Advanced search

Message boards : Number crunching : All WUs failing with "computation error"

Author Message
YoDude.SETI.USA [TopGun]
Send message
Joined: 3 Jul 09
Posts: 4
Credit: 8,848,685
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwat
Message 12293 - Posted: 3 Sep 2009 | 2:25:15 UTC

Here's the setup

2x9800GTs non-sli
A recent nvidia driver, 190.62
Boinc 6.6.2
Vista 64 U

Every WU is failing as soon as it begins to run. I've tried everything I can think of to make this setup work. I was able to run these WUs with these cards about a month ago and going back to an older driver doesn't help.

Any ideas?

Profile Zydor
Send message
Joined: 8 Feb 09
Posts: 252
Credit: 1,309,451
RAC: 0
Level
Ala
Scientific publications
watwatwatwat
Message 12308 - Posted: 3 Sep 2009 | 13:35:23 UTC - in response to Message 12293.

Try this:

http://www.gpugrid.net/forum_thread.php?id=1293

Its easy to do .... takes a few minutes only. Often what will happen is the belief in the message "Unstall Complete" from windows becomes absolute, its far from the case. Usually bits of drivers get left behind, eventually .... those bits build up and you dont know they are there - and it goes boom.

No guarantees, but, its worked with many, depends if you have old driver bits left behind, and what they are.

Regards
Zy

YoDude.SETI.USA [TopGun]
Send message
Joined: 3 Jul 09
Posts: 4
Credit: 8,848,685
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwat
Message 12311 - Posted: 3 Sep 2009 | 14:41:39 UTC - in response to Message 12308.

Thanks. I have it downloaded but must leave for work so, I'll give 'er a try this evening and let you know the results.

Actually I think this may fix the problem (just a feeling though) as I've had problems with this system (Evga 790i - "Rodan") since I installed ATI 4870s in it a couple weeks ago to see if I could get them work in x-fire. No luck with that. Since then, those cards have moved to another system (X58 Classified - "Godzilla") that supports them in x-fire and they are doing quite well there but Rodan has been giving me serious driver issues since then.

Thanks again,

Yo-

Oh and if you're wondering, there are 2 more systems aptly named, "Mothra" and "Gammera".

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 12334 - Posted: 3 Sep 2009 | 20:47:27 UTC

Are you using remote desktop to access the machine? Because all of your WUs which run for a while switch to "Device Emulation (CPU)" just before they fail. That means either the driver crashed, can't find the GPUs any more or has been replaced by a different driver. The latter is the case if you access a machine by remote desktop (and some other remote control software): the proper driver gets replaced by some virtual driver for the remote session and the GPU crunching fails (also happens with ATIs).

If that's the case you can use REAL VNC. It's more sloppy but doesn't break anything.

MrS
____________
Scanning for our furry friends since Jan 2002

YoDude.SETI.USA [TopGun]
Send message
Joined: 3 Jul 09
Posts: 4
Credit: 8,848,685
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwat
Message 12354 - Posted: 4 Sep 2009 | 2:13:56 UTC - in response to Message 12334.

In fact I am using remote desktop. Here's what I've done.

1. Ran "Driver Sweeper" and cleared out everything from all old graphics drivers.
This seemed to work. The system came up and I reinstalled the 190.62 nvidia
drivers without issue. (I felt better after doing this too :)

2. GPUGrid WUs downloaded and actually started running (I was pretty happy at
that point) and I let them go for a few minutes, then shutdown the system.

3. Moved my monitor cable back to my main system and restarted the other.

4. Went to see the amazing accomplishment via remote desktop and WTH? ALL
6 WUs were just gone. Boinc reported No CUDA devices found and there
was some messages that it couldn't run on non-existent devices or something.

5. Moved the monitor cable back to the offending system and all was well
(after another reboot of course) CUDA devices found and everything good.

6. Discovered that something is very much amis if you switch your monitor cable.


So....I'm still a bit unhappy, now tell me about REAL VNC

Yo-

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 12361 - Posted: 4 Sep 2009 | 20:24:20 UTC - in response to Message 12354.

Yo Dude, I just told you that the WUs are going to fail as soon as you connect to the machine via remote desktop. Are you really surprised that this is just what you're seeing?
(Real VNC)

MrS
____________
Scanning for our furry friends since Jan 2002

YoDude.SETI.USA [TopGun]
Send message
Joined: 3 Jul 09
Posts: 4
Credit: 8,848,685
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwat
Message 12380 - Posted: 5 Sep 2009 | 15:24:15 UTC - in response to Message 12361.

MrS,

I've installed REAL VNC as you suggested and can now report, all is well.
Thank you for your enlightenment with this problem.
I've successfully completed one WU without issue.

Thank you again, this looks to have done the trick.

Yo-

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 12381 - Posted: 5 Sep 2009 | 15:46:53 UTC - in response to Message 12380.

That's nice to hear :)

MrS
____________
Scanning for our furry friends since Jan 2002

Post to thread

Message boards : Number crunching : All WUs failing with "computation error"

//