Which core does the GPU client try to use?

2bullish

2bullish

2bullish

Associate
Joined
21 Jan 2005
Posts
762
Location
Manchester
If I have a dual core system am I right in thinking that the gpu client will try to access core1 or does it not matter? The reason I ask is that I want to stop folding on one of the cores so that I can run the gpu client and was waiting for core 1 to finish but then went out and its just downloaded a 600 pointer but core2 is finished in another couple of hours, so could switch that off if it doesnt matter which core the gpu client uses. Make sense?
 
CPU client 1 = Machine ID 1
CPU client 2 = Machine ID 2
GPU client = Machine ID 3

It doesn't matter which CPU client you kill, the Machine IDs are still different so the GPU client will take whichever core isn't being used.

SiriusB
 
If you have dual core and run one GPU client and one CPU client you'll find that you have:

F@H instance 1 CPU using 25% of core 1 and core 2 (50%)
and
F@H instance 2 GPU using 25% of core 1 and core 2 (50%) and 100% of the GPU.

I think.
 
Well theoretically it would be good if you could assign each fah process an affinity, thereby meaning Windows wouldn't have to manage the two processes. However it's not possible at the moment you just have to let Windows manage it.

I don't actually know that much as I don't even have a motherboard that has a PCI-express slot.

However, I think, from what people have said, that running one CPU client at 100% and one GPU client is best assuming you have dual core. There is of course the fact that if you have a nice clocked Conroe you may well achieve higher PPD with two CPU clients depending on the WU's around at the moment.

I think that basically you need to have one free physical CPU for ever GPU client you want to run.

The main thing with the GPU client is that it can churn out WU's like nobody's business compared to a CPU, which from a science point of view is fantastic.
 
oceaness said:
Well theoretically it would be good if you could assign each fah process an affinity, thereby meaning Windows wouldn't have to manage the two processes. However it's not possible at the moment you just have to let Windows manage it.

I don't actually know that much as I don't even have a motherboard that has a PCI-express slot.

However, I think, from what people have said, that running one CPU client at 100% and one GPU client is best assuming you have dual core. There is of course the fact that if you have a nice clocked Conroe you may well achieve higher PPD with two CPU clients depending on the WU's around at the moment.

I think that basically you need to have one free physical CPU for ever GPU client you want to run.

The main thing with the GPU client is that it can churn out WU's like nobody's business compared to a CPU, which from a science point of view is fantastic.
Click to expand...

can't you set the core affinity in task manager?
 
You must log in or register to reply here.
Back
Top Bottom