AMD Radeon R9 Fury Review: Sapphire Tri-X Overclocked

How We Test

Test System

Tom’s Hardware standardized the benchmarking test system across all of our various locations and reviewers for 2015. The idea being that we can compare the tests that everyone does and rule out any discrepancies from using different components. The PC we all use has an Intel Core i7-5930K, 16GB of DDR4, two 500GB SSDs from Crucial and an MSI X99S XPower motherboard. We also use an 850W Platinum-rated power supply from be quiet!, which gives us the headroom to test even the most power-hungry graphics cards.

Software & Drivers

We’ll be making use of the results recently garnered from our GTX 980 Ti and R9 Fury X evaluations. Some of the tests use different drivers. For instance, Nvidia's 352.90 build was used for all tests on the GTX 980 Ti. It was also used for all Nvidia GPUs while testing Grand Theft Auto V. All other games were tested with GeForce driver 347.84 for the Titan X, 980 and 780 Ti.

AMD’s Radeon R9 Fury X and Fury were both tested using Catalyst 15.15 Beta, since it was the only driver with support for Fiji when testing was conducted. Catalyst 15.5 Beta was used to test the 290X, as 15.15 does not support it.

As we noted earlier, though, two days before the Fury release, AMD published Catalyst 15.7. Igor managed to test GTA V, Shadow of Mordor, Battlefield 4, Metro Last Light and Thief. In all of his spot checks, there were no appreciable performance differences. The new driver primarily focuses on features like DX12 and Windows 10 support.

In addition to running all of the tests using Sapphire's stock settings, the Fury Tri-X was tested at AMD’s reference specifications, which gives us a nice baseline for comparing other reference-class Fury cards (like the other model Sapphire intends to launch). Sapphire pointed out that initial stock for these cards will be limited, while its lower-clocked model will be more obtainable.

Benchmarks

Noise

As usual, we’re using a high-quality measurement microphone that is positioned perpendicular to the center of the graphics card in question at a distance of 50cm for our noise measurements. Smaart v.7 is used for the analysis. The ambient noise at night during our measurements never exceeded 26dB(A) and was recorded and accounted for during each individual noise measurement.

Power Test System