2015 has been a good year for those interested in high-performance small form factor (SFF) computing. Some of the most noteworthy releases have been ASRock's X99E-ITX/ac motherboard which brought an entirely new level of performance to consumer mini-ITX systems. And let's not forget about the array of mITX cases that we've seen from manufacturers such as Phanteks, Corsair, and Silverstone.
AMD is aiming to make its stamp on the SFF marketplace by giving gamers a graphics card based on its flagship GPU but in a mITX-capable form factor. That card is the 6-inch air-cooled R9 Nano which uses the same full-fat Fiji GPU found on the red team's flagship R9 Fury X water-cooled offering. At a little over 7.5 inches, the R9 Fury X was a small graphics card in its own right, but the Fury Nano takes over 1.5 inches off that length and mitigates installation headaches by not using the AIO liquid cooling unit.
The question that everybody is asking is ‘how will AMD's flagship Fiji GPU fair when deployed on a 6-inch air-cooled card?'
One of the most noteworthy points about the R9 Nano is that AMD deploys a full-fat Fiji GPU on the 6-inch card. Unlike the R9 Fury (non-X), there is no removal of processors or texture units when comparing the R9 Nano to AMD's flagship R9 Fury X. So how exactly is the TDP cut from 275W on the Fury X to 175W on the R9 Nano?
Firstly, the GPU Core clock is cut to a maximum of 1000MHz, which is down from the R9 Fury X's solid 1050MHz. The critical point here is the word maximum, because AMD's PowerTune algorithms dynamically adjust the actual core clock based on factors such as power usage and GPU temperature. Cutting the R9 Nano's board TDP to 175W gives the PowerTune process a target to aim at.
AMD also says that the 175W TDP is achievable thanks to the way in which the Fiji GPU scales. The chip architects have suggested that the Fiji GPU settings deployed on the R9 Nano act in the ASIC's ‘efficient' range, while the Fury X pushes the ASIC past that range in search of greater performance over core efficiency. There's also the HBM factor, which helps to shave precious Watts off the board's overall power consumption in comparison to the GDDR5 alternative.
Performance-per-inch is probably the single point that AMD would like to push with the R9 Nano. An early look at the specification sheet makes it hard to argue against that point. On paper, at least, the six-inch card opens up a whole new level of gaming performance to a SFF audience who have, up until now, been able to choose from a mini-ITX variant of Nvidia's GTX 970 as their highest-end small graphics card.
The R9 Nano's target is to give SFF gamers a miniature graphics card that is realistically capable of pushing playable frame rates at a 4K resolution without significant reductions in image quality settings. That's something that none of the current mITX-intended graphics cards on the market can realistically achieve.
But specifications and intentions don't translate into real-world performance, so let's take a look at how AMD's Radeon R9 Nano performs in the real world.
GPU | R9 390X | R9 290X | R9 390 | R9 290 | R9 380 | R9 285 | Fury X | Fury | R9 Nano |
Launch | June 2015 | Oct 2013 | June 2015 | Nov 2013 | June 2015 | Sep 2014 | June 2015 | June 2015 | Sep 2015 |
DX Support | 12 | 12 | 12 | 12 | 12 | 12 | 12 | 12 | 12 |
Process Node (nm) | 28 | 28 | 28 | 28 | 28 | 28 | 28 | 28 | 28 |
Processors | 2816 | 2816 | 2560 | 2560 | 1792 | 1792 | 4096 | 3584 | 4096 |
Texture Units | 176 | 176 | 160 | 160 | 112 | 112 | 256 | 224 | 256 |
ROP’s | 64 | 64 | 64 | 64 | 32 | 32 | 64 | 64 | 64 |
GPU Clock (Max) (MHz) |
1050 | 1000 | 1000 | 947 | 970 | 918 | 1050 | 1000 | 1000 |
Memory Clock (MHz) |
6000 | 5000 | 6000 | 5000 | 5700 | 5500 | 500 | 500 | 500 |
Memory Bus (bits) | 512 | 512 | 512 | 512 | 256 | 256 | 4096 | 4096 | 4096 |
Max Bandwidth (GB/s) | 384 | 320 | 384 | 320 | 182.4 | 176 | 512 | 512 | 512 |
Memory Size (MB) | 8192 | 4096 | 8192 | 4096 | 4096 | 2048 | 4096 | 4096 | 4096 |
Transistors (mn) | 6200 | 6200 | 6200 | 6200 | 5000 | 5000 | 8900 | 8900 | 8900 |
TDP (Watts) | 275 | 290 | 275 | 275 | 190 | 190 | 275 | 275 | 175 |
We have already taken a more in-depth look at the Fiji GPU in our review of the R9 Fury X, so head over to that page if you want to know more.
We received our sample card without retail packaging or an accessory bundle. The items that are bundled will depend upon the board partner's choices, although we would expect something along the lines of a HDMI-to-DVI plug, power cable adapters, and generic documentation.
At first look it is clear that AMD's Radeon R9 Nano is a tiny card in many senses. The aesthetic design is likely to appeal to gamers, with a largely black metal shroud contrasted by bold red written text. AMD uses a dark black fan which has a reflective finish.
The Fury X shroud CAD models were given to consumers who wanted to 3D print their own cover, and representatives on AMD's recent press call suggested that a similar thing could be done with the R9 Nano.
A single (roughly-92mm) axial fan is used to cool the Radeon R9 Nano. The entire black, brushed aluminium shroud is sturdy and acts as a thermal control mechanism. The black colouring helps to manage the transfer of heat via radiation, and the aluminium material acts as additional surface area to be used for convective cooling.
The target temperature set for the R9 Nano by AMD is 75°C, but the card will throttle and heavily ramp up fan speeds when (if) 85°C is reached. Our testing showed the board running happily at 75°C, and when a higher thermal load is applied (overclocking) the fan simply runs faster to compensate. Core clock speed dynamics seem heavily weighted towards power usage of the card, rather than GPU temperature, for the default settings.
AMD's reference design does not utilise a backplate, and it seems that board partners are suggesting that this is not only related to early press samples of the card. I would have liked to see a backplate on the card, not for rigidity or aesthetics, but to spread the thermal load across a large cooling plate.
There would also be the benefit of a backplate acting as protection for any straggling cables inside a SFF build touching the rear PCB and suffering thermal issues (melting).
Vents on the side of the card run parallel to the heatsink's fin array. AMD representatives were keen to indicate that the fin array has been designed run in the card's front-to-back direction, rather than side-to-side.
The goal here is to ensure that more hot air is expelled out of the case – the side-to-side fin orientation typically spills heated air onto the motherboard PCB and nearby chassis side panel.
This is where that front-to-back fin array orientation starts to show a minor sign of potential weakness. AMD chose to leave the front end of the card completely open, rather than enclose it with a shroud or extension of the metal plate. With that said, there was little choice because enclosing the front end would likely stagnate the flow of air and create bigger cooling headaches.
The open design gives hot air that has been processed through the fin array freedom to escape inside the chassis with very little resistance (as my simple hand testing and our thermal imaging proved later in the review). While there is the argument that the overall pressure gradient makes it slightly more favourable for air to travel out the rear of the chassis rather than inside it (due to temperature differences between the enclosed volume of chassis air and the outside environment), a design tweak could have made certain that less hot air spills directly into the case.
AMD would likely have conducted internal testing that made this design most favourable for the design goals, however it will have the consequence of increasing internal chassis temperatures as opposed to blower-style coolers which dump the majority of their hot air straight out the back of the case.
With the R9 Nano's 175W typical board power rating, a single 8-pin power connector is allocated. The single 8-pin connector is a favourable approach (over dual 6-pins) for cards intended to be used inside SFF systems. It is far easier to route one cable in a tight space than two, even if the 8-pin PCIe connector is slightly less common.
The frontwards-facing (rather than side-mounted) orientation of the 8-pin power connector dictates an additional few centimeters of clearance to ensure that the power plug can be inserted. This may be part of the reason for AMD opting for a six-inch card, rather than use the full 6.7″ length that a mini-ITX motherboard is designed upon.
AMD's R9 Nano is a smidgen wider than the PCIe slot, so width-related interference issues are unlikely to be of any concern.
Reference R9 Nano models feature a BIOS switch which is likely deployed for redundancy purposes.
Three full-sized DisplayPort 1.2 connections and a single HDMI output form the connectivity options. Unfortunately the HDMI port does not conform to the HDMI 2.0 standard, so users hoping to get 4K 60Hz gameplay on a large TV will have to go searching for a DisplayPort-to-HDMI 2.0 adapter.
No HDMI 2.0 is unquestionably a problem to users wanting the R9 Nano for use inside a powerful gaming HTPC. AMD did, however, point out that the necessary DisplayPort adapters will be hitting the market, but the time scale is still unknown.
All four display connectors are squeezed onto a single row, with very little of the precious airflow grille space being wasted.
A look at the naked board shows how AMD utilises PCB space. Forming the 4GB VRAM capacity, four stacks of 500MHz HBM chips can be seen merging the interposer and the Fiji GPU. The application of on-package HBM chips is critical in reducing PCB sizing requirements and also power consumption compared to GDDR5. They also give the R9 Nano a 4096-bit bus which creates 512GBps of memory bandwidth when combined with the 500MHz clock.
Four primary power delivery phases feed the GPU solution and these are managed by an International Rectifiers controller. The MOSFETs are cooled by their own dedicated heatpipe and fin array solution which receive near-direct airflow from the 92mm fan.
It is good to see MOSFET cooling being taken seriously on a high-powered graphics card. AMD and Nvidia's reference coolers usually take good care of the power delivery components, however many board partner solutions rely upon small passive heatsinks or even incidental airflow to cool un-covered MOSFETs.
An indentation is cut into the aluminium fin array in order to make room for the relatively thick 92mm fan. As we have already discussed, the fin orientation is from the front to the rear of the card. This helps to reduce the amount of hot air that spills inside a chassis, although it does mean that a significant proportion will be blown directly out of the front of the card, as well as through the rear vent.
AMD uses a vapour chamber cooling system in order to manage the GPU's roughly 175W thermal load. A large copper plate makes direct contact with the GPU and, further up in the system, flattened heatpipes transfer thermal energy to outer edges of the fin array.
A vapour chamber cooling system is typically seen as an optimal solution for power-hungry but small graphics cards. The thermal capacity that the vapour chamber system provides in a relatively small volume makes it a common selection for such applications.
Even by the standards of SFF-intended graphics cards, the R9 Nano is still notably small. AMD's air-cooled Fiji-based card shaves about 1.5 inches off the length of a R9 Fury X, which in itself is particularly small for a high-end graphics card.
More importantly to many users with compact SFF cases, the difficulty of allocating a 120mm AIO radiator mounting location inside a mITX chassis is alleviated with the R9 Nano.
As a note, the six-inch R9 Nano is less than half the length of Sapphire's R9 390X Tri-X. Reference cooler models of Nvidia's similarly-priced GTX 980 Ti are close to 4.75-inches longer than the R9 Nano.
No banana for scale, however a 3.5″ HDD, 4.7″ LG Nexus 4, and 2.5″ SSD may be more fitting comparisons.
The card's six-inch length means that it does not overhang the edge of a 6.7″x6.7″ mini-ITX motherboard such as Asus' Z170i Pro Gaming. That additional seven tenths of an inch is useful for allocating clearance to the 8-pin PCIe power cable.
On paper (or screen), AMD's R9 Nano shows up featuring a GPU core clock speed of 1000MHz. In reality, this figure is actually up to 1000MHz, meaning that the ASIC is capable of a Gigahertz but the PowerTune algorithms will decide the instantaneous GPU clock speed based primarily on power consumption and temperature factors.
We know that the Fiji GPU is capable of around 1100-1150MHz when deployed on the R9 Fury X, so we pushed to see how far we could overclock that same GPU when it is installed on the R9 Nano card.
Excuse the spoiler, but given the R9 Nano's particularly dynamic nature when it comes to operational core clock speed, we added a section in our chart which presents the ‘real' core clock speed observed. We must point out that throughout our testing, the games' 1440P resolutions (which Heaven ran at due to its lack of a 4K option) allowed AMD's card to hit a higher average clock speed than the identical 4K tests did.
With that said, we found that Unigine Heaven's 1440P test was a good indicator of the actual core clock speed that the card was typically hitting in games, on average; Tomb Raider and Shadow of Mordor would sink a little lower on the core clock graph at 4K, while GTA V was happy to allow a higher core clock at the 8MP resolution.
The overclocking chart above tells only a small portion of the story. I spent the best part of a day playing with various frequency-related settings for the R9 Nano board and core clock. My findings were that simply increasing the allowable power limit through AMD's Catalyst Control Center (or third party software) was the quickest and most effective way of netting a speed boost which actually held during games and benchmarks.
That final part of the sentence is critical due to the complexity and operational nature of the PowerTune algorithms that AMD has deployed in the R9 Nano. I saw that better performance gains were provided by applying a +20% board power limit than increasing the maximum core clock to 1050MHz. The card seems very much power-limited in order to meet AMD's usage scenario targets, so playing with the power percentage slider nets the best overclocking results – this is certainly the case in my experience.
With that said, some of the tests (GTA5 was one and 3DMark was another), especially at a sub-4K resolution, were able to hit the 1000MHz core clock limit when the card was given a +30% power allocation. Increasing the core clock to 1050MHz (1100MHz proved unstable for our card) netted healthy gains for GTA V and 3DMark, although it presented no difference for the likes of Tomb Raider and Shadow of Mordor – both of which could not hit the 1GHz maximum core clock, never mind 1050MHz. This was using a 4K test resolution.
Comparing the card's frequency logs for a stock-clocked and an overclocked 4K run of GTA V shows how the power setting and increased maximum core clock affect the usage dynamics. With 30% added to the power budget and the target core clock upped to 1050MHz, the final run scene of GTA V's benchmark at 4K is able to see the average a little over 1GHz on the core clock. That number is almost 100MHz lower when using default settings.
We are not saying that allowing the board to receive 30% more power is sensible for daily usage, however it does highlight that AMD has deployed a capable power delivery system with a strong amount of overhead. We had no overheating issues throughout our overclocked testing – the fan speed simply ramped up to a greater velocity in order to dissipate the increased thermal load.
Let me make that perfectly clear again. We are not saying that a +30% power limit is the most sensible setting to apply while also considering the long term preservation of the hardware. But it does serve as a quick method of netting higher performance results that show what the R9 Nano (more specifically, our single sample card) can handle if you give it enough power.
Overclocked performance will be outlined throughout the review.
For the review today we are using the Catalyst 15.201.1102 driver that was supplied to us by AMD for the R9 Nano launch. Unfortunately due to a very tight schedule with the card, we did not have enough time to re-test every other piece of Nvidia and AMD hardware using the latest drivers. With that said, even the oldest drivers that we used are very recent (AMD Catalyst 15.7.1 and Nvidia 355.60).
Our testing will take place inside a well-cooled NZXT Phantom 630 chassis with its side panels removed. This is the exact same test configuration that we have used for all comparison graphics cards, so it ensures consistency, even if the R9 Nano is unlikely to be used inside such a large full tower chassis. The point of this is to initially show R9 Nano performance without any thermal handicaps induced by a SFF test environment.
We test using the 2560×1440 and 4K resolutions because gamers purchasing a £500+ card are likely to be running monitors with those screen resolutions.
Overclocked performance for the R9 Nano will be included in the 4K tests for two reasons; 1 – 4K demands every last slither of performance that a graphics card can deliver and 2 – we simply did not have enough time to test (and re-test) the overclocked card through our 1440P performance suite.
We will also be taking a look at how the R9 Nano performs inside a high performance SFF gaming build that will be at home under a large TV or next to a high-res monitor. For this test scenario, we will be running a small number of benchmarks at 1920×1080 and 4K, as they are the most popular TV resolutions for users with a gaming HTPC.
The objective here is to see if the card shows significant performance degradation and signs of thermal throttling when the card is crammed inside its most applicable usage environment.
We test using the AOC U2868PQU 4K monitor that scored highly in our review that can be found HERE. AOC's 28″ offering is one of the most affordable 4K monitors on the market and makes the realms of 4K gaming more accessible to a wider audience.
Test System
- Processor: Intel Core i7 5960X ES (4.4GHz OC).
- Memory: 16GB (4x 4GB) Corsair Vengeance LPX 3200MHz DDR4.
- Motherboard: Asus X99-Deluxe.
- System Drive: 500GB Samsung 840.
- CPU Cooler: Corsair H100i.
- Case: NZXT Phantom 630.
- Power Supply: Seasonic Platinum 1000W.
- Operating System: Windows 7 Professional with SP1 64-bit.
Graphics cards (comparisons):
- AMD R9 295X2 (Catalyst 15.7.1)
- AMD R9 Fury X (Catalyst 15.7.1)
- Asus R9 390X STRIX OC (Catalyst 15.201.1102 – R9 Nano launch)
- Sapphire R9 380 ITX Compact (Catalyst 15.201.1102 – R9 Nano launch)
- Nvidia GTX Titan X reference (ForceWare 355.60)
- EVGA GTX 980 Ti Classified ACX 2.0+ (ForceWare 355.60)
- Nvidia GTX 980 Ti reference (ForceWare 355.60)
- Asus GTX 980 STRIX OC (ForceWare 355.82)
- Asus GTX 970 DCU Mini (ForceWare 355.82)
Software:
Unigine Heaven Benchmark
3DMark
Fraps
Steam Client
MSI Afterburner
TechPowerUp GPU-Z
Games:
Battlefield 4
Bioshock Infinite
Grand Theft Auto V
Metro: Last Light
Middle Earth: Shadow of Mordor
Tomb Raider
Game descriptions edited with courtesy from Wikipedia.
3DMark is an essential tool used by millions of gamers, hundreds of hardware review sites and many of the world’s leading manufacturers to measure PC gaming performance.
Futuremark say “Use it to test your PC’s limits and measure the impact of overclocking and tweaking your system. Search our massive results database and see how your PC compares or just admire the graphics and wonder why all PC games don’t look this good.
To get more out of your PC, put 3DMark in your PC.”
3DMark has the R9 Nano slightly outperforming a factory-overclocked GTX 980 at 1080P, and widening that performance lead at 4K. The Fiji GPU is, by now, a known entity when it comes to resolution scaling – at 4K is where the Fiji- and HBM-based cards show their best relative performance, while the competition is typically able to claw back performance at lower resolutions.
AMD's water-cooled R9 Fury X shows what the unconstrained Fiji GPU is capable of in 3DMark. While the R9 Nano does sit roughly between the factory OC GTX 980 R9 Fury X in terms of performance, its numbers are closer towards the former than the higher-performing latter.
Unigine provides an interesting way to test hardware. It can be easily adapted to various projects due to its elaborated software design and flexible toolset. A lot of their customers claim that they have never seen such extremely-effective code, which is so easy to understand.
Heaven Benchmark is a DirectX 11 GPU benchmark based on advanced Unigine engine from Unigine Corp. It reveals the enchanting magic of floating islands with a tiny village hidden in the cloudy skies. Interactive mode provides emerging experience of exploring the intricate world of steampunk. Efficient and well-architected framework makes Unigine highly scalable:
- Multiple API (DirectX 9 / DirectX 10 / DirectX 11 / OpenGL) render
- Cross-platform: MS Windows (XP, Vista, Windows 7) / Linux
- Full support of 32bit and 64bit systems
- Multicore CPU support
- Little / big endian support (ready for game consoles)
- Powerful C++ API
- Comprehensive performance profiling system
- Flexible XML-based data structures
We set Quality to ‘Ultra', Tessellation to ‘Normal', Anti Aliasing to 2x, and the resolution to 2560×1440.
With Unigine Heaven running at a 1440P resolution, not 4K, the stock-clocked R9 Nano is able to match the performance of Asus' factory-overclocked GTX 980 STRIX.
Opening the power taps and increasing the target clock speed sees the R9 Nano jump above Asus' GTX 980 STRIX and sit directly between performance of the factory-overclocked Nvidia solution and AMD's flagship R9 Fury X.
The AMD R9 Nano's average core clock speed for the roughly 250 second run of Unigine Heaven was around 900MHz when using the default card settings. In this usage case, power delivery limits seem to be the primary driver for PowerTune core clock management, as the GPU temperature remained below 75°C.
Battlefield 4 ’s single-player Campaign takes place in 2020, six years after the events of its predecessor. Tensions between Russia and the United States have been running at a record high, due to a conflict between the two countries that has been running for the last six years. (Wikipedia).
For 2560×1440 testing, we opted for Battlefield 4's Ultra preset. When testing with the demanding 4K resolution, we opted for Battlefield 4's High IQ preset in order to make playable frame rates a greater possibility. Our frame rates are recorded using a section of the game.
Performance of the R9 Nano sits between that of a factory-overclocked Asus R9 390X STRIX and the reference R9 Fury X when running Battlefield 4 at 1440P. Hitting an average frame rate of 54 FPS when using the game's Ultra preset is solid performance and realistically translates into a perfectly enjoyable gameplay experience from this six-inch card. Asus' GTX 970 DCU Mini is significantly slower than the R9 Nano at 1440P, although that is to be expected given the price differential.
4K performance still places the R9 Nano between a Fury X and R9 390X, however the increased resolution allows AMD's flagship single-GPU card to widen its performance lead. Why is this given that the performance gap should be constant between the two Fiji-equipped graphics cards? It's simply a case of the higher-stress 4K resolution forcing a reduction in the average GPU core clock that the R9 Nano reaches. The 275W water-cooled Fury X has no such clock-speed constraints to contend with.
AMD's R9 Nano out-guns the Asus GTX 970 DCU Mini by 32% at 4K, making it the fastest SFF card in Battlefield 4. It's also significantly more expensive than the Nvidia-based solution and further image quality reduction would be required to achieve enjoyable frame rates even on the R9 Nano at 4K. The overclocked R9 Nano does, however, manage to sit in a comfortable range of frame rates for good usage with a 4K FreeSync screen.
Set in 1912, in Bioshock Infinite, players assume the role of former Pinkerton agent Booker DeWitt, sent to the flying city of Columbia on a rescue mission. His target? Elizabeth, imprisoned since childhood.
Bioshock Infinite is set to its highest image quality settings at both resolutions. We test a section of the game. In our previous graphics card review (of EVGA's GTX 980 Ti Classified) we removed the game's minimum FPS reading as our current test method seems to be imposing an unjustified penalty against AMD cards so we are looking into this issue. The same charting process is conducted for this review.
More of the same is shown in Bioshock Infinite. The R9 Nano sits between a factory-overclocked GTX 980 and the reference R9 Fury X at 1440P, with performance being slightly closer to the former. Focusing on Nvidia's reference GTX 980 Ti, which is about the same price as the R9 Nano, an 18% performance increase is delivered for the green team's offering. But given the GTX 980 Ti's increased power draw, higher operating temperatures, and physically large sizing, that could be considered an apples-to-oranges comparison.
4K performance shows an almost identical trend to that of Battlefield 4. The well-cooled, power-happy R9 Fury X is able to widen its performance lead over the R9 Nano thanks to fewer GPU core clock constraints.
Grand Theft Auto V is an action-adventure game played from either a first-person or third-person view. Players complete missions—linear scenarios with set objectives—to progress through the story.
Outside of missions, players can freely roam the open world. Composed of the San Andreas open countryside area and the fictional city of Los Santos, the world of Grand Theft Auto V is much larger in area than earlier entries in the series.
The world may be fully explored from the beginning of the game without restrictions, although story progress unlocks more gameplay content. (Wikipedia).
We set all of GTA V's settings to their highest levels for both resolutions but made sure to disable the performance-crippling MSAA settings. We saw the game use almost 4GB of VRAM at 2560×1440 and around 4.5GB at 4K (where possible with the relevant video cards). The built-in benchmark is used to gather performance data.
Performance data from the AMD R9 295X2 unfortunately could not make it into the charts as the card has a bug with our test monitor that prevents GTA V's benchmark from operating.
GTA V shows a preference towards Nvidia's hardware at 2560×1440. The Asus GTX 980 STRIX leapfrogs the R9 Nano, which now sits above the factory-overclocked R9 390X. A frame rate of almost 60 FPS means that the stock-clocked R9 Nano will deliver an enjoyable gaming experience in GTA V while using the 1440P resolution and Very High image settings.
4K sees the Fiji-based R9 Nano outperforming the GTX 980. Given that these GTA V settings and the 4K resolution will happily use more than 4.5GB of VRAM, the R9 Nano seems to be enjoying some of the same benefits that the Fury X saw thanks to its high-bandwidth memory. GTA V seems to be allowing the 4GB frame buffer to use its speed as a way of alleviating the higher VRAM usage that other graphics cards were hit with.
GTA V was one of the more positive game engines when it comes to allowing the R9 Nano's GPU core clock to flex its speed. At 1440P the R9 Nano was able to hit its 1GHz maximum core clock and the average reading was in the high-900MHz range. 4K also saw the card run at an average core clock of more than 900MHz.
The overclocked R9 Nano configuration saw a healthy performance gain at 4K. The average GPU core clock speed sat a little over the 1GHz mark, although this came at the price of a noticeably higher fan speed increasing noise output.
Metro: Last Light takes place one year after the events of Metro 2033, proceeding from the ending where Artyom chose to call down the missile strike on the Dark Ones. The Rangers have since occupied the D6 military facility, with Artyom having become an official member of the group. Khan, the nomad mystic, arrives at D6 to inform Artyom and the Rangers that a single Dark One survived the missile strike.
4A Games’ proprietary 4A Engine is capable of rendering breathtaking vistas, such as those showing the ruined remnants of Moscow, as well as immersive indoor areas that play with light and shadow, creating hauntingly beautiful scenes akin to those from modern-day photos of Pripyat’s abandoned factories and schools.
We opted for the High quality setting at 4K and Very High for 1440P testing. AA was kept off, AF was applied as 16x, and tessellation was set to normal. The minimum FPS readings do not highlight anything more important than very occasional drops in one of the three runs.
Metro: Last Light sees Asus GTX 980 STRIX outperforming the R9 Nano at 1440P. The similarly-priced GTX 980 Ti is 17 FPS (27%) faster than the R9 Nano at this resolution.
Running the 4K resolution sees the R9 Nano jump back above Asus' GTX 980. AMD's six-inch R9 Nano outperforms the significantly cheaper mini-ITX GTX 970 by 8 FPS (24%) at 4K.
In Middle-earth: Shadow of Mordor, the player plays as a ranger by the name of Talion who has wraith-like abilities.In this open world video game, players have the freedom to pursue side quests and roam around the world.
We use a mixture of Ultra and High settings in Shadow of Mordor's for both resolutions and tested using the built-in benchmark. These quality settings were initially seen as the Ultra preset on our test system, although an adjustment in the game has shown to now be a custom profile.
This time it's the Asus R9 390X that overtakes the R9 Nano in terms of 1440P performance. AMD's R9 Nano is slightly faster than the GTX 980 STRIX at 1440P, although it cannot come close to the performance of the liquid-cooled Fury X.
The factory-overclocked R9 390X holds its position at 4K, which is likely thanks to its large – 8GB – frame buffer which Shadow of Mordor will happily eat into.
Shadow of Mordor was one of the games that put a particularly high tax on the R9 Nano in regards to GPU core clock. The 1440P setting saw the R9 Nano's GPU core clock average a little over 900MHz for the short benchmark run, while 4K dropped the core to below 900MHz on average.
Tomb Raider received much acclaim from critics, who praised the graphics, the gameplay and Camilla Luddington’s performance as Lara with many critics agreeing that the game is a solid and much needed reboot of the franchise. Much criticism went to the addition of the multiplayer which many felt was unnecessary. Tomb Raider went on to sell one million copies in forty-eight hours of its release, and has sold 3.4 million copies worldwide so far. (Wikipedia).
We test using the Ultimate quality profile for both resolutions. The in-game benchmark is used.
Tomb Raider has the R9 Nano slightly ahead of Asus' GTX 980 STRIX, although this result is close enough to be considered a draw. The R9 Fury X and GTX 980 Ti shows healthy performance leads over the similarly-priced R9 Nano.
4K testing opens a slight gap for the R9 Nano over its R9 390X and GTX 980 competition. As with 1440P though, the similarly-priced GTX 980 Ti and R9 Fury X are noticeably faster than the R9 Nano.
The logic behind testing the R9 Nano and its nearest price competitor – Nvidia's reference GTX 980 Ti – inside a high-end SFF build is to see how the AMD card performs in its intended environment and to see if Nvidia's competitor can maintain its performance lead with constricted cooling space.
We see this test as logical because many SFF cases (especially those targeting gamers and enthusiasts) are perfectly capable of housing cards as large as the GTX 980 Ti. 1920×1080 and 4K were the resolutions used as these are the most popular TV resolutions for gamers interested in using the graphics cards inside a SFF gaming HTPC to power a large screen.
In order to alleviate bottlenecks and build a realistic high-performance gaming SFF system, we opted for a mixture of enthusiast gaming hardware. The Intel Core i7-6700K processor was stock-clocked (because that's fast enough for gaming) and sat on Asus' Z170I Pro Gaming mITX motherboard with 16GB of Crucial DDR4-2400MHz memory.
Storage consisted of a Crucial BX100 250GB SSD for the OS and some games, as well as a 256GB Plextor M6e for additional games. Seasonic's excellent 760W Platinum unit was used for power.
Star of the show was Silverstone's SUGO SG13 mITX chassis. This is a superb case with smart design features from front to back. Most importantly, it is one of the smallest mITX cases on the market but is actually able to fit graphics cards as large as the reference GTX 980 Ti or Titan X.
Unsurprisingly, the faster GTX 980 Ti reference graphics card comfortably out-paces the R9 Nano at 1080P testing. Both cards comfortably deliver playable frame rates with taxing image quality settings, however the GTX 980 Ti is simply quicker.
Perhaps less obviously, the R9 Nano is also outpaced by the reference GTX 980 Ti at 4K. The higher thermal stress of 4K game rendering simply forces both cards to increase the fan speeds in order to provide adequate cooling.
We initially thought that Nvidia's blower-style cooler may force the GTX 980 Ti to start throttling with the demanding thermal load. But in reality this wasn't an issue – the fan speed was instead increased to keep the GPU core temperature in check. Pretty much the same logic applies to AMD's R9 Nano.
This is where we start to see usability differences between the two similarly-priced solutions. Nvidia's 250W TDP GTX 980 Ti is dealing with a significantly greater heat output, and this causes high temperatures towards the rear of the chassis. What I will say, however, is that the blower cooler design is very effective in venting the vast majority of the hot air outside of the chassis.
AMD's R9 Nano, on the other hand, runs with significantly lower surrounding temperatures thanks to its 175W typical board power rating. The caveat is that a large proportion of the hot air is expelled directly into the chassis due to AMD's cooler and shroud design.
I highlighted earlier in this review that the open front side of the card would be an escape for hot air to work its way into the chassis environment. Thermal imaging proves this point because the hottest part of the case's side panel was towards the front section.
With a large portion of the heated coolant air being exhausted inside the chassis, a system based on the R9 Nano is likely to experience higher CPU and motherboard temperatures than a system with a blower-style GPU cooler like that used on the reference GTX 980 Ti.
There is indeed a good amount of heat going through the rear vents, although the previous thermal imaging shots prove that the heat is travelling in both directions – out of the case and back into it.
Looking at clock speeds throughout testing in the SFF chassis, there was little difference between those displayed inside the well-cooled ATX case. Silverstone's smart design (and many other mITX cases are similar) gives a direct breathing point for the graphics card fan. This allows the temperature levels to be kept in check and the GPU core clocks to operate without throttling.
We did, however, notice that both the GTX 980 Ti and the R9 Nano were operating with higher fan speeds in the SFF system (compared to the ATX system) while under heavy load.
The tests were performed in a controlled room with temperatures maintained at a constant 25°C – a comfortable summer environment for the majority of people reading this. Idle temperatures were measured after sitting at the desktop for 15 minutes. Gaming measurements were acquired by repeating the GTA V benchmark for 15 minutes and taking the stabilised reading. All fan settings were left on automatic.
AMD's temperature target is 75°C and that's pretty much where the card will reach after a long enough session of gaming. The fan speed ramps up to a user-definable maximum (65% by default) in order to hit the target GPU temperature and stay well away from the 85°C throttling point.
The rear PCB hits a temperature level in the 60s very quickly after load is applied. The card's hot-spot is directly behind the VRM components, and temperatures climbing towards the high-80s quickly after load being applied are borderline concerning.
I say ‘borderline' because the VRM components themselves are perfectly capable of running at temperatures that most people would perceive as worrying (100°C plus is fine for certain MOSFETs and inductors). But that heat spreads along the PCB and soon reaches its way to the GPU area, forcing temperatures higher. There's also the concern of a fan cable or some other connector contacting the PCB inside a SFF system. This could cause the cable plastic to melt, and that's a bigger headache.
I would have liked to see AMD make use of a backplate on the card. Not only would this help to effectively (and easily) distribute heat away from the MOSFET area, it would also act as a protection barrier for cables that are touching the card. I'm not saying that these temperature levels should have you worried about melting cables, but if the plastic lead from one's cheap fan is going to melt, I'd rather it do so on a metal backplate than the card's expensive PCB.
Due to the city-centre positioning of our test location, constantly high ambient noise makes it very difficult to accurately measure the difference in the noise level of each graphics card. As such, we have instead decided to record a short video to show how the fan speed of AMD's R9 Nano operates when it is housed inside a SFF chassis.
The R9 Nano is barely audible when operating at its idle fan speed level (19-22%) in our testing. Once the card's fan moves towards 40%, the noise output is clearly more noticeable, although it's not intrusive. I would be hard pushed to notice the load fan speed when sat 2m+ away from a large screen TV and the SFF gaming system.
Subjectively, I found the loaded R9 Nano slightly quieter than the loaded GTX 980 Ti reference inside the SFF test system. However this is only a subjective point (not scientifically tested) and the real-world difference is very small.
What was noticeable was the coil whine that our R9 Nano sample exhibited. As soon as a GPU load was applied, the inductors would put out a high-pitched squeal that was undeniably frustrating to hear. Did this act as an annoyance during the gaming experience? Without headphones on, in my opinion, yes it did. But that is just my opinion and you may have different tolerance levels.
We must point out that we only have one sample to test – other reviewers' R9 Nano cards may not suffer any coil whine. It's worth noting that retail samples or revisions may alleviate the coil whine that we heard. My colleague Allan heard coil whine on his early R9 Fury X card, however the newer Sapphire R9 Fury X that I tested with did not output any noticeable whine.
We measure the amount of power drawn from the wall by the entire test system. Our Core i7-5960X CPU is heavily overclocked and has a large bearing on the power draw readings when it is heavily loaded in GTA V (our chosen game for system readings). The important information is the difference in power draw levels between each card.
The GTX 980 is a 165 TDP part and AMD's R9 Nano is a 175W TDP card. So it comes as little surprise that overall system power consumption is just under 10W more than that of a GTX 980, albeit a factory overclocked one.
AMD has done a very good job at constricting the power usage of the R9 Nano. We cannot accurately validate AMD's claims that the R9 Nano is utilising the ‘efficient' range of the Fiji GPU, but power draw numbers compared to a R9 Fury X would suggest that is indeed the case.
The AMD Radeon R9 Nano is a unique graphics card in many ways. It utilises a flagship GPU but houses it on a board which measures in at six inches long to make it a true mITX graphics card.
As far as small form factor graphics cards go, AMD has comfortably taken the performance crown with the R9 Nano. The previous fastest offerings were mITX GTX 970 designs from the likes of Asus. The fastest offering on AMD's side was the Sapphire R9 380 Compact ITX. Both of those cards were comfortably outperformed by the R9 Nano, but that is to be expected given the latter's significant price premium.
Overall performance sees the AMD R9 Nano typically sat between factory-overclocked GTX 980 and stock-clocked R9 Fury X levels, with most frame rates being slightly closer to the GTX 980 than the R9 Fury X. While both of those cards are valid offerings in their own right, they are not the most convenient solutions for application in many SFF cases. And that's where the AMD R9 Nano is able to offer something different – generally acceptable 4K performance from a truly mini-ITX graphics card.
Power consumption of the R9 Nano has been handled well by AMD. The card requires around 10W more than the GTX 980 and it uses that increased power budget to deliver greater performance on average. Temperatures have also been handled well, in general. The core is happily maintained at its 75C target without the fan having to run at annoying levels.
Temperatures of the rear components (namely the rear VRM area) are high though, so I would have liked to see AMD apply a heat-spreading backplate to the card. There's also the cooler design which exhausts a sizeable proportion of its hot air directly inside a chassis.
The R9 Nano uses the same full-fat Fiji GPU found inside the R9 Fury X. But the latter is consistently faster. Why? That's because AMD rates the R9 Nano GPU for a core clock of ‘up to 1000MHz‘. Our testing showed that (namely) the card power allocation forced the GPU to operate around the 900MHz mark on average, although this varies from game to game, with GTA V allowing higher clocks and Shadow of Mordor and Tomb Raider being less friendly to the GPU.
Pricing is going to be the R9 Nano's big talking point. AMD has set the retail price at £515, which is pretty much identical to that of the R9 Fury X and similar to the retailer selling price of reference GTX 980 Ti models. Indeed those other two cards are different to the R9 Nano in many respects, but they are also competitors in many senses.
When I was looking for a mITX chassis to purchase, it struck me just how many are designed to accommodate large graphics cards. Even our tiny Silverstone Sugo SG13 is able to house a GTX 980 Ti comfortably, and a R9 Fury X (and its water-cooler) at a push, although the latter is arguably less advisable or convenient. By that logic, they're both competitors to the R9 Nano for users with appropriate SFF cases.
The AMD Radeon R9 Nano is a smart graphics card from AMD that brings a new level of performance to the market for users interested in buying a truly SFF graphics card. It does, however, charge a significant pricing premium for the luxury of reduced size when comparing it to its closest performance competitor (the GTX 980).
Power usage is good, temperatures are in check, and noise output was bearable, so AMD's R9 Nano is a valid option is you are happy to pay the price premium to get a high ratio of performance-per-inch without having to deal with the AIO liquid cooling solution of the Fury X.
Discuss on our Facebook page, over HERE.
Pros:
- Brings new levels of performance to the SFF graphics card market.
- Comfortably outperforms other SFF offerings (GTX 970 and R9 380).
- Capable of playable 4K performance.
- Cooling solution keeps the GPU temperatures in line.
- Cooling solution is not loud enough to cause annoyance.
- Power draw levels are very good at stock clocks.
- Increasing the power budget nets solid performance gains for increased fan speed and power draw.
Cons:
- Retail price is significantly above its closest performance competitor (the GTX 980).
- Cooler design exhausts some air directly into the chassis.
- Rear components can get hot and there is no backplate.
- No native HDMI 2.0
KitGuru says: The AMD R9 Nano is a unique product that caters directly to the SFF gaming market. With solid power consumption numbers, good performance even at 4K, and a convenient form factor, the R9 Nano should not be overlooked if you want a high-performance gaming HTPC. It's mainly the £515 asking price that is hard to swallow.