• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Inconsistent GPU folding

W

wurmy

Limp Gawd
Joined
Dec 8, 2004
Messages
383
I have a dual SLI2 mobo (A75A-G35) with an Llano A2 (dual0core) CPU doing GPU-only folding.

The two cards are "identical" EVGA 650Tis OC'd to 1097MHz.

Right now, the rig is on an older Rosewill 650w (I think) "green unit" new out of the box and I have it on a kill-a-watt to monitor performance. On 807x units, it uses about 183w, and on 762x units the wattage goes up to about 193w, about 39-41K ppd on average.

Strange thing is that sometimes I check my kill-a-watt and when it's changed from 180-190 down to 150w, I check my TPFs and one my GPU slots will have a high TPF even with a similar WU. Task manager performance doesn't show anything pegged and the only way I've found to fix it is to reboot whenever I catch it. When I've caught this so far, the WUs were not in the initial states of execution. I'm running the non-beta v7 client with the 2.25 core.

I've re-flashed the BIOS to the latest version, used a recemt Rosewill Stallion 670w PSU, changed the RAM in single-channel mode, used the latest nVidia drivers, and yet sometimes the box just isn't folding the way it should. I'm keeping the kill-a-watt on there until I figure this out, and it's kind of annoying. I'm could run single-GPU "nodes" at about 110w per GPU consistently, and I have a Bulldozer that will run two GPUs at about 220w consistently.

Any ideas? The mobo only cost me $40 + shipping and I'm starting to wonder if it's a dud.
 
What's the stock clocks on those 650Ti boards?
I have an OC'd 560Ti that when it get's too hot, will drop the clock in half and not recover until rebooted.
Even so, the GPU monitor program can still falsely "show" the clock as being full speed.
It's usually obvious because that card will suddenly be very cool compared to normal.
 
I assume you're not running in SLI mode? I've seen posts where folks have had to disable SLI for folding performance. How about the power state in the NVidia driver? I think the new 6xx series cards have more aggressive power saving features, try setting it to 'prefer maximum performance' in the NV control panel - Manage 3D/Global Settings tab . BTW decent ppd on those 650Ti's, I was thinking of picking up one myself.
 
Thanks guys! The stock clock is 927Mhz. I've run different cards in different rigs and it seems like 1097MHz is the sweet spot. The cheap motherboard actually has good spacing between the GPUs, so has better cooling than on my bulldozer rig, and the cards can tolerate more heat than they're currently running at.

The cards don't support SLI. I changed the NV setting from "adaptive" to "maximum performance. The next time it happens, I'll check the power state of the GPUs and take it from there.

It doesn't happen every day though, but before I order another of the same mobo, I just want the rig to work more consistently.
 
Pretty large OC for F@H I would say. I can always get my cards "stable" at good OCs for a few days across many WU types.
Then an odd 76xx series WU will come along and boink it up.
I finally backed off to a small under-volt + reasonable OC that keeps the power/temps down.
Now the cards are stable for months at a time.
I find that highly reliable, slightly lower PPD nets me more points in the long run than pushing for maximum instantaneous PPD with unpredictable crashes.
Just my experience of course.
 
Core32 said:
Pretty large OC for F@H I would say. I can always get my cards "stable" at good OCs for a few days across many WU types.
Then an odd 76xx series WU will come along and boink it up.
I finally backed off to a small under-volt + reasonable OC that keeps the power/temps down.
Now the cards are stable for months at a time.
I find that highly reliable, slightly lower PPD nets me more points in the long run than pushing for maximum instantaneous PPD with unpredictable crashes.
Just my experience of course.
Click to expand...

I second this. Core is bang-on in his observation and advice. My GTX 560Ti Twin Frozr will not tolerate OC's much higher than 900 Mhz for folding - it'll fold fine for a week then crash the driver and folding grinds to a halt. It'll run games just fine at 950 MHz+ but not folding.
 
It might have been the performance setting being at "adaptive" but since I first posted, it hasn't happened again. I'm keeping the kill-a-watt on the rig for a week or so.

I've ran the same model of cards consistently for weeks at the same OC on different rigs ranging from a Sempron, Athlon II X4, Phenom II x4, FX 8150, Llano A6, and a Llano A2 (which is hosting the two GPUs right now).

I guess I need to wait a few days to see if the NV control panel setting worked or not. Next time it happens, I'll do a better job of checking the folding log and event viewer to get more hints.
 
You must log in or register to reply here.
Back
Top