NVIDIA GeForce GTX 1070 On Linux: Testing With OpenGL, OpenCL, CUDA & Vulkan
Like the GeForce GTX 1080 Linux review, the GTX 1070 testing is primarily focused on OpenCL and OpenGL testing along with a Dota 2 Vulkan test. CUDA 8.0-RC1 + cuDNN was also tested on the NVIDIA hardware for this article if you are interested in NVIDIA's GPGPU API. First up are the raw performance results followed by some performance-per-Watt / power consumptiob results and GPU thermal testing at the end of this article.
For putting the GeForce GTX 1070 performance into perspective, I compared it on the green side to the GeForce GTX 680, GTX 760, GTX 770, GTX 780 Ti, GTX 950, GTX 960, GTX 970, GTX 980, GTX 980 Ti, GTX TITAN X, and GTX 1080. All of these NVIDIA cards were tested on the NVIDIA 367.17 driver. On the AMD side was the AMDGPU-PRO Beta 2 driver and tested there for some reference was the R9 285, R9 290, and R9 Fury. Not many cards were tested there due to what's supported by AMDGPU-PRO and my smaller collection of AMD GPUs plus as most Phoronix readers will know they currently don't provide much competition to NVIDIA under Linux.
All of the tests happened on the same system with an Intel Xeon E3-1280 v5 Skylake processor, MSI C236A Workstation motherboard, Samsung 950 PRO M.2 NVMe SSD, and 16GB of DDR4-2133MHz EUDIMM memory. Common on the software side was Ubuntu 16.04 LTS x86_64 with the Linux 4.4 kernel and Unity 7.4 desktop.
All of the benchmarks in this article were run in a fully-automated and reproducible manner using the open-source Phoronix Test Suite benchmarking software.