Test Methods
Benchmark testing where we use a scripted or specific benchmark utility that has no user interaction other than setting the graphics properties and hitting ‘start’. The utility then gives us a report of the relevant statistics.
We also only graph the average frame rates as minimum and maximum frame rates can be misleading.
The following tests were all reproducible or scripted or in-game benchmarks that are consistently executed without human intervention.
Futuremark Firestrike
We’ve moved on from the older 3DMark11 benchmarks that still hold relevance for some but the latest benchmarking tests from Futuremark give modern day GPUs a serious belting at the extreme, high end and mid-low end levels.
In these synthetic benchmarks, we see the AMD Radeon R9 Nano consistently taking the points with the Firestrike series.
Firestrike Ultra
Firestrike Ultra is first up and gives a good indication of how ‘4K-ready’ a GPU is.
Firestrike Extreme
Next up Firestrike Extreme is geared to test 2560×1440 gaming capability
Firestrike
Firestrike gives an indication of performance at 1080p
Skydiver
Finally we tested SkyDiver in the interests of completeness but the result is aimed at low end graphics platforms.
Unigine Heaven
Unigine Heaven Benchmark 4.0 Basic is one of the programs that we run to temperature test the cards and we also do a benchmark run during the process.
Interestingly, in the Unigine Heaven benchmark, we see the Nano performing a little better than the GTX970 but not as well as the GTX980. A score of 1824 in Unigine Heaven Extreme is still an excellent result for a card this size.
Tomb Raider
For the in-game benchmarking utility, the following graphics settings were used:
1920×1080 & 2560×1440, Motion Blur:On, Screen Effects:On, Quality: Ultimate,Texture Quality: Ultra,Texture Filter: Anisotropic 16X, Hair Quality: TRESSFX, Anti-Aliasing: FXAA Shadows: Normal, Shadow Resolution: High, Level of Detail: Ultra, Reflections: High, Depth of Field: Ultra, SSAO: Ultra, Post Processing: On, Tessellation: On, High Precision: On
Although TRESSFX is an AMD feature we had it enabled because gamers will want to enjoy it in this title and we have other benchmarks so it all evens out in the end.
Despite Tomb Raider being considered an ‘AMD friendly’ title, the full sized GTX 980 takes the points in the benchmark. The margin is 6FPS at 1080p but narrows down to only 1.5FPS at 1440p.d
Batman Arkham City
The settings used for the Arkham City in-game benchmark were:
1920×1080 & 2560×1440, V-Sync: Off, Anti-Aliasing: FXAA (High), Directx 11 Features: MVSS and HBAO, Dx11 Tessellation: Normal, Detail Level: Very High Dynamic Shadows: Yes, Motion Blur: Yes, Distortion: Yes, Lens Flares: Yes, Light Shafts: Yes, Reflections: Yes, Ambient Occlusion: Yes, Hardware Accelerated PhysX: Normal
We ran this one 3 times and got the same results in each run at the standard test settings. It turns out that it’s the NVIDIA PhysX hardware acceleration setting as expected. If this is set to “Off”, at 1920×1080 we see a minimum of 42FPS, Maximum of 231FPS and an average of 155FPS. At 2560×1440 with NVIDIA PhysX disabled, we saw a minimum of 49FPS, a maximum of 169FPS and an average of 124FPS.
The benchmark settings need to be consistent so the original scores stand. We have hardware PhysX set to Normal for Batman Arkham City and TressFX enabled in Tomb Raider to balance the developer bias out.
Hitman Absolution
The following settings were used in the in-game benchmark:
1920×1080 & 2560×1440, MSAA:8x, Vsync: Off, Quality Setting: Ultra
The Nano pulled ahead of the pack here but more so in the 1440p benchmark where the margin is 7FPS better than the GTX 980. Seeing almost double the performance of the R9 380 4G in a card half the size is a little surreal.
Metro 2033
The benchmarking script for Metro 2033 was set using the following options:
1920×1080 & 2560×1440, DX11, Quality: Very High, AA: MSAA 4X, Texture Filtering: AF 16X, Advanced PhysX: Disabled, Tessellation: Enabled, DOF: Enabled
The GTX 980 and 970 both show up the Nano in the Metro 2033 benchmark but the difference is significantly less at 2560×1440.
Bioshock Infinite
The in-game benchmark for Bioshock Infinite was set using the following options:
1920×1080 & 2560×1440, DX11, All settings maxed out
In Bioshock Infinite, we can see that the Nano actually sits in between the GTX 970 and GTX 980 performance-wise.
Sniper Elite III
The in-game benchmark for Sniper Elite III was set using the following options:
1920×1080 & 2560×1440, All settings maxed out
Once again, the R9 Nano splits the difference between a GTX 970 and a GTX 980 at 1080p but comes out on top in the 1440p benchmark by a mere 1.1FPS.
GTA V
The in-game benchmark for GTA V was run using the following options:
1920×1080 & 2560×1440, DX11, Tessellation: 3, ShadowQuality: 3, ReflectionQuality: 2, Reflection MSAA: 2, SSAO: 2, Anisotropic Filtering: 16x, MSAA: 2x, TextureQuality: 2, ParticleQuality: 2, WaterQuality: 2, GrassQuality: 2, ShaderQuality: 2, Soft Shadows: 1, UltraShadows_Enabled: false, Particle Shadows: on, Shadow Distance: 1, LongShadows: off, Reflection_MipBlur: on, FXAA Enabled, TXAA Disabled, Lighting_FogVolumes: true, Shader_SSA: true, PostFX: 2, DoF: off, MotionBlurStrength: off
I’m happy to call this a tie with the GTX980 at 1440p but at 1080p, the GTX 980 takes the points by almost 6FPS.
What we have seen in all of the benchmarks is a common theme that the AMD Radeon R9 Nano copes better in relative terms at 1440p resolution.
