NVIDIA GeForce GTX 980 Dominates With OpenCL On Linux
Earlier this week I published the Linux review of the NVIDIA GeForce GTX 980 high-end Maxwell graphics card. While there were several OpenCL benchmarks included in that review, most of the tests focused upon the Linux OpenGL performance. For those that requested more GPGPU benchmarks, in this article are many Linux OpenCL compute benchmarks of the GTX 980 and other high-end AMD and NVIDIA GPUs from recent generations.
The GeForce GTX 980 dominates as well under OpenCL as it does with OpenGL. While the GTX 980 has just 2048 CUDA cores compared to 2,880 on the GTX 780 Ti, the Maxwell architecture is a heck of a lot more efficient and powerful than Kepler. My earlier GTX 980 Linux review is worth reading if you're not yet up to speed with NVIDIA's new GeForce GTX 900 series as the new graphics cards are real winners for Linux users as long as you don't mind obliging to using proprietary hardware drivers on your systems.
The assortment of AMD and NVIDIA graphics cards tested under Linux for this OpenCL comparison included the following:
NVIDIA GeForce GTX 680 2048MB (1006/3004MHz)
eVGA NVIDIA GeForce GTX 750 1024MB (1019/2505MHz)
NVIDIA GeForce GTX 750 Ti 2048MB (1019/2700MHz)
NVIDIA GeForce GTX TITAN 6144MB (836/3004MHz)
NVIDIA GeForce GTX 780 Ti 3072MB (875/3500MHz)
NVIDIA GeForce GTX 980 4096MB (1126/3505MHz)
Sapphire AMD Radeon HD 6950 2048MB (800/1250MHz)
ASUS AMD Radeon HD 7850 1024MB (860/1200MHz)
XFX AMD Radeon HD 7950 3072MB (900/1375MHz)
Gigabyte AMD Radeon R9 270X 2048MB (1100/1400MHz)
XFX AMD Radeon R9 290 4096MB (947/1250MHz)
The NVIDIA 343.22 binary driver was used with the GeForce hardware while the AMD Catalyst fglrx 14.40.1 / OpenGL 4.3.13179 Linux driver was used for the Radeon hardware. The Intel Core i7 5960X Haswell-E system was called into action with its Ubuntu 14.04 LTS + Linux 3.15 kernel installation for doing this compute-focused article.