Folding Forum

Is there a FAH hardware usage statistics?
I mean e.g. how many AMD Tahiti GPUs are used in % of all GPUs and how many PPD they produce in % of all GPUs?

As far as I know, there's nothing more detailed than this page : http://fah-web.stanford.edu/cgi-bin/mai ... pe=osstats

toTOW wrote:As far as I know, there's nothing more detailed than this page : http://fah-web.stanford.edu/cgi-bin/mai ... pe=osstats

I use another link and the table shows other information: http://fah-web.stanford.edu/cgi-bin/mai ... e=osstats2

artoar_11 wrote:

toTOW wrote:As far as I know, there's nothing more detailed than this page : http://fah-web.stanford.edu/cgi-bin/mai ... pe=osstats

I use another link and the table shows other information: http://fah-web.stanford.edu/cgi-bin/mai ... e=osstats2

osstats2 is linked from the Teams & Statistics. I would think it's the more accurate version.

I didn't know this one ... it just splits NV GPUs in two categories : NV GPU is probably G8x/G9x/G20x and NV Fermi GPU are probably Fermi and newer GPUs models.

What's a little concerning: Although both pages osstats and osstats2 were updated today, the numbers do not match.
e.g. ATI GPU in osstats show 3114 Active CPUs while osstats2 shows 6207.
I believe osstats2 is more accurate...

There is also another page osstats3 which shows the history of April 2014.
Compared to osstats2 of today we have lost 75% of the Windows Clients but Linux Clients have trippled.
The old Nvidia GPUs mostly disappeared but current AMD and Nvidia GPUs count stayed the same.

I interpret this that more server machines running Linux are folding while the Windows gamer power users leaved folding.

The osstats page was deprecated some time ago. Osstats2 is the current page and is based on the data collection changes adopted as part of the updated serverstats, project summary pages, and other related changes done then.

foldy wrote:Compared to osstats2 of today we have lost 75% of the Windows Clients but Linux Clients have tripled.

Hmmm.
Many GPU-centric Donors decide their CPU slot isn't worth it and at the very least, there are only (N-1) CPUs available which effectively shuts sown all Dual CPU or uniprocessor clients -- or moves them to Chrome.

I wonder how migrations from V7-CPU to Chrome/NaCl shows up in those statistics.

@bruce: That's it. I myself only fold on GPUs because of the much times higher performance.
So the statistics only count a Windows or Linux client when it folds on a CPU?
We cannot see if GPU only folders use Windows or Linux?

When the V6 client was popular, folks would run both a CPU client and a GPU client, counting at two clients. When V7 combined that same function into two slots, Stanford should still count that as two clients, and I suppose they do (running two different WUs from two different projects).

My earlier question was whether when the NaCl client was released, did those count too or just V7slots + V6 clients.

That stats page is do for another overhaul considering all the new platforms running now. But not a high priority.

And as noted, the FLOPs calc on that page was done with code running in 2013. One would hope the newer cores since then are more efficient.

As suggested on Reddit: if PG makes the information available in a compressed CSV file I'm sure there will be two or three in the community happy to help with the data mining. We are here to help .... With the low priority stuff ...

Publishing data hasn't typically been the issue. Taking time away from other higher priority issues to ensure proper collection of those stats from the current clients, and then coding how to collect from the new clients and new platforms is more the issue if I had to guess.