Folding Forum

here's my rig stats:
ati 3870 vid card max clocked in catalyst
asus p5e-vm hdmi board
xeon x3350 quad cpu clocked to 3.4 ghz
4 gigs memory
raptor hd
running vista ultimate 64-bit
all running on custom water

normally i run 2 smp clients on this rig and make roughly 3400-3500 ppd, but i wanted to try the new 6.20 gpu client out. so i dropped the second smp client and started running the gpu client and i'm only getting about 2900 ppd now. is it because the single 3870 just isnt as good as 2 cores running an smp client? or am i doing something wrong? i set the core priority of the gpu client to low so as not to disturb the smp client (it wouldnt run without errors until i did this). as a side note, when i look in catalyst to see my card temps, it shows that the gpu is only loaded at about 29-37%. any help would be greatly appreciated guys. thanx.

when i use catalyst control center to check the temp of my 3870 card while folding with the new 6.20 console client, why does it show that the gpu load is always less than 40%? does this sound wrong or is this the norm? thanx guys.


I guessing this might all be related, so I merged the two threads. -7im

yes they are, just thought they'd fit in separate topics. thanx!!

2900 isn't right at all for this hardware. You should be seeing 4000+. Make sure you have set your GPU client to run at "Slightly Higher Priority" in the configuration. It sounds like it isn't being fed properly.

i do that by running the -configonly flag from the administrator comand prompt and changing the advanced options correct?

For Some reason, the SMP client and GPU2 don't play nice in Vista.

Your GPU2 client is stalling your SMP.
One core is trying to run two clients ... which henders preformance.

I would suggest running a VMServer on your system.
That will load Core 0 & 1 with SMP ... then run GPU2 on Core 3 ... and a core to spare .... to run what ever else you want (Standard Client)

Running VMServer is very easy ...... esp if you run "notfred" distro in there.
I did this and my SMP points went from 3000 -> to 5000 PPD .... well worth the effort.

The other way is to run 3 x Standard Clients and 1 x GPU2 to max the quad core.