CPU question.

Denithor

Diamond Member
Apr 11, 2004
6,300
23
81
Ok, I have three processors available to use for a folding box, I need help choosing which one is the best for this purpose.

1) Pentium 4 2.8GHZ (800FSB/512K cache)
2) Celeron D 2.93GHz (533FSB/256K cache)
3) Athalon 64 3000+ 1.8GHz

This box will also contain a Radeon X1900GT video card and will be running the gpu client. I need to know which cpu will push this video card the most to get the best results out of the system. The rest of the system is basically the same, 512MB DDR400 RAM and whatever motherboard I can scrounge that is suitable.

Thanks!
 

GLeeM

Elite Member
Apr 2, 2004
7,199
128
106
I have no experience with the GPU client, but I have read at the F@H forum that the GPU client does not need much of a processor to feed it data. I guess I would pick the combo that will use the least electricity and run the coolest, as any of those processors should be more than enough for the GPU client.
 

Denithor

Diamond Member
Apr 11, 2004
6,300
23
81
Well, then, as a follow-up question, which one would be the fastest alone for folding? I could always set up two boxes I suppose, one with GPU and one with the fastest cpu.
 

Denithor

Diamond Member
Apr 11, 2004
6,300
23
81
Yes, this is a Northwood P4 with Hyperthreading. Should this feature be enabled or disabled?
 

BlackMountainCow

Diamond Member
May 28, 2003
5,759
0
0
Now this is a guess into the blue: If the GPU client needs only a few CPU cylces, maybe you can run a GPU and a full CPU client on the P4 HT. I know, the P4 HT only has one physical core, but two logical ones. I guess you'd have to test this theory, but I'd give it a shot: Run the GPU and another CPU client on the HT and see what times you get per frame.
 

Alyx

Golden Member
Apr 28, 2007
1,181
0
0
The P4 is probably your best bet. The GPU client actually uses quite a few cycles, on my Core 2 Duo it uses about 25% of the processor. I guess it needs that much to constantly "poll" the GPU core to feed information and get results. Your best productivity is probably the P4 running both the GPU client as well as the single core client. The GPU client should finish about 1 WU a day, so roughly 330ppd and the CPU client I imagine will take around a day and a half to two days per WU pending on size so roughly 250ppd there. All told a 400-600ppd average system, not half bad!

Once you've set it up you may need to have the CPU client run only small work units, between both cores running it might run out of ram if you are using it for internet or anything.