![]() There isn't much more to add besides: enjoy and be sure to run your own comparison tests if you're interested in Linux GPGPU performance. Starting to use Phonornix 3.0 (beta 1) is done in 4 easy steps: Download the latest beta-version from Extract. The tests are based on David Bucciarelli ‘s OpenCL demos. With that said, on the following pages are all of these CUDA and OpenCL performance figures. On Ubuntu Phoronix version 2.8 is in the Ubuntu app store (Synaptic), but 3.0 has those nice OpenCL-tests. At this time I don't have any plans of running a similar AMD OpenCL round-up considering I have far fewer Radeon graphics cards than all of these NVIDIA samples due to AMD's lack of interest in seeing Linux GPU reviews and thus needing to buy my AMD graphics cards retail. Obviously if you don't have CUDA setup or are using a non-NVIDIA GPU, only the OpenCL sub-set of the tests will execute. After installing the Phoronix Test Suite on the Linux distribution of your choice, simply run phoronix-test-suite benchmark 1511113-PTS-GPGPUNVI62 to conduct your own, fully-automated, side-by-side performance comparison against all of the results in this article. OpenCL is rarely used for machine learning. With these fresh numbers being put out here mainly for reference purposes, you're encouraged to see how the performance of your own Linux GPGPU setup compares to the results in this article. A study that directly compared CUDA programs with OpenCL on NVIDIA GPUs showed that CUDA was 30 faster than OpenCL. All of these tests were run in a fully-automated and reproducible manner using the open-source Phoronix Test Suite benchmarking software. The GPGPU benchmarks ran for our latest Linux OpenCL comparison included a mix of our old and new compute tests: SHOC (Scalable HeterOgeneous Computing suite), ASKAP tConvolveCuda, CUDA Mini-Nbody, JuliaGPU, MandelbulbGPU, and LuxMark. Tests were done from Ubuntu 14.04.3 LTS on Linux 3.19 with CUDA 7.5 Tool-Kit and the NVIDIA 352.39 binary driver. ![]() The cards included the entire range of GeForce GTX 900 "Maxwell" graphics cards as the latest-generation solutions along with several older GeForce GTX 600 and GTX 700 Kepler graphics cards. Mostly for reference purposes, I ran a variety of new and old OpenCL and CUDA open-source Linux benchmarks on ten different NVIDIA graphics cards. With having just added some new OpenCL/CUDA benchmarks to the Phoronix Test Suite and, I took this opportunity to run a variety of OpenCL/CUDA GPGPU tests on a wide-range of NVIDIA GeForce graphics cards. ![]()
0 Comments
Leave a Reply. |