Strange PPD

R

relic

[H]ard|DCer of the Month - August 2007
Joined
Mar 30, 2001
Messages
9,318
I am running two nearly identical systems.

GPU2 client:
One with a 5008 is getting 4750 PPD
the other with a 5008 is getting 2575 PPD

These systems have been within 100PPD of each other for 3 weeks.

WTF?
 
since yesterday i have been getting strange results on my new 8800GT box, it seems like every other WU it fluctuates 1000PPD, it started at 4700 and some of the WU's are 3600, then it will switch back, i don't know what is going on
 
it´s your folding mojo you are loosing it, to replenish it you´ll have to buy GTX280s for everyone ;)
 
I would check both clock speeds and temps.
Also check CPU & kernel loading.
The third thing to check is how many cores each GPU client is folding on or are they locking to one core.
Can you see any difference between the boxen bar the PpD.

If you dont find anything wrong, then stop watching **** on the slow computer ........... :p

Luck .................. :D
 
I bet it's a core affinity issue. There is a way to tweak the affinity to avoid core 0 and 3 (cores windows love to pick for foreground and background tasks).

 
Wish it was something that easy, but no.
Both boxes have identical WU and configurations and have been identical in PPD for several weeks


After reviewing both systems there are no changes.

The sytem that is getting the higher PPD is actually the one I have been using for web browsing, the other one is purely folding.

No changes, same hardware, same OS. Changed today for no apparent reason.
 
My major points fluxes come when I get a 2665 and 2653, Points jump to about 4k, from the normal 3k for dual 2665's on my quads... Do that time 9 quads and things can get wonky... I've peaked at 55k and hit a low of 49k..

 
OK, I just let the strange 5008 WU run and the PPD is back up after it was turned in.
No changes at all on the system.

I checked the logs and the steps for that particular WU were almost double the time of a previous 5008 on the same machine, later 5008s are running at normal speed.

Looks like a wonky WU, or a potential "re-benched" one got out in the wild. ;)
 
relic, I've been having the same issues as well, it's been crazy how much my GPU2 clients PPD fluctuates. I don't know if some WU's have harder parts then others, or if things are just funky as they are still trying to set everything up, I just don't know what it might be.
 
This is me guessing but I would not be supprised if thoes same flaky WU preformed beter on ATI cards. Maybe something is in the works :confused:
 
Kendrak said:
This is me guessing but I would not be supprised if thoes same flaky WU preformed beter on ATI cards. Maybe something is in the works :confused:
Click to expand...

Possible, that was my original thought.
But who knows?
 
I've just rebooted one of my boxen that was folding slow.
The GPU2 client was only doing ~4,000 PpD vs 5,500 PpD when running full speed.
Also the VM's where running slow at ~1,600 PpD vs 2,000 PpD.
One thing I did notice the cpu loading was down and the kernel loading was up.
Normally the CPU loading is at 99-100%, when running slow it was only ~95-97%.
The kernal loading on the core which runs the GPU client is normaly only around 10%, when running slow it was around 40%.
All loading look back to normal after the reboot.

I've also upped the FSB up a couple of hertz.
I've now got my 5450's at 3.45 Ghz.

Luck ........... :D
 
Always good advice, Tigerbiten.
I didn't bounce that box since it was specifig to one WU, but whenever I see a trend downward in PPD on a folding box I usually reboot it.

More often than not, I had left it running for a month and the OS was spazzing.
 
jbzx86 said:
GPU clients for Linux would be nice
Click to expand...

So would free booze, not going to happen. I would like this to, it save me 70 bucks per folding box. I can normaly find a deal on Vista, but $70 is a moherboard or 1/2 a GPU.
 
You must log in or register to reply here.
Back
Top