Advanced search

Message boards : Graphics cards (GPUs) : 65nm GTX260-216 issue.

Author Message
lkiller123
Send message
Joined: 24 Dec 09
Posts: 22
Credit: 15,875,809
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 18598 - Posted: 10 Sep 2010 | 1:32:15 UTC
Last modified: 10 Sep 2010 | 1:41:41 UTC

I've just purchased 2 55nm GTX260s and 1 65nm GTX260; they are solely for GPUGRID purposes.

Now the 55nm all crunch fine, they all blow their way through the WU without any problem. Now the weird thing is, on the WUs that the 65nm is crunching on always error out in Windows, saying something like "ACEMD2 6.05 has stopped working."

What is going on?

Profile liveonc
Avatar
Send message
Joined: 1 Jan 10
Posts: 292
Credit: 41,567,650
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwat
Message 18599 - Posted: 10 Sep 2010 | 1:48:40 UTC - in response to Message 18598.
Last modified: 10 Sep 2010 | 2:29:17 UTC

The 65nm was the old version of the GTX260 with 192SP or maybe the 65nm version with 216SP, which still uses more more & produces more heat. I've read they're troublesome. Try different driver versions until you find one that's working well. It might not even be possible, but you never know if you try hard enough & try to optimize airflow as much as possible to reduce heat as much as possible. Try downloading NVIDIA Control Panel & manually set your fan speed to 100%. Stability "may" also be gained by either downclocking or increasing voltage, but DON'T do this unless you're confident with what you know your GPU can take & understand the risk there still is, this will also produce more heat if you increase voltage.
____________

lkiller123
Send message
Joined: 24 Dec 09
Posts: 22
Credit: 15,875,809
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 18600 - Posted: 10 Sep 2010 | 2:43:32 UTC - in response to Message 18598.

It is a GTX260 core216 with a 65nm chip. I always keep the temps under 75 degrees and I have multiple fans blowing onto the cards.

Can you recommend me a driver that works best with GPUGRID?

Profile liveonc
Avatar
Send message
Joined: 1 Jan 10
Posts: 292
Credit: 41,567,650
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwat
Message 18601 - Posted: 10 Sep 2010 | 3:23:16 UTC - in response to Message 18600.
Last modified: 10 Sep 2010 | 3:36:07 UTC

I'd guess that the same one that works with a GTX280 would work well with a GTX260 core216 65nm. But I don't know who uses that anymore. Try the 197.45 driver & move downwards in slow, painful, trial & error in search for a stable version.

BTW did you mix your 65nm with your 55nm? I see you have two PC's or have you not disabled SLI? If you have two PC's try using the two 55nm in the same PC, but also be sure that SLI is disabled, because SLI guarantees failed WU's.
____________

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 18604 - Posted: 10 Sep 2010 | 10:01:47 UTC - in response to Message 18598.

lkiller123, It is unlikely that your 65nm GTX260 will work well.
I suggest you return it or sell it.

lkiller123
Send message
Joined: 24 Dec 09
Posts: 22
Credit: 15,875,809
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 18610 - Posted: 10 Sep 2010 | 15:07:16 UTC - in response to Message 18604.

Will there be a fix of some sort for this?

Or is it really just a driver problem?

lkiller123
Send message
Joined: 24 Dec 09
Posts: 22
Credit: 15,875,809
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 18612 - Posted: 10 Sep 2010 | 15:37:07 UTC - in response to Message 18610.

Another update:

I was running the 65nm stock-clocked for last night. It actually finishes some WUs, and some other WU fail at just about 90-95% finished.

I am guessing that some specific type of WUs will be able to run on the 65nm, and some does not. Am I right?

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 18613 - Posted: 10 Sep 2010 | 15:51:30 UTC - in response to Message 18612.
Last modified: 10 Sep 2010 | 16:10:22 UTC

It does not help that you put one 55nm and one 65nm GTX260 into the same system.

Only one task validated on device 0. So all tasks failed on device 1.

Having 2 big GPU's in the same case comes with it's own set of problems; overheating/cooling, and power requirements. Other components (motherboard mainly) can also overheat.

Try the GTX 260 (65nm) in a system all of its own, to see if it actually gets any tasks complete. Then try the 55nm card.
My guess is that the 55nm card completed the work unit that did finish.

You seem to be restarting the system many times, or perhaps you have configured Boinc to not use the GPU when the computer is in use?
If you are restarting, that explains the failures:
Tasks are running on either GPU randomly, so it looks like a task is doing fine, but when you restart it has a 50/50 chance of running on the other card. 4 or 5 restarts and the chances are that any given task will have run on both cards, and failed on the 65nm card.

coldFuSion
Send message
Joined: 22 May 10
Posts: 20
Credit: 85,355,427
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 18616 - Posted: 10 Sep 2010 | 18:12:30 UTC - in response to Message 18613.

It's also possible that the stock GPU voltage is insufficient

I had this problem with my GTX 470

You could try using MSI Afterburner to up the GPU voltage one notch at a time and see if it stabilizes it.

lkiller123
Send message
Joined: 24 Dec 09
Posts: 22
Credit: 15,875,809
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 18619 - Posted: 10 Sep 2010 | 22:42:30 UTC

It isn't the problem with the voltage. I've bumped up the 65nm up to 1.2V, and the tasks still fail.

I've noticed the pattern; every time I turn back on the monitor, the WU fails right away. There might be some problem linking to that...

I will be putting the 65nm by itself in the system. Let's see how it goes.

Profile liveonc
Avatar
Send message
Joined: 1 Jan 10
Posts: 292
Credit: 41,567,650
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwat
Message 18623 - Posted: 11 Sep 2010 | 10:02:36 UTC - in response to Message 18619.

Giving it more juice, isn't the same as drowning it. Don't kill your GPU on voltage & believe that it was just a crappy card. You give it what it needs & not more than it needs. Maybe it even got too much to start out with... But if everybody else says that they simply can't get anything running on that model, maybe it wasn't the voltage at all, but that it indeed was a crappy card...
____________

lkiller123
Send message
Joined: 24 Dec 09
Posts: 22
Credit: 15,875,809
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 18631 - Posted: 11 Sep 2010 | 17:38:16 UTC

Finally believed that it is the designs fault. Most 65nms simply can't work with GPUGRID yet..

At 1.2V and shader clocked at 1600, the card is definitely stable under Furmark. I will just take this card over to F@H until this problem gets fixed.

Thanks guys.

Post to thread

Message boards : Graphics cards (GPUs) : 65nm GTX260-216 issue.

//