CPU + GPU bottleneck?

[Ryo129]

I've been folding on my Hd 4870 and getting about ~4200 ppd on it according to FAHmon. I just recently set up a smp client to fold with my p2 920 and my ppd on the GPU dropped to 3k? it could just be the WU but i wanna be sure if it is bottlenecking and so i can try and fix it.

Secondly anyone know of a way to hide the smp console window? i don't want to start it as a service and i've already shut the client down ~5 on accident

 

[stanhemi]

Download ''get and set cpu affinity'' lock 3 core for your smp and the last core for your gpu client.

 

[Ryo129]

hmm that program is crashing on startup =\ think it would be fine if i just manually set the affinity in the task manager?

 

[Kursah]

Let's start with this, what kind of PPD is being reported for the PII? I'm not sure the drop from one core, but add that to the 3k and compare to just the vid card, should be pretty damn solid. TBH, my card scores the same whether I am running SMP for my quad or not, I would assume that's how yours should be...though maybe the WU changed, that can have a drastic affect on reported PPD. I would say leave it alone and fold away tbh. It could be an ATI deal, but I would assume letting the quad run on full dice should be pretty solid, I average around 3.1kppd from my q6600 at 3.6ghz, I would assume that PII should do pretty well and be close, maybe even better than my chip. Though you can try changing affinity in task manager, it should work for ya. Keep us posted!

 

[Ryo129]

i'm getting close to 2k on my pII 920 (@stock 2.8) and with both i'm getting a total of ~5k ppd. I'm not quite sure on exactly how the affinity should be set so heres how i set it:
FAH6 (CPU smp) - cpu 1,2,3
FahCore_a1 (all the smp clilents) - cpu 1,2,3
FahCore_11 (GPU client) - cpu 0

FAHmon reports a total ppd of 4300 (where it was before i changed affinity)

EDIT: hmm i also installed that F@H SMP affinity changer how well does that work with a GPU client?

 

[Kursah]

The affinity changer is for getting more done with the CPU, though it might help you with the GPU client I'm not sure of that. I do know that it will take one SMP client and use it on cores one and 2 for all processes in that client, then you download and install a 2nd SMP client (say add a 1 to the install dir), and run that, it'll go on the 3rd and 4th cores. That's sort of how it works, it's also based on client load for which cores are to utilize it.

Really, now that I think about it, no point in running the SMP client on a quad unless you plan to run a 2nd SMP client at the same time, WU's will take longer, but PPD should increase at least on Intel cpu's, I'm unsure on the AMD end of it...I forgot about that thread and it's link in my sig tbh.

 

[Lionheart]

hmmmmmmmmmmm

 

[Ryo129]

Ooooh update: i re overclocked my pc for more ppd and all my numbers shot up. I'm now running at 3.3ghz core and 480mhz on my ram. GPU is averaging 3.2k ppd and smp client hits 2.4k ppd. All is good