- Joined
- May 18, 1997
- Messages
- 58,175
No Folding for Fermi yet. Got excited over nothing.
Here is the 411.
I looked on NV's internal Fermi site this morning, but could not get the CUDA folder to DL.
I was hoping it would run on GPU2. Here is what NV has to say now. So yeah, I thought it MIGHT work for me. I will keep trying to get what I need from NV but no public workload downloads right now. Damn.
FAQ with the PR stuff.
Full 2560 rez shot.
Here is the 411.
I looked on NV's internal Fermi site this morning, but could not get the CUDA folder to DL.
Folding@home GF100/GT200 ‐ Compatible Preview Test Application
What it is
NVIDIA and Folding@home have worked together to create a test application that is useful for measuring the relative speed between NVIDIA GeForce GTX 470 / 480 and NVIDIA GeForce GTX 280 graphics cards running the Folding@home algorithms. This is being made available to press only, and we request that you do not re‐distribute it.
Because the Folding@home software is evolving, a comparison of performance scaling with earlier versions of the Folding@home software is not valid. The test software runs on one of three included molecules and does not get work units from the web. Results are written to a log file. We are working to achieve even greater performance gains and further qualification of the software is ongoing. The Folding@home team plans to release a public build sometime in April.
FAQ with the PR stuff.
FAQ
Q: How do I interpret the results?
A: If you divide the ns/day for the GTX 480 measurement by the ns/day for the GTX285, you’ll get a
speedup multiplication factor – e.g. 1.7X faster. If you subtract one from that and multiply by 100,
you’ll get percent speedup – e.g. 70% faster.
Q: Why do the different molecule simulations produce different results?
A: The different simulations exercise different aspects of the software and hardware and consequently
take more or less advantage of the architectural advances of GF100 GPUs.
Q: How do the resulting nanosecond per day (ns / day) metrics compare to points per day that a user
might accrue using the real folding application when it’s deployed?
A: A GTX280 does about 7700 points per day (PPD), so the PPD for the GTX480 would be based on the
speed up. So, if the GTX 480 is doing 70% better, then it would get 1.7 x 7700 = 13,090 PPD. Here's a
useful url with PPD comparisons: http://forum.xcpus.com/xtreme‐folding‐home/12939‐gpu‐ppdcomparison.
html
Q: Can this FAH client work with Work Units downloaded online from the Folding@Home server?
A: No. This build does not automatically get workloads from the Folding@home server. The
Folding@home team will be deploying an official FAH client based on this preliminary build shortly.
Q: Does this special build mean that it’s difficult to get existing software to work with GF100 GPUs?
A: No. We were able to get the Folding@home algorithm to work with GF100 GPUs very quickly. There
are a few reasons that it’s taking longer to deploy an end‐user client for Folding@home
I. The Folding@home team is evolving their software architecture and wanted to deploy GF100
GPU support on their latest architecture
II. NVIDIA and the Folding@home team wanted to take this opportunity to further optimize the
performance of the NVIDIA GPU client software
III. NVIDIA and the Folding@home team are committed to doing sufficient QA to ensure troublefree
end‐user experiences with the software
Q: Does this special build work with other NVIDIA cards like the Geforce GTX 280?
A: Yes. This build was developed in conjunction with CUDA C and will work with older CUDA capable
GPUs.
Q: Does this build work with ATI cards?
A: No. This build was specifically developed to run on NVIDIA cards and will not run on an ATI.
Full 2560 rez shot.
Last edited:




