what does each of them represent? It is possible, but difficult — the number of possible configurations is finite, but large. Yes, and it’s not too interesting to do this: it has long been known that the HDG 4000, even in its “best form”, is not a full-fledged gaming solution, but for solving most other tasks, as a rule, older and weaker GPUs are enough — up to HD Graphics processors Celeron on the Sandy Bridge core. On the other hand, you can try to estimate the approximate range where most solutions should fall — this is not so difficult. Yes, and in the process of various testing, we have accumulated a certain set of useful information. In any case, it turned out that recently, using the same driver version (which is relevant in this case), we have tested five different computer configurations for different purposes that have just the desired graphics subsystem. Thus, in this article, we will simply collect the results together and try to evaluate the impact of various factors on the performance of the Intel HD Graphics 4000 graphics core.

Test stand configuration

We have already indicated the range of potentially possible clock frequencies above — from 850 MHz in Y-series processors to 1350 MHz in Core i7 Extreme Mobile. Thus, from the point of view of theory, the most correct approach would be to take two systems: on Core i3-3229Y (there is nowhere lower) and Core i7-3940XM (there is no higher) and test them with different memory configurations — at least one and two channels , and as a maximum also with different frequencies. Which is not feasible in practice. Firstly, it is still difficult to find anything on the Y-processor: such models have appeared quite recently, so most tablets in retail chains are equipped with the more familiar U or even M Core. Secondly, there is still no point in looking for it: the design of the tablet does not imply a flexible configuration of the memory system — here you can just “run into” memory modules soldered on the board and / or fatal single-channel. Thirdly, not everything is smooth on the top edge — top laptops do not have the problems described above, however, processors of the XM and QM families (where the maximum graphics frequency is 1. 3 GHz) are usually found on sale exclusively paired with discrete video cards , which can not always be turned off. On the other hand, this also leads to the fact that it is simply not necessary to test the extreme options — since the probability of meeting them in practice is zero or (in the case of Y) there are still no choices.

And here is the range 1. 05-11.15 GHC, here is extremely interesting in that it fits most of the possible options. It is easy to see that three of the five configurations have already been tested by us — today, just the video-related results will be “deployed”. And supplemented by two more implementations — in the Core i7-3770S and i7-3770K processors. The clock speed of the video core is typical for many Core i7 1.15 GHz, but two different memory frequencies. Plus a huge spread in terms of processor performance — let’s see how it can affect the graphics results. And for comparison, we added the results of one processor with HDG 2500, but a powerful processor part — it suddenly turns out that ultra-mobile solutions, despite the top-end (formally) graphics, are still significantly slower. With the equality of the processor part, this, of course, is not observed, but with such a difference, everything can be.

And an important point is the different TDP levels of the tested processors, since five out of six support Turbo Boost technology for processor cores and all for GPUs. Why is it important? You may recall that in our power consumption tests, loading the GPU increased it for the Core i7-3770K by 17 watts. Naturally, a lot depends on the specific instance of the processor, especially since different series are subjected to the selection of different degrees of rigidity according to this parameter — we also saw 20 W from the HDG 2500 in the budget i5-3450. But the order of magnitude itself is understandable and, in general, not small — dual-core U-series processors are limited to the same 17 W for the entire processor. Yes, and 12 W of the official difference between the 3770S and 3770K must also affect the operation of Turbo Boost when using the processor «as a whole», and, therefore, on performance.

Aliens vs. Predator

As we have already written more than once, no integrated graphics can handle this game in this mode, so we get a pure stress test of the video core operating at the limit of its capabilities. Moreover, anything can turn out to be a limiter to these possibilities: the equality of the results of the Core i3-3217U and i7-3517U is very indicative — despite the potential differences, both models “rested” on the same TDP. But two qualitative effects are clearly visible — firstly, single-channel memory is similar to death even for U-family processors (we have already seen that this is true for top models), and secondly, even in this mode, the HDG 4000 is still faster, than 2500.

In low-quality mode, you can even try and play on any of the subjects. But in different ways: a low-frequency dual-core processor with single-channel DDR3-1333, but with HDG 4000, as it turned out, is suitable for this almost to the same extent as one of the older desktop models with HDG 2500! Despite the fact that in this mode the processor is also working, it is not for nothing that two quad-core Core i7s are in the first places. The difference between them is already relatively small, despite the fact that one model is generally top-end and works with faster memory, and the second is energy efficient. 3217U and 3517U are much slower, although in their case there is some performance margin that can slightly improve the picture quality.

Batman: Arkham Asylum GOTY Edition

Relatively old and «light» graphics engine «loads» the GPU to a lesser extent, but it has increased requirements for the processor component due to good multi-threaded optimization. As a result, desktop Core i7 already «pull» the quality mode, and ultra-mobile processors are only close to this level. But they are very close, so with a slight decrease in quality they can reach the “playable” level. Unless, of course, you «clamp» the memory system — in single-channel mode, the HDG 4000 is reduced to almost 2500. But, by the way, not lower — the i5-3570S overtook the i5-3317U only due to the «full» four cores at a higher clock frequency and twice as much L3 cache.

At minimum quality, everything turns into a processor competition. What is worth noting here is that such settings, as we can see, still cannot be called completely irrelevant — for top-end processors with integrated graphics, the frame rate begins to “go off scale” beyond the threshold of sufficiency, but you have to test not only them. On models for nettops and ultrabooks, the FPS is high, but not to say that it is “excessive”.

Crysis: Warhead x64

Another stress test, where, firstly, the complete incompetence of systems with single-channel memory, that of the HDG 2500, and secondly, that the processor component still matters, even in such conditions, affecting the overall performance. On the other hand, first of all, nevertheless, the GPU, and then everything else.

Including — and in potentially suitable for practical use (if, of course, it is a pleasure for someone to admire such a picture) video modes. In any case, the Core i7-3517U still managed to overtake the Core i5-3570S due to the advantage in the graphics component, despite the fundamentally different processor.

F1 2010

As we have already written more than once, the same frame rate in this game does not mean anything if it is equal to 12.5 FPS — a feature of the game engine that tries to keep it at this level, discarding the insignificant (in his opinion).

In low quality, you can sometimes play on the HDG 4000, however, as we see, for this you need at least a Core i7-3517U (in its class, by no means the worst, to put it mildly, and not cheap), and equipped with dual-channel memory with a frequency of 1600 MHz. Failure to comply with any of these conditions is fraught with consequences. Overshoot — will change the picture to a lesser extent than the overshoot size 🙂

Far Cry 2

HDG 4000 performance is still not enough for this old game (which is not news for a long time), but to a lesser extent than for Crysis or AvP , certainly. No wonder that the performance of the older and younger of the tested processors differs by one and a half times. On the other hand, from the point of view of worldly wisdom, we would not be surprised at the greater difference — after all, the CPU parts differ too much. You could even say, fundamentally and in all respects.

And in the minimum quality mode, it just comes to the fore. And the most curious result is that the Core i3-3217U, even in this case, could not reach the comfort threshold. That is, this game of almost five years ago is still in no way amenable not only to Atom or Brazos, but in general to very many platforms of increased efficiency. And it doesn’t matter whether with integrated video or with any discrete: there is not enough performance and the processor part itself. So progress is progress, and a certain minimum of system requirements must be provided. As you can see, even the older CULV processors cope without a special margin of safety, while the younger ones cannot cope at all (it will be interesting to see how Kabini and the younger Haswell are doing with this). In general, a “fresh” tablet or a budget ultrabook will not necessarily allow you to play even very old games and even “at the minimum wage”.

Metro 2033

Returning to the origins in the form of the first diagram — it is clear that none of the test subjects is enough for the quality mode of this game, and fundamentally not enough. But the effect of performance characteristics on performance is very clear, so we will not describe everything in detail — it is easy to draw all the conclusions on your own.

Metro 2033 appeared a year and a half later than FC2, so the minimum hardware requirements for the game are higher. To be fair, the “plinth” quality mode itself has a much higher quality 🙂 The minimum for it is Core i3-3225, i.e. to somehow play this game, we need a processor with a frequency above 3 GHz and HDG 4000, both of which are essential. The HDG 2500 won’t run the game even with these settings, regardless of the processor. And weak models with any graphics will not cope with it precisely because they are weak.

We advise many laptop buyers to think about the latter 😉 Firstly, in the light of these trends, the attempts of some manufacturers to complete their products on CULV processors with discrete video cards start to look a little strange. In particular, we have seen models based on the Core i3-3217U paired with the GeForce GT 740M. The last video card is another example of renaming and optimizations, since it is almost the same 640M that many people have known for a long time, but with slightly increased frequencies. God knows what, of course, but potentially a couple of times faster than the same HDG 4000. However, as we can see, the “processor independence” of games has its limit, especially when it comes to more or less modern projects, i.e. here for Metro 2033 is already a little low-voltage dual-core models. Thus, a configuration similar to the one indicated will allow the user, perhaps, to increase the quality of the picture in old games, but not to play (at least somehow) in new ones — you must admit, this is not an achievement for which it makes sense to pay for discrete graphics.

The second problem is from the same area: AMD does not get tired of repeating that, although its APU has lower processor performance, the graphics are more powerful than those of Intel. As you can see, there are limits to everything — including the weak dependence of the results on the processor. And then the partners add fuel to the fire by adding to some A8-4555M (which can at least feed the built-in GPU) a discrete video card based on something like the Radeon HD 7550M/8550M. No doubt — Dual Graphics is sometimes the only way to improve the performance of the graphics subsystem, but this is only true when it is not enough. As you can see, not only this is possible in the low-consuming segment.

Summary results

Let’s try to assess the situation in general, and also look not only at games, for which we will use charts with average results for a group of tests/applications (you can read the full testing methodology in detail in a separate article). The results in the diagrams are given in points, for 100 points in this article the performance of the Core i3-3217U is taken as the slowest of the four processors tested. Those who are interested in more detailed information are again traditionally invited to download a table in Microsoft Excel format, in which all the results are shown both in converted points and in «natural» form.

Let’s start with the games. It’s immediately obvious that the single-channel memory mode instantly drops the HDG 4000 to the level of 2500 and other similar solutions, so it’s not very relevant for practical use. Under normal conditions, the difference in results is 33%. On the one hand — a lot, on the other — everything is different. Even the TDP is 4.5 times different. But if you do not give such freedom, and use the same DDR3-1333 memory, then 15% is not gained. Which is easily explained — after all, the video core itself is the same (adjusted for the effect of the heat pack on its real clock frequency), and taking into account its power, heavy gaming applications are its stress test in the first place.

But in practice, as we have already seen, in such conditions the frame rate is almost universally too low to be used, so reduced graphics quality modes are more relevant. For many solutions, it is reduced to a minimum: this mode is too easy for top solutions, but CULV processors even in it, as we see, do not always cope. And here the dependence of the results on the processor part is already visible to the naked eye, so 33% turn into 128% — comments are unnecessary. Moreover, we note that a “normal desktop” processor with an HDG 2500 bypasses even the CULV Core i7 (3517U, of course, is a junior model, however, the older 3687U differs only in a maximum clock frequency increased by 10%, which may well not be enough), but one and a half times behind a «normal desktop» processor with an HDG 4000.

If this load were multi-threaded, most likely we would have received a spread of results as in the previous case, but “only” 1.87 times. But the alignment inside is different: there is practically no difference between the HDG 2500 and 4000. It’s no wonder that the memory mode has an effect, but only slightly — the higher clock frequency of the processor more than covers this difference.

In the days of GMA and the first versions of HDG, these results also depended on the video core, but now, as we can see, they have stopped. Well — we will take this into account when developing the next versions of test methods 🙂

Total

So, as expected, we have confirmed the dependence of the performance of integrated graphics solutions on the processors in which they are integrated. However, we note that it is not always so strong. As expected, when the load falls on the GPU, a large spread of results can be found only when comparing processors with a fundamentally different heat pack, since it also affects the frequencies of the graphics core. But such modes are guaranteed to be too «heavy» not only for the IGP, but also for the lower models of discrete video cards, so in order to play them in practice (and not just watch a slide show), you have to reduce the image quality, because. That is, reduce the load on the GPU and increase it on the CPU. While the latter belong to the same class, the power of the graphics core itself continues to determine (which we have already observed on the example of desktop solutions, where a pair of high-frequency cores and a TDP margin allowed the same HDG 4000 to deploy to the full extent of its weak forces and paired with different processors ), but expecting the same level of performance from ultrabook and desktop processors is no longer worth it. In principle, it would be difficult to assume otherwise, but it is never superfluous to make sure that this is exactly the state of affairs. The love to name similar architecturally, but different in performance solutions, of course, did not start with Intel, but in most cases, manufacturers still somehow hint at the existence of a difference. Yes, the company itself adheres to the same practice in the naming system of processors — giving them non-intersecting numbers and not forgetting to add the letter “M” or “U” at the end, sometimes sharply affecting the family number (a hackneyed example: the vast majority of desktop Core i5 are to quad-core processors, but all Core i5-Ms are dual-core only). And with the graphics, there is not even such clarity: one can judge only by indirect signs — such as the name of the processor where it is built.

Is there any hope of stopping the resulting mess in the future? Maybe in a distant, but definitely not in the next generation of processors. That is, we certainly have no doubt that the Iris 5100 is a more powerful GPU than the HDG 4600. However, will this allow you to play on the Core i7-4558U (dual-core SoC with a TDP of 15W) with more comfort than on the Core i7-4700HQ, not to mention the older desktop Core i7-4770K (quad-core processors, which also outperform the 4558U in terms of clock speed and are less «squeezed» by the heat pack) — the question is open. And the complete equality of processors with the equally integrated GPU called the same is all the more doubtful. However, it is impossible to deal with these issues without direct testing, and this is a topic for completely different testing.

We thank the companies Corsair and Palit
for their help in completing the test benches

Intel GMA 950: characteristics, review, overclocking

One of the first integrated graphics accelerators from Intel — GMA 950 was presented in this solution. 2008. At that time, this video accelerator was perfect for creating entry-level personal computers. Even some 3D toys that were demanding on the hardware resources of the graphics adapter were launched on it with a resolution of 640 x 480, and even 50-60 frames per second could be obtained. It is about this product that will be discussed in this material.

What users is this accelerator aimed at?

Niche The Intel GMA 950 is an entry-level personal computer with minimal graphics performance requirements. Such a system unit is perfect for working with office applications, watching videos and running the simplest toys (for example, turn-based strategies). But something more to get from it in terms of speed will not work. For such PCs, the presence of a discrete accelerator is mandatory, and this will already allow you to run resource-intensive tasks: three-dimensional games, design packages or graphics processing programs.

Chipset and socket

The Intel GMA 950 graphics card was part of the G45 chipset for the LGA775 processor socket. This solution is focused on the mass segment and had a fairly low cost. The presence of an integrated graphics system made it possible in this case to assemble excellent office system units. Also, this platform was great for creating a basic multimedia system. The presence of a PCI — Express 16X slot made it possible to install discrete graphics solutions in such a personal computer and get a full-fledged gaming PC. But more often, the P45 chipset was used for these purposes, which was much cheaper and did not have an integrated video card, but at the same time, it was practically not inferior to the G45 in terms of functionality. Today, in terms of performance, the LGA775 is an outdated socket and it is not advisable to assemble a new system unit on its basis.

Modes of operation

The maximum output resolution for a PC display for this solution is 2048px X 1536px. 1920px X 1080px (1080p, Full HD), 1280px X 720px (720i) modes are also supported. All this makes this accelerator compatible with any modern monitor or TV. That is, it can be connected to any such device and display a picture on it in any of the previously indicated modes. Also in the list of supported modes of this video card there are such image output modes as 1024px X 768px, 800px X 600px, 640px X 480px. In the latter case, you can even launch quite demanding toys by the standards of 2008.

Technical specifications

At a clock frequency of 250 MHz, the Intel GMA 950 graphics processor operates. Characteristics indicate that it can process 256 bits of information in one cycle of its operation. The device includes 4 independent pixel processing units. The thermal package of the solution is 7 W, and its semiconductor crystal is manufactured according to the technical process, which corresponds to 130 nm.

Memory

Intel GMA 950, like any other integrated graphics accelerator, does not have its own RAM. For his needs, a part of the total RAM is reserved, which is installed in the computer system. And this nuance is the bottleneck that limits the capabilities of the adapter. In discrete graphic solutions for information exchange with RAM, a separate bus with an increased frequency and increased bit depth is used. It is due to this that a significant increase in performance is obtained in the latter case. The maximum amount of RAM that the video accelerator can use for its needs is 224 MB. The adapter itself can work with both DDR2 and DDR3 modules. The type of RAM depends on the capabilities of the motherboard.

Drivers

Officially, the Intel GMA 950 video card could function under any operating system of the NT family. The characteristics of the drivers for this video solution indicated this. But for Windows 9X and even for Windows Me, specialized drivers from Intel were not presented. As a result, it was possible to install some kind of unified software of this kind, but at the same time, the functionality of the graphic solution was significantly reduced.

Overclocking

Developers for the Intel GMA 950 provided certain opportunities for increasing performance. Overclocking in this case required the installation of an additional GMA Booster program. This small utility made it possible to increase the frequency of the integrated accelerator of this model, and this, in turn, made it possible to obtain a small increase in performance. The difference was so small that it didn’t make much sense. Therefore, if it was possible to meet a PC with such an overclocked graphics processor, it is extremely rare.

Analogues

GeForce 9300 from NVidia and Radeon X300 from ATI are the main competitors in the integrated graphics market at the time of the release of the Intel GMA 950. The characteristics of these graphics accelerators are shown in the table.