Optimising output

Associate
Joined
19 Jan 2007
Posts
2,066
Location
Northampton
Hello chaps

I've now got my E6700 running 2 GPU clients and 1 CPU client, I have noticed that one core of the CPU reads a consistant temperature and the other fluctuates, my guess is the requirements of the GFX cards are not a steady load.

Is there any way of setting a second CPU client up that will take a lower priority than the GFX cards so it will use the little gaps to crunch, even if a WU took a month it's extra points!!

I had a go, but I didn't seem to get it right, the CPU client ate all the resources and the GFX cards sat idle.

Ideas please.
 
It's been a while since I GPU folded, but I think it's still the general consensus that no more than one CPU client should be run alongside GPU clients. You need a free CPU core to aid the GPU clients.
 
You should stop running the CPU client as all it will be doing is slowing down one or both the GPU clients which is bad for the science. Stanford need the results in as fast as possible and having two GPUs running at half speed is a waste.

When I GPU folded I did some experimenting with running a standard client on the same core as the GPU client and the results were never good. I got more PPD by running just the GPU client on one core and a standard client on the other.
 
Cheers chaps.:)

It was worth a punt though.:)

With 2 GPU clients running and 1 CPU client it seems fine, so I'll leave it that way. The GPU's are each knocking out a frame every 8 minutes and the CPU a frame every 10 to 15 minutes depending on WU.

fah.jpg


If that sounds OK I shall leave it at that.
 
Last edited:
That sounds better than I would have expected given that neither GPU is getting a full core. Either the C2Ds are able to give enough time to the GPUs and have some spare or the client has been updated to not necessarily require 100% of a core/cpu.
 
Mine are quite old now, the new R600 series might have more folding power, we'll find out soon.:)
 
Back
Top Bottom