AMD Radeon R9 Nano CrossFire Performance Revealed
The first proper Crossfire review for the Radeon R9 Nano has been published by TechPowerUp. The Radeon R9 Nano is based on the Fiji graphics core and we should expect around the same performance scaling as the Fury based cards but so far none of the major sites had done proper CrossFire testing of the card probably because AMD had sent a single review sample to each tech site.
TechPowerUp was one of the sites that didn’t receive a review sample at launch following the «Fair Reviews» drama brought up by AMD’s Roy Taylor that might have been the reason his account got deleted on Twitter a few days ago. With the dust settled and the Nano launch passed, AMD seems to have started sending samples to the sites who were unable to receive them. Some of these sites include the much respected and well-known TechReport, TechPowerUp itself and also, HardOCP. So far, TechPowerUp has published their review just nine days after the first reviews appeared on the web. TPU not only managed to get a review sample but bought themselves a retail Nano too to test the Crossfire performance of AMD’s latest and most powerful SFF (Small Form Factor) card.
The performance results show that the Radeon R9 Nano scales really well at 4K (3840×2160) resolution with up to 90% scaling in AMD optimized titles but also does that are NVIDIA optimized that include GTA V and Witcher 3. CryEngine powered titles that take advantage of GCN core architecture also get 80%+ scaling. Majority of the titles show great performance from the XDMA architecture that is found on the latest GCN graphics cards and while there are some instances where the games show no performance improvement over a single Radeon R9 Nano card, that is mostly due to the fact that the games don’t have Multi-GPU functionality integrated within them. TPU noted that Far Cry 4 was the only example from several titles where all graphics card performed well in terms of CrossFire scaling except the Fiji based parts that include the Radeon R9 Nano and that is because AMD is still refining the drivers for their Fiji based cards that include the Radeon R9 Nano, Radeon R9 Fury X, Radeon R9 Fury and the upcoming, Radeon R9 Fury X2.
2 of 9
Image Credits: TechPowerUp
The site also goes on to show Crossfire performance of a Radeon R9 Nano coupled with a Radeon R9 Fury X. It was reported before the Nano launch that both the R9 Fury X and Radeon R9 Nano will feature CrossFire XDMA compatibility and that is the case but when running together, the difference is quite small when compared to a CrossFire Nano configuration even though the Radeon R9 Fury X is around 15% faster than the Radeon R9 Nano due to faster clock speeds. It seems to be that the drivers are making both the Radeon R9 Fury X and Radeon R9 Nano to operate at Nano clock speeds rather than Fury X clock speeds. This clock-speed parity between two cards affect performance so there’s no point in adding a Fury X to a CrossFire configuration with Nano as you’ll be getting the same performance as you do with a second Nano. Nevertheless, its an interesting configuration and signals towards the Fury X2 which might either use a full blown Fiji XT GPU configuration or go ahead with a Nano based Fiji XT variant that conserves clock speeds.
So this review actually does a good job providing an early sneak peak at the Fury X2.
It’s mentioned at the site that while the performance improvement is good enough for 4K gaming, a CrossFire setup takes away the idea of Mini-ITX setups which is what the Nano is aimed at. A Fury X still seems as a better choice for the AMD crowd if they want to go Multi-GPU inside their rigs as the latter offers faster performance. Do go ahead and review TPUs review for a detailed performance overview of Radeon R9 Nano in Crossfire.
2 of 9
Share this story
Updated GPU Benchmarks, and Saying Goodbye to the R9 290
Goodbye, old friend.
Greetings! We have two updates we’d like to share.
First, we’ll start with the sad news…
Goodbye, R9 290
The last of the mighty R9 200-series GPUs from AMD has fallen, for today we remove the R9 290 from the Logical Increments parts guide.
The 290 was arguably the better card when compared to the 290X, being cooler, quieter, and cheaper, while providing nearly the same performance. It outlasted the majority of the other 200-series cards, but this has been a long time coming. It is mostly sold out everywhere, and the remaining units are extremely overpriced.
Of course, the R9 290 is not dead. The same chip that was inside the 290 powers the R9 390, which is $10-$15 more expensive, but also ~10% faster. So, you can still buy the same card, albeit with a new name.
Next, some better news!
New GPU descriptions and benchmarks
GPU descriptions now look like this.
We last updated our graphics card descriptions and benchmark information more than a year ago. Each card was compared to the R9 290 and GTX 780, and we noted each card’s performance in World of Warcraft and Crysis 3.
Since then, both the 290 and the 780 have been replaced, and WoW’s engine changed a bit. At some point, WoW removed anti-aliasing options, so reviews for new cards showed WoW with higher FPS.
But then WoW apparently brought back anti-aliasing, so cards showed a huge variation in benchmarks, depending on when the benchmark was taken. For example, on 1440p the GTX 970 gets 166 FPS in this review, and only 82 FPS in this review, even though both reviews come from the same site!
Next, Crysis 3 is no longer the “most GPU crushing game around,” as Witcher 3 has taken the throne. So we now show each card’s performance in Witcher 3 instead of Crysis 3.
We now compare GPU power to the Fury X and 980 Ti, the flagship cards currently available from AMD and NVIDIA (not counting the Titan X). WoW and Witcher 3 benchmarks are taken from modern, consistent reviews. The “relative power” and “FPS in [game X]” data were taken from a wide collection of sources, so we now give ranges instead of specific values, This should make our description more accurate and realistic, though slightly less precise.
Sources for data are:
- TechPowerUp Witcher 3 Benchmarks – GTX 980 Ti
- TechPowerUp Witcher 3 Benchmarks – GTX 950
- TechPowerUp Witcher 3 Benchmarks – R9 380X
- Hardware.
info Witcher 3 Benchmarks – R7/R9 300-series - Guru3D Crysis 3 Benchmarks – GTX Titan X
- Maximum PC GTX 980 Ti SLI Benchmarks
- TechPowerUp Crysis 3 Benchmarks – GTX 980 SLI
- TechPowerUp Witcher 3 Benchmarks – R9 Nano CrossFire
- HardOCP R9 Fury CrossFire Benchmarks
- GamersNexus GTX 980 Ti SLI Benchmarks
- HardOCP R9 Fury X CrossFire Benchmarks
- Toms Hardware World of Warcraft Cataclysm Benchmarks – Radeon HD 7770 and 7750
- Toms Hardware R7 240 and 250 Benchmarks
- TechPowerUp World of Warcraft Cataclysm Benchmarks – Radeon HD 6670
- YouTube: World of Warcraft with an R7 360
- YouTube: World of Warcraft with an R7 260X
AMD Radeon R9 Nano video card review and testing GECID.com. Page 1
::>Video cards
>2015
> AMD Radeon R9 Nano
09/16/2015
Page 1
Page 2
One page
With the release of the NVIDIA GeForce GTX 9 series graphics cardsxx, NVIDIA has significantly shaken the position of its competitors in the graphics adapter market.
The release of the AMD Radeon R9 3xx line can hardly be called a worthy answer, because all of its representatives are the usual rebranding of models from previous generations. Of course, the attractive price has done its job — despite the lack of new technical solutions and relatively high power consumption, the renamed video cards are in good demand among users. However, such a marketing strategy works well only in cases where the “cost / opportunity” indicator comes to the fore, that is, in the lower and middle price segment of the market.
If we consider the niche of gaming solutions, then here the price factor fades into the background, and the main characteristic, of course, is performance. In such a situation, the usual rebranding would no longer help (and by and large there was no longer anything to rename), so AMD employees had to “sit down at the drawing board” and come up with something new.
The result is the AMD Radeon R9 Fury line, which nominally belongs to the AMD Radeon 3xx family, but is technologically superior to it.
Only three models have been announced so far (AMD Radeon R9Nano, AMD Radeon R9 Fury and AMD Radeon R9 Fury X), but we do not exclude that their number will increase in the future. At a minimum, we expect a dual-chip version of the AMD Radeon R9 Fury. We will talk about each model in more detail in the relevant reviews, but now let’s see what is inside these «furies».
General information about new video cards
climbed up to around 300 watts. It would seem that AMD should have drawn the appropriate conclusions and followed the same path as their competitor — to reduce TDP. In principle, the “red” might have succeeded if NVIDIA had not planned to launch two 250-watt monsters (NVIDIA GeForce GTX 980 Ti and NVIDIA GeForce GTX TITAN X). In fact, this meant: «turn a blind eye to energy consumption and increase power as much as possible.» And, as you know, AMD has never had problems with the latter.
As a result, we have the largest video chip (area — 596 mm 2 ) ever released by AMD, with the largest number of transistors (8.
9 billion). The novelty was called «AMD Fiji», thanks to which it perfectly fit into the company’s island terminology. It is noteworthy that the graphics core is still based on the 28nm process technology. By the way, the competitor also had certain problems with the transition to thinner lithography, so in this area there is a certain parity between the two companies.
Having made such a digression, it’s time to move on to such a favorite indicator of all fans of NVIDIA products as TDP. Recall that with a die area of 438 mm 2 , based on the same 28-nm process technology, and 6 billion transistors, the power consumption level of the AMD Radeon R9 290X was 250 watts. The flagship of the new family received a 36% larger chip and 48% more transistors. It would seem that with such introductory his TDP should have skyrocketed. But no, only 275 W, which is only 10% more than the previous figure. In other words, AMD did not just build up raw power, but also thought about how to do it as efficiently as possible.
Someone might say that NVIDIA products still look better in terms of performance / power consumption. But wait a minute. We are talking about the flagship of the AMD Radeon R9 line.Fury. But it also has less «gluttonous» models. Take at least the AMD Radeon R9 Nano graphics adapter: the same AMD Fiji chip, the same 4 GB of HBM memory, and the output TDP level is only 175 watts. This is a kind of answer to all skeptics: «if we want, we can also make a fast and efficient video card.» True, another question is that such a desire rolls on AMD extremely rarely, but these, as they say, are already nuances.
AMD Fiji Graphics Core Features
So, we have already figured out above that the AMD Fiji chip is amplified by a rather large number of transistors compared to AMD Hawaii. But where did they go? To answer this question, we suggest looking at the block diagram of the core itself.
Structural scheme GPU AMD Hawaii XT ( AMD Radeon R 9 0019
Struk t urnaya schematic In the maximum version, the AMD Fiji chip includes four unified shader engines (Shader Engine, SE), built on the basis of the GCN 1.
2 microarchitecture, with separate geometric processors (Geometry Processor, GP), rasterization units (Rasterizer) and sets of computing modules (Compute Unit , CU). The latter still have 64 stream processors (ALU) and 4 texture units (Texture Unit, TU). It seems to be all the same. However, due to the use of 16 compute units (CU) in each shader engine (SE) instead of 11, the number of streaming units (ALU) and texture units (TU) increased by 45%: from 2816 to 4096 and from 176 to 256 respectively.
However, do not think that in practice you will get the same increase in performance. After all, the number of blocks that perform triangle rasterization remained unchanged (4). And this means that a maximum of 4 primitives per clock is processed, while the flagships from NVIDIA are capable of simultaneously calculating 24 elements at once. The bottleneck of the AMD Fiji core can also be the object finalization stage, since the number of ROP pipelines has also not increased and is still 64.
For comparison: NVIDIA GeForce GTX 980 Ti these blocks are 1.5 times larger (96). The only consolation in this situation is the fact that AMD has improved the geometry processing algorithms, which, all other things being equal, makes the AMD Fiji chip faster than its predecessor. True, the manufacturer is silent about any details or at least about the percentage of the speed of performing settlement operations.
Another point that I would like to pay attention to in the context of this section is the decrease in the speed of double precision operations (FP64): from 1/2 to 1/16 relative to the speed of FP32 calculations. Translated into public language, this means that the performance of the AMD Fiji graphics chip in complex professional calculations will be relatively low, but the rendering of 3D scenes will not be affected at the moment. And since in the vast majority of cases, such video cards are bought to play the gameplay, then desktop PC users can sleep peacefully. If you use a video card to speed up calculations, then it is better to look towards professional solutions from the AMD FirePro series or analogs from NVIDIA.
By the way, the “greens” chose exactly the same strategy when releasing their tops. In our opinion, this is absolutely the right move, because a specialized product always shows better results than a universal one.
The number of memory controllers has not changed compared to the AMD Hawaii graphics processor, but now they work not with GDDR5 chips, but with HBM memory soldered on a substrate common with the video chip. We decided to dedicate a separate section to this, without exaggeration, innovative technology.
High Bandwidth Memory — a new step in the development of video memory
With the growing popularity of high resolutions and the active use of HD textures by game developers, the issue of video memory bandwidth has become very acute. Simply increasing the width is quite a resource-intensive task (a large number of GPU pins are involved and the PCB layout is much more complicated), and increasing the frequency of memory chips is too costly in terms of energy and finances (requires selected microcircuits that operate at increased speed and supply voltage) .
In a word, the GDDR5 standard has actually exhausted its potential, and we need to start moving in a different direction. NVIDIA engineers did not dare to take such a step, but their competitors from AMD, on the contrary, announced last year that they were working at full speed on a new type of memory, which was called «High Bandwidth Memory» or «HBM» for short.
Its fundamental difference from GDDR5 is that a very wide bus (4096 bits) and a set of relatively slow memory chips (operating at an effective frequency of 1 GHz) are used to connect to the GPU. With conventional GDDR5 video memory, the opposite is true.
Naturally, all the complexities of bus wiring of such a large bit depth are preserved here. To solve this problem, the memory was combined into stacks of four chips, which are piled on top of each other, and moved as close as possible to the GPU on a common silicon substrate. The latter was called «interposer». It is a kind of ordinary crystal, on which, instead of logical components, there are layers and connections for signal transmission.
In other words, «interposer» is a kind of analogue of a printed circuit board, but with a much denser laying of lines.
As a result, it became possible not only to spread a bus several thousand bits wide, but also to reduce energy losses by reducing the length of submarine lines and using memory chips with a low frequency of operation. In addition, the area occupied by the GPU and the memory subsystem on the printed circuit board has been reduced by about a third. This simplifies the design of cooling systems, improves their efficiency and allows you to create compact yet powerful video cards.
In theory, everything looks very nice, but how are things in practice? You can be calm, because this is also a complete order. At the moment, the maximum HBM memory configuration is represented by four 1024-bit stacks of four chips, which operate at a frequency of 1 GHz. Through simple calculations, we get a bandwidth of 128 GB / s for each such stack, or 512 GB / s for all at the same time.
For comparison: AMD Radeon R9290X the corresponding figure is 320 GB / s, while the NVIDIA GeForce GTX TITAN X has 336 GB / s. In both cases, the difference exceeds 50%. Agree, it looks impressive. But we are talking only about the first generation HBM memory, while the release of the second is not far off, where instead of four memory chips in the stack, eight microcircuits will be used.
In terms of energy efficiency, the results are also amazing — a threefold advantage of the HBM standard over GDDR5. In absolute terms, this will be expressed as a saving of 15-25 W of consumed energy.
The only disadvantage of this technology at this stage of development is the limitation on the maximum amount of memory — 4 GB. As practice shows, some modern games (such as GTA V) sometimes require more space to store textures, which theoretically can lead to friezes or a drop in FPS levels. However, looking back at the number of advantages that the new video memory standard brings, you can close your eyes to this shortcoming.
Moreover, in the second generation, all 8 GB will already be available.
Technology support
The well-known AMD TrueAudio, AMD Eyefinity, AMD Mantle, AMD FreeSync and multi-GPU synchronization via the PCI Express interface have been supplemented with support for hardware video decoding in H.265 format and the DirectX 12 API. But the standard For some reason, AMD again bypassed HDMI 2.0, although a competitor has already implemented it on representatives of the NVIDIA GeForce GTX 9xx family a long time ago. Recall that with the help of this interface it becomes possible to transfer video content via the HDMI port in 4K Ultra HD format at a frequency of 60 Hz.
As for the technologies implemented at the software level, first of all noteworthy is the support for virtual resolution AMD VSR — an analogue of the DSR function from NVIDIA. Their essence is as follows: a high-resolution image is formed at the output, and then it is adjusted to the capabilities of the output device, thereby increasing the clarity of the picture.
Another interesting option is AMD Frame Rate Targeting Control (AMD FRTC). As you might guess from its name, it allows you to set the maximum FPS value for 3D applications. If the boundary mark is exceeded, the video card will automatically slow down, simultaneously reducing power consumption and the speed of rotation of the fans on the cooler. In our opinion, this is a rather popular option, because it is not uncommon for old games to be played at a frame rate of about 80 — 100 FPS, although there is no point in such high rates. It is much more appropriate to limit yourself to a comfortable 60 FPS, while reducing the load on the video card.
Indeed, the manufacturer promises us a performance increase of 20-30% compared to the flagship of the previous generation AMD Radeon R9 290X. At the same time, the novelty will be almost half as long and will comply with the Mini-ITX standard. What other video card can boast such a set of features?
NVIDIA, of course, also has compact solutions, for example, a mini version of the NVIDIA GeForce GTX 970.
But, judging by the above slides, it can hardly be considered a serious competitor for the AMD Radeon R9Nano. The only indicator by which it is able to get ahead of the tested novelty is a slightly lower TDP level — 160 W versus 175 W. Although in practice the difference of 15 W will be almost imperceptible, as evidenced by the thermal images of the PC during the gameplay.
As you can see, AMD Radeon R9 Nano has even a slight advantage in terms of heating. True, here you still need to make an allowance for the fact that the experiment was carried out at AMD, and the measurements were carried out not on an open stand, but through the ventilation grill of the computer case cover. However, this does not change the fact that the AMD Radeon R9Nano and NVIDIA GeForce GTX 970 are about equal in terms of heat dissipation.
If, according to this parameter, we compare the new product with the flagship of the previous generation, then the advantage will definitely be on the side of the «baby» AMD Radeon R9 Nano.
It consumes less energy and does not heat up as much.
In most cases, the difference in the maximum temperature of the GPU reaches 18 — 20°C.
AMD Radeon R9 Nano Specification vs. AMD Radeon R9 290X:
|
Model |
AMD Radeon R 9 Nano |
AMD Radeon R9 290X |
|
Graphics core |
AMD Fiji |
AMD Hawaii XT |
|
Microarchitecture |
AMD GCN 1.2 |
AMD GCN 1.1 |
|
Technical process, nm |
28 |
|
|
Number of transistors, billion |
8. |
6.02 |
|
Number of Stream Processors (ALUs) |
4096 |
2816 |
|
Number of texture units (TU) |
256 |
176 |
|
Number of Raster Operations Units (ROPs) |
64 |
|
|
Graphics core frequency (base / in Boost mode), MHz |
— / 1000 |
— / 1000 |
|
Memory type |
HBM |
GDDR5 |
|
Video memory size, GB |
4 |
|
|
Effective video memory frequency, MHz |
1000 |
5000 |
|
Video memory bus width, bit |
4096 |
512 |
|
Video memory bandwidth, GB/s |
512 |
320 |
|
Texture fill rate, GTexel/s |
256 |
176 |
|
Approximate capacity, TFLOPS |
8. |
5.6 |
|
TDP level, W |
175 |
250 |
|
Tire type |
PCI Express 3.0 |
|
|
Auxiliary Power Supply |
1 x 8-pin PCIe |
1 x 6-pin PCIe 1 x 8-pin PCIe |
|
Video interface kit |
3 x DisplayPort 1.2 1 x HDMI 1.4a |
2 x DVI-D 1 x HDMI 1.4a 1 x DisplayPort 1.2 |
|
Maximum resolution |
4096 x 2160 @ 60 Hz |
|
|
Dimensions measured in our test lab mm |
166 x 112 |
290 x 112 |
|
Manufacturer website |
AMD |
|
9 
19 Appearance and cooling system
At the same time, the cooling system does not protrude beyond the textolite and the interface panel. So there should be no problems with placing it in a Mini-ITX case. The main thing is that it should be equipped with at least two slots for expansion cards.
The entire top of the video card is covered by an aluminum casing with a hole in the center for installing a fan. As you can see, AMD abandoned its famous «turbine» in favor of a more familiar option.
The following set of interfaces is provided for image output:
- 3 x DisplayPort 1.2;
- 1 x HDMI 1.4a.
Since the AMD Radeon R9 Nano belongs to the category of top video cards, it is quite logical that the manufacturer abandoned analog interfaces and bulky DVI-D during its design. In all cases, 4K Ultra HD (4096 x 2160), however, with a different maximum frequency. For HDMI 1.4a video output it is 30Hz, and for DisplayPort 1.2 it is 60Hz. Through the use of AMD Eyefinity technology, the image can be displayed simultaneously on up to six monitors.
The active part of the cooling system used here is a 4. 2W FirstDo FDC10h22D9-C 86mm low profile fan. Such turntables have been installed everywhere by different manufacturers since the days of AMD Radeon R9 models.280 / 280X. Therefore, in case of failure, it can be easily replaced with another one.
Hidden behind the shroud is a massive heatsink, which structurally consists of a copper evaporation chamber and aluminum fins attached to it. For faster and more efficient heat dissipation, two 8mm heat pipes are also used. Please note that not only the GPU is in contact with the cooler, but also four stacks of HBM memory.
The next component of the cooling system is a metal plate covering most of the textolite and in contact with the power elements of the power converters. Again, for greater efficiency in the area of \u200b\u200bthe VRM phases, the manufacturer used a heat pipe and an additional radiator.
PCB Implementation Features
When AMD Radeon R9 Nano was designed, the PCB was built from scratch. Moreover, the manufacturer tried to use the highest quality and highly efficient components (mainly for surface mounting). The power converter for the graphics core includes four phases that are controlled by the IR3564B controller. The voltage for the HBM memory is generated using a simpler single-phase module. The back side of the board is not hidden from view behind a traditional metal plate, as it does not make much sense. After all, now the memory chips are located near the video core and are covered with a common radiator. Nevertheless, the X-shaped stiffening plate is present here. Still, the cooling system weighs almost half a kilogram, so strengthening the textolite will obviously not be superfluous.
The novelty is powered by a PCI Express slot and an 8-pin PCIe connector. This configuration can easily withstand a load of 225 watts, while the TDP of the AMD Radeon R9 Nano is 175 watts. This backlog will be in demand during overclocking of the video card.
The textolite does not have the usual connector for connecting the AMD CrossFireX bridge, since here the synchronization of several graphics adapters combined in one bundle occurs through the PCI Express interface. But the DIP switch has not gone away, which allows you to activate the backup BIOS in the event of a failure or unsuccessful firmware of the main chip.
The AMD Radeon R9 Nano is based on the AMD Fiji graphics chip, manufactured in 28nm using the advanced AMD GCN 1.2 microarchitecture. It includes 4096 stream processors with 256 texture units and 64 ROPs. The maximum speed of the GPU is 1000 MHz, the base frequency has not yet been reported.
The memory subsystem is represented by four stacks of SK Hynix chips with a total capacity of 4 GB, operating at an effective frequency of 1000 MHz. To connect them to the GPU, a bus with a width of 409 is used.6 bits, which provides a bandwidth of 512 GB / s. Such figures are a record for video cards available on the market today.
Cooling system efficiency test in practice
Unigine Heaven (left) and MSI Kombustor (right) stress test results
75°C mark, which is the target for the AMD Radeon R9 modelNano. The critical value, which the manufacturer does not recommend to overcome, is 85°C. As you can see from the screenshots, the video card itself selects such an operating mode so that the heating does not exceed 75 ° C. On the one hand, it seems to be good — you can be sure that the graphics chip will not overheat during operation. But on the other hand, such an algorithm actually slows down the video card itself, reducing its capabilities. So, for example, when the load was created by the Unigine Heaven benchmark, the GPU speed changed within 930 — 950 MHz. When running the more demanding MSI Kombustor test, the graphics core clock automatically dropped to 800-810 MHz. Recall that at the maximum it can be 1000 MHz. It is unlikely that a 20% slowdown will greatly please users. In our opinion, it was necessary to somehow set the limits differently so as not to limit the speed of the GPU so much. We do not exclude that this algorithm will be corrected in non-reference versions of AMD Radeon R9 Nano from third-party manufacturers.
But back to the cooling system. Depending on the stress test, the fan speed varied between 2000 and 2200 rpm. At the same time, the video card created a noticeable, but quite acceptable noise for a long pastime at the PC.
In the maximum fan speed mode (approximately 3600 rpm according to the built-in tachometer), the GPU temperature dropped to 62°C. The noise created at the same time stood out noticeably against the general background, and it’s by no means comfortable to call it comfortable. Although from a low-profile 86 mm fan, one could hardly expect anything else.
But in this case, this is not the main thing. Look at the frequency of the GPU. As before, it did not rise to 1000 MHz, but was at 815 MHz. Although the target temperature (75°C) was still as much as 13°C. From this we can conclude that the power saving mechanism here has too aggressive settings: the maximum load immediately leads to a drop in GPU speed. I am glad that real applications do not create such harsh conditions as the MSI Afteburner utility, and accordingly, during the launch of games, you can count on high frequency rates: 900 — 950 MHz.
However, it’s a little disappointing that we never saw the 1000 MHz value that the spec claims.
In the absence of load, the frequency and voltage on the GPU were automatically lowered, which led to its lower power consumption and heat dissipation. The fan in this mode rotated at a speed of about 1500 rpm according to the built-in tachometer. At the same time, there was practically no noise, and the temperature of the GPU did not exceed 33 ° C.
AMD Radeon R9 Nano — one small but very fast graphics card
The AMD Radeon R9 Nano graphics card does not have DVI connectors, but today this is hardly a disadvantage. This type of connection is used less and less
The case when you do not need to judge the performance of a video card by its appearance. AMD Radeon R9 Nano can really surprise with its high speed.
Over the past few years, you can see the growing popularity of compact desktop PCs in the world. If previously a small computer case almost immediately spoke of the modest capabilities of the system, today this is not at all the case. Even powerful gaming PCs can be quite compact.
Motherboard manufacturers produce many miniature boards in the Mini-ITX form factor with top chipsets that can run the fastest processors. Manufacturers of computer cases are not far behind, offering their own solutions.
Video cards are more difficult. Powerful external graphics accelerators have a fairly large heat dissipation and require advanced bulky cooling systems. Until recently, the fastest ITX graphics cards on the market were solutions based on the NVIDIA GeForce GTX 970.
Yes, the GTX 970 is a very good and fast graphics card, but it’s more of a midrange solution.
What if you need even more graphics performance, all on a small ITX system?
AMD Radeon R9 Nano (left) will easily overtake any other graphics card in this photo» /> The case when you should not judge the speed of a video card by its size. AMD Radeon R9 Nano (left) will easily overtake any other video card on this photo
AMD Radeon R9 Nano is a graphics card specially designed for compact systems. It is equipped with the new Fiji XT GPU. The same GPU is installed in the flagship AMD Radeon R9 Fury X video card, introduced this summer. AMD Radeon R9 Nano almost completely repeats the characteristics of AMD Radeon R9 Fury X (see table)
AMD Radeon R9 Nano data CPUZ
To begin with, it must be said that the new R9 video cardsThe Nano, R9 Fury X and R9 Fury X use a completely new type of HBM (High-Bandwidth Memory). Pay attention to the huge 4096-bit memory bus used here. Even in combination with a relatively low frequency, this provides the highest throughput.
Compared to GDDR5, HBM memory is much more economical in terms of power consumption, and its modules take up much less space, which allows you to make the video card more compact. This is what the developers took advantage of when designing the R9.Nano. But the R9 Nano is not only compact, but also economical (if you compare it with the R9 Fury X). The TDP difference is 100W!
Apparently, even a slight reduction in frequency compared to the R9 Fury X allowed us to use lower voltage levels to power the GPU. In addition, under load, the video card does not always work at the maximum GPU frequency. It can also be assumed that the most successful GPU crystals are selected for the R9 Nano.
AMD Radeon R9 maximum GPU frequency
As far as functionality goes, AMD Radeon R9 Nano is fine with that. Up to six monitors can be connected, the card has powerful hardware video decoding capabilities, x265 codec support, H.265 video data decoding, interesting proprietary technologies TrueAudio and AMD FreeSync, Mantle, etc. are supported. R9 onlyNano, but also other current AMD video cards — Virtual Super Resolution (VSR) and Frame Rate Target Control (FRTC).
Activation of Virtual Super Resolution (VSR) technology in the control panel
VSR — improves the image quality in games (according to the manufacturer) by processing the picture at a higher resolution and then downsizing to the screen resolution. In some games, turning on VSR also allows you to get a wider viewing angle.
Enable Frame Rate Target Control (FRTC) in Control Panel
FRTC is a manual FPS limit at a given level to reduce graphics card heat generation. For example, a graphics card is capable of 100 FPS in a game, but if you limit the level to 70 FPS, it will not work at full capacity, generating less heat and running with less noise.
Unfortunately, the AMD Radeon R9 Nano does not support HDMI 2.0, only HDMI HDMI 1.4a. With today’s rise in popularity of 4K panels, this can be called a disadvantage. HDMI 1.4a will not allow you to use a frequency above 30 Hz at 4K resolution, and not all devices have DisplayPort ports.
We called this video card economical, but this definition is relative in this case. TDP video card — 175 watts. To dissipate heat in such a small video card, an advanced cooling system is used. Under the outer casing there is a small recess for the fan, and almost all the space is occupied by the radiator. Heatpipes are installed at the base of the heatsink. By the way, most likely absolutely all AMD Radeon R9 Nano video cards on the market will be equipped with just such a reference cooling system.
The cooling system does its job well. Under load, the temperature of the GPU rises to around 74 ° C, which is quite normal, and at the same time the video card is not at all noisy.
There is only one power connector, but it is eight-pin
Traditionally for many AMD video cards, the R9 Nano has a switch that allows you to select one of the two BIOSes for operation
According to information on the net, many complain about the noisy R9 Nano chokes. Yes, we have noticed this in some (very few) applications. It is curious that in these same applications, to one degree or another, we noticed the squeak of throttles on most powerful video cards that passed through our Test Lab. Therefore, somehow separately blame AMD Radeon R9 for this0580
Test histograms:
* — the MSI GTX 970 Gaming 4G graphics card and MSI GTX 980 Gaming 4G were tested in December 2014. Accordingly, the results of these video adapters were obtained using the Drisen.
Specifications AMD Radeon R9 Nano
| Interface |