For those more interested in Linux GPU performance for CUDA/OpenCL GPGPU computing than Linux gaming, this article is for you with a fresh round of results across my available GeForce Kepler/Maxwell/Pascal cards using the latest NVIDIA 375.10 binary driver paired with CUDA 8.0 on Ubuntu 16.04 LTS Linux.
I started off yesterday by delivering some fresh Caffe NVIDIA deep learning benchmark results while in this article is a wide-range of OpenCL and CUDA performance data for the 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. Unfortunately I have no access to any Quadro/Tesla accelerators for testing thus just benchmarking with these consumer cards.
The 14 NVIDIA GPUs were tested with a wide array of OpenCL and CUDA 8 benchmarks using the open-source, automated Phoronix Test Suite benchmarking software. In addition to looking at the raw performance data, the AC power consumption was monitored via a WattsUp Pro to also offer performance-per-Watt metrics where applicable. Continue on to see these latest benchmark results.