Folding@Home Performance Is Looking Good On The GeForce RTX 2080 Ti
Folding@Home was accidentally left out of yesterday's RTX 2080 Ti CUDA/OpenCL comparison with simply forgetting to add the FAHBench test profile to the run queue. But as there is often interest in seeing the FAHBench performance on new GPUs by at least a few of the premium enablers, I ran some extra tests just looking at the Folding@Home performance and here are those results today.
With the same Ubuntu 18.04 system and using the NVIDIA 410.57 driver, the GeForce GTX 680, GTX 780 Ti, GTX 980 Ti, GTX 1070, GTX 1080, GTX 1080 Ti, and RTX 2080 Ti graphics cards were tested.
The Phoronix Test Suite was also monitoring the GPU core temperature and overall AC system power consumption during this Folding@Home performance benchmarking on these NVIDIA GPUs under Linux.
Like we saw with many of the other OpenCL and CUDA GPGPU tests, the RTX 2080 Ti is a big leap ahead of the GTX 1080 Ti in Folding@Home... The RTX 2080 Ti was 47% faster than the GTX 1080 Ti! Or if going from the GTX 980 Ti to RTX 2080 Ti, there is 2.53x the performance.
This GeForce RTX 2080 Ti Founder's Edition card had an average GPU core temperature under FAHBench at 54 degrees and a peak of 67 degrees, a few degrees lower than the GeForce GTX 1080 Ti Founder's Edition.
The AC system power consumption when running FAHBench was 214 Watts on average with a peak of 319 Watts, compared to 196 Watts with the GTX 1080 Ti and a peak of 260 Watts.
While the RTX 2080 Ti was consuming more power, this Turing graphics card ended up delivering about 35% superior performance-per-Watt over the GTX 1080 Ti.
As always, if you have any other ideas for any scripted/automated-friendly GPU benchmarks to add to the Phoronix Test Suite and OpenBenchmarking.org, feel free to let me know (ideally also including links to said programs / test scripts). Thanks to our test automation framework, if you want to see how your own system's Folding@Home GPU performance compares, simply run phoronix-test-suite benchmark 1809219-RA-FAHBENCHR84 for your own fully-automated, side-by-side performance comparison.