When AMD announced the Fury family of GPUs at E3 back in June, it promised that its new graphics hardware would ship in multiple high-end flavors. We've already covered the launch of the water-cooled, 7.5-inch Fury X and the larger, air-cooled Fury, but today's announcement covers what was arguably the most interesting card of all -- the six-inch Radeon R9 Nano. (The GPU was initially referred to as the AMD Radeon R9 Fury Nano, but AMD's marketing folks realized that's a bit of a mouthful).
It's very small
AMD has set a high bar for the Nano, with some aggressive performance claims and a tight power envelope. Now that the company is formally revealing the card, let's take a look at its specs and capabilities.
Positioning and specifications
The Fury Nano is aimed at customers who want an extremely small graphics card but aren't willing to sacrifice performance. At six inches long, it's exactly the size of a standard PCIe x16 port -- AMD executives told us at E3 that they could've made the board even smaller, if not for that limitation. Size isn't the only feature of the card, however. According to AMD, the Nano has a TDP of just 175W. The GPU's power configuration confirms that; the Nano has a single eight-pin PCIe connector, compared to two eight-pin connectors on the standard Fury X.
The Radeon Nano in a mini-ITX system.
Given its aggressive TDP target, one might expect AMD to have aggressively trimmed core counts and other features of the R9 Nano compared to the full Fury X. That's not, however, what the company chose to do. Specs on the Radeon R9 Nano are shown below:
The R9 Nano is a full Fury X, only on a smaller PCB. The TDP difference between this card and its higher-power sibling can be explained by two things. First, AMD is using a lower core clock speed for the GPU. While the Nano is capable of bursting up to 1GHz in certain scenarios, we expect its real-world clock to be in the 850-880MHz range. This can vary by application, however, and we don't have final silicon in-hand yet, so treat this as a ballpark.
You might not think that trimming a relatively modest 200-250MHz off the GPU core clock could slash power consumption by nearly 50%, but there's reason to believe AMD's figures. Generally speaking, power consumption in modern semiconductors rises as the square or cube of the voltage increase and in direct proportion to frequency. Since higher voltages are required for higher frequencies, this means that increasing the clock speed by a small amount can create a much larger increase in power consumption. AMD's FX-9590 (5GHz Turbo) has a 220W TDP, compared to the FX-8350's (4.2GHz Turbo) 120W TDP. A 19% increase in frequency requires an 1.83x increase in maximum dissipation. Intel CPUs suffer from similar trends, though the absolute figures are smaller, as we've previously covered.
The other reason that AMD can hit these power targets has to do with GPU binning. Any time AMD, Intel, or Nvidia builds a processor, they test the batch to determine which chips can operate at which frequency and voltage settings. While AMD didn't specifically comment on its binning practices with Nano, it's a safe bet that the company is reserving the very best of its Fiji GPUs for this product line.
As far as absolute performance is concerned, AMD is predicting that the Radeon Nano will deliver 30% more performance than the R9 290X in a fraction the space and 16dBA less noise.
Estimated performance
AMD is projecting that the Radeon R9 Nano will compare extremely well against the GTX 970 Mini, as shown below:
After some of the issues surrounding the Fury X's early benchmark leaks, many people will give these results a gimlet eye. According to the footnotes published in AMD's presentation, Far Cry 4 was tested at 4K at the Ultra High preset with 0x MSAA and 0x AF. AMD used its own Catalyst 15.201 and compared against Nvidia's 355.60 WHQL driver. All of the additional games were tested at High detail and 4K, with no MSAA or AF forced on or manually selected. FXAA was used in several titles, but it has a minimal impact on GPU performance.
Price, initial thoughts
AMD has announced that the Radeon R9 Nano will debut at $649, just like its highest-end Radeon R9 Fury X. If we take AMD's claims of offering 30% more performance than the R9 290X at face value, it means that the R9 Nano should be within 10% of the Fury X in the majority of benchmarks. Factor in the card's tiny form factor and 175W TDP, and it's easy to see why AMD wants to put a premium on this card. There's literally nothing like it on the market today.
AMD wants to put 4K-capable systems into living rooms -- without writing it into your prenup.
By keeping the Nano at $649, however, AMD is also setting a very high bar for itself. Even if we assume that the R9 Nano beats the GTX the 30-40% estimates that the graphs show, that still leaves AMD charging almost twice as much for its premium solution. Again, we'll have to wait for hardware to see if this is justified. At 175W, the R9 Nano still draws significantly more power than the GTX 970 Mini it compares against -- though it would still win past Nvidia's card in terms of performance per watt if AMD's test results are accurate.
AMD will ship the card the week of September 7, which is when reviews should begin to roll as well. We're very interested to see how this plays out.
