Last week I published some fresh AMD Linux 4.9 + Mesa 13.1-dev benchmarks on many different AMD Radeon GPUs going all the way back to the Radeon HD 4800 series days. Today those numbers are being complemented by an extensive NVIDIA GeForce Fermi / Kepler / Maxwell / Pascal comparison to make up a 31-way NVIDIA/AMD Linux OpenGL performance comparison. If you are curious how the NVIDIA and AMD Linux performance is with the very latest drivers and going back several hardware generations, this holiday article is for you.
As already mentioned, the testing on the AMD side was with their fully open-source driver stack and using the newest code (Linux 4.9 + Mesa 13.1-dev) where this mainline code continues to support all AMD GPUs, even going back to the RV770 days compared to AMDGPU-PRO now just supporting GCN hardware. Overall the open-source AMD Linux driver stack advanced a heck of a lot this year as covered in the many Phoronix articles. The AMD cards tested based upon what I had available and limiting it to the HD 4890 and newer were: the Radeon HD 4890, HD 5830, HD 6870, HD 6950, HD 7750, HD 7950, R7 260X, R9 270X, R9 285, R7 370, RX 460, RX 470, RX 480, and R9 Fury. On the FirePro side I was also able to test the V8750 and V8800 as additional data points but unfortunately have no newer FirePro / Radeon Pro hardware.
On the NVIDIA side, of course, I was using the newest NVIDIA 375.26 driver released last week. Tested GeForce GPUs were the GeForce GTX 460, GTX 650, GTX 680, GTX 760, GTX 780 Ti, GTX 950, GTX 960, GTX 970, GTX 980, GTX 980 Ti, GTX 1050, GTX 1050 Ti, GTX 1060, GTX 1070, and GTX 1080. A nearly complete Maxwell and Pascal coverage of the consumer cards plus some interesting Fermi and Kepler cards for reference.
In this article we're just looking at the raw OpenGL Linux performance while in an article coming out this weekend is taking all of these cards and looking at their performance-per-Watt and overall power efficiency and some other interesting metrics. All of these Linux graphics benchmarks are carried out in a fully-automated and reproducible manner using the open-source Phoronix Test Suite benchmarking software.