AMD R9 Nano review: Stellar performance in a pint-sized graphics card

AMD R9 Nano review: Stellar performance in a pint-sized graphics card
At E3 this past summer, AMD announced four new GPUs that would make up its next-generation product family: The Radeon Fury X, the Fury, the Fury Nano, and the as-yet-unreleased dual-GPU Fury.

At E3 this past summer, AMD announced four new GPUs that would make up its next-generation product family: The Radeon Fury X, the Fury, the Fury Nano, and the as-yet-unreleased dual-GPU Fury. We’ve covered each launch in turn, but the R9 Nano (AMD opted to shorten the name) is arguably the pinnacle of the entire Fury family. While it’s not expected to match the Radeon Fury X’s performance, it’s even shorter than the Fury X’s 7.5 inches. At just six inches, the Nano is exactly the same size as a PCIe x16 slot. Any shorter, and AMD would’ve had to compromise on bus bandwidth.

The Radeon Nano

In person, the card is every bit as impressive as it looks in pictures. At six inches long, it’s actually shorter than the first 3D accelerator I ever bought, the Diamond Monster II. GPUs have gotten larger and smaller over the years, but I don’t ever recall AMD or Nvidia fielding a GPU this small at the high end of the product stack. This opens up all manner of interesting system possibilities for mini-ITX or small form factor enthusiasts, and that’s the audience that AMD wants to target with this card.

The Radeon Nano in a mini-ITX system.

One note: We don’t currently have data to share on how the Nano performs specifically in a mini-ITX configuration. The chassis that we ordered for the review shipped late and missed its plane. As a result, our Coolermaster Elite 110 is currently sitting in Newark, NJ. For now, we’ve decided to review the card against both the GTX 970 Mini and its full-size cousins and competitors from AMD and Nvidia in a full-size ATX chassis. AMD may have a different market in mind for this GPU, but there’s no reason it can’t be used in a full-size chassis — and users interested in purchasing one may want to know its performance characteristics in different thermal environments.

A tiny GPU with a steep hill to climb

The Nano is the smallest high-end GPU we’ve ever tested and the card’s cooler and heatsink are clearly high-class, but the price gap between the R9 Nano and its closest competitor is nothing to sneeze at. The GTX 970 Mini is a $355 GPU, going up against a $649 Nano. That’s a high bar to clear, and if you’re looking to the Nano to justify it strictly on performance grounds, you’re going to be disappointed. The Nano is more complicated than that, for reasons we’re going to be discussing.

Like the R9 Nano, the Asus GTX 970 Mini is a small GPU at 6.7-inches long with a double-wide PCIe slot, two DVI ports, HDMI, and a DisplayPort. Asus clocks the card at 1088MHz base and 1228MHz boost — a hair above Nvidia’s stock 1050MHz / 1178MHz configuration.

Test and Nano configuration

We tested the Nano in our Core i7-5960X testbed (Haswell-E) with 16GB of DDR4-2667 and Windows 8.1 with all patches and updates installed. While we intend to update to Windows 10 in the near future, our previous testing was all done with Windows 8.1, and we chose to maintain that continuity rather than upgrading mid-cycle.

We’ve made one change to our test suite and replaced the older Metro Last Light , which debuted in 2013, with the upgraded Metro Last Light Redux , which shipped a year later. All of our AMD and Nvidia cards have been retested in the new title, and our power consumption figures were recalculated using the newer version of the game as well.

When we decided to test the R9 Nano in both a mini-ITX box and standard chassis, we decided to make one additional modification to the GPU’s default configuration. Instead of leaving the GPU at its default power settings, we used Catalyst Control Center to tell the GPU to use up to 50% more power than normal. This was done in order to measure the performance delta between Nano and the Fury X in a more conventional tower as opposed to a thermally constrained mini-ITX system.

In order to present at least some additional information on how Nano performs in a standard configuration, we backtracked and tested the R9 Nano at its default power level in Metro Last Light Redux. While this is just one data point at present, using that game let us make an apples-to-apples comparison with all the other GPUs we evaluated.

We attempted to use third-party drivers to modify the GTX 970’s maximum power consumption, but the Asus GTX 970 Mini doesn’t appear to support this feature in Asus’ own GPU Tweak or MSI Afterburner. Nvidia Inspector reported that it could change the GPU to 110% of standard, but that utility hasn’t been updated in more than a year and we saw no performance differences between 100% and 110%. Either the utility no longer functions or the GTX 970 Mini doesn’t need it.

1 of 4