Advanced search

Message boards : Number crunching : Recommended driver for GF660?

Author Message
Martin Tomasek
Send message
Joined: 20 Apr 11
Posts: 2
Credit: 108,074,017
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 29756 - Posted: 8 May 2013 | 18:13:36 UTC

Hello,

a month ago I bought a new video card Gigabyte Geforce 660 (N660OC-2GD). Now I have still problems with calculations and I lost a lot of time. Please can anyone help with this?

Here is my old computer system (Win 7):
http://www.gpugrid.net/results.php?hostid=151270

and here is the newly installed system (Win 8):
http://www.gpugrid.net/results.php?hostid=151360

Whenever I pause a job, display driver crashes and displays the message that "Display driver stopped working and has recovered". But it happens without pause. Normally I work on the computer (using mail, word, excel, text editor), but it happens at night, when I'm not at the computer.

I have a problem here on the GPU Grid, my other projects go well (DistrGen, PrimeGrid). I tried to stop the calculations on the CPU (project Asteroids), but it did not help.

I also tried older drivers but without success. I tried in the Windows registry set a longer TdrDelay from 2s to hardcore 120s. Now when I pause work, so the picture freezes for a few minutes (can not move the mouse) and after two minutes restarts the video driver and everything is working again.

Is in a BOINC an error log that would help me find out what happened?

Sorry for my English, it is Google Translate.

Thanks for reply.

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 29759 - Posted: 8 May 2013 | 19:34:20 UTC - in response to Message 29756.

Your driver 314.22 generally works quite well. I don't think we could recommend any better one. However, I see that you jumped right in with the new Noelia klebe WUs. These are producing errors notoriously. One of them is the driver reset upon suspending / restarting tasks, so this is actually not caused by your system. Sorry, we don't know yet if there's anything we can do to avoid this (except not pausing the WUs).

MrS
____________
Scanning for our furry friends since Jan 2002

Martin Tomasek
Send message
Joined: 20 Apr 11
Posts: 2
Credit: 108,074,017
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 29763 - Posted: 8 May 2013 | 20:21:54 UTC - in response to Message 29759.

Ok, thank you for reply.

Sometimes I play games and I need to pause boinc. I'll wait for short runs, until it is somehow resolved (eg new drivers).

Sometimes the driver restarts without suspension work.

nanoprobe
Send message
Joined: 26 Feb 12
Posts: 184
Credit: 222,376,233
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 29764 - Posted: 8 May 2013 | 21:55:04 UTC - in response to Message 29763.
Last modified: 8 May 2013 | 21:56:04 UTC

Ok, thank you for reply.

Sometimes I play games and I need to pause boinc. I'll wait for short runs, until it is somehow resolved (eg new drivers).

Sometimes the driver restarts without suspension work.

If you're running Windows there is a very easy regedit that can help stop the driver resets/restarts. It's a known issue in Windows, especially on the high end cards, that has received very little publicity.

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 29771 - Posted: 9 May 2013 | 9:40:42 UTC - in response to Message 29764.

If you get a driver restart, when the GPU app stops running for example, and you get an app error, close Boinc, wait a minute and open Boinc again. The GPUGrid WU will usually resume from it's last check point. If it doesn't then the registry hack might be useful. I suggest people only try it if they are getting repeated app crashes.

I think this is the reg hack you were referring to, but there might be alternatives:
https://forums.geforce.com/default/topic/503962/tdr-fix-here-for-nvidia-driver-crashing-randomly-in-firefox/
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

nanoprobe
Send message
Joined: 26 Feb 12
Posts: 184
Credit: 222,376,233
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 29781 - Posted: 9 May 2013 | 15:42:25 UTC - in response to Message 29771.

If you get a driver restart, when the GPU app stops running for example, and you get an app error, close Boinc, wait a minute and open Boinc again. The GPUGrid WU will usually resume from it's last check point. If it doesn't then the registry hack might be useful. I suggest people only try it if they are getting repeated app crashes.

I think this is the reg hack you were referring to, but there might be alternatives:
https://forums.geforce.com/default/topic/503962/tdr-fix-here-for-nvidia-driver-crashing-randomly-in-firefox/

That is not the regedit I was referring to. I'll post the one I use on all my machines when I get home in case someone would like to try it.

nanoprobe
Send message
Joined: 26 Feb 12
Posts: 184
Credit: 222,376,233
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 29802 - Posted: 10 May 2013 | 21:29:47 UTC

Here is the regedit that should stop the driver has stopped responding/restart issues in Windows. Copy and paste all the text below into noterpad. Rename it fix.reg (or any name you'd prefer with .reg extension)
Right click on it and open with it Regeditor. You'll get warnings about editing the registry can botch things up, do you want to proceed. Hit yes and then reboot. Hopefully that will stop the issue. There is one negative I have found when doing this. Some GPU projects can stumble when you OC your card too much. If that happens you may get a BSOD instead of a driver reset. Happened to me once on Einstein@Home. I lowered the OC and no more problems.

Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Watchdog]
"DisableBugCheck"="1"

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Watchdog\Display]
"EaRecovery"="0"
____________

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 30030 - Posted: 16 May 2013 | 19:26:29 UTC
Last modified: 16 May 2013 | 19:27:10 UTC

Do I understand it correctly that this change avoids "fake" driver resets, when the GPU wasn't actually hanging. But if you get a real error you're f*cked and probably need a hard reset, just like in the "good old times", before vid drivers could recover a GPU?

MrS
____________
Scanning for our furry friends since Jan 2002

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 30048 - Posted: 17 May 2013 | 13:32:23 UTC - in response to Message 30030.
Last modified: 17 May 2013 | 13:43:15 UTC

Looks like it to me. I think something along the reg edit I linked to would be safer as it basically just increases the time before the driver restarts (not that this is the only issue, otherwise I probably wouldn't get blue screens as well).
I expect that Boinc and some Longer WU's need more than 2 seconds to close down gracefully.

I'm just after testing this and it looks like it works/is a fix:

Open a registry editor and goto the following location

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers]


Create a REG_DWORD file called TdrDelay and set the value to 20 seconds


Then navigate to

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\ GraphicsDrivers\DCI]


and Create or Edit (mine was at 7) a Value called


"Timeout"=dword:00000020


Then reboot to apply the settings.


So tested and works (for me so far):
After restarting I repeatedly suspended and enabled the Noelia WU's without issue. No driver crashes/restarts and no blue screens. Also tested closing and opening Boinc while GPU and CPU WU's were running. Again no issues.
I suggest others test this!
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

nanoprobe
Send message
Joined: 26 Feb 12
Posts: 184
Credit: 222,376,233
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 30053 - Posted: 17 May 2013 | 15:24:17 UTC - in response to Message 30030.

Do I understand it correctly that this change avoids "fake" driver resets, when the GPU wasn't actually hanging. But if you get a real error you're f*cked and probably need a hard reset, just like in the "good old times", before vid drivers could recover a GPU?

MrS

This may explain it better than I could.
http://msdn.microsoft.com/en-us/library/windows/hardware/ff553893(v=vs.85).aspx

I only made the note about 1 BSOD that I had while running a highly OC card at E@H because I thought it might be related to the regedit.

Jacob Klein
Send message
Joined: 11 Oct 08
Posts: 1127
Credit: 1,901,927,545
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30057 - Posted: 18 May 2013 | 6:01:35 UTC - in response to Message 30053.

I also found some good information here:
http://msdn.microsoft.com/en-us/library/windows/hardware/ff570088%28v=vs.85%29.aspx

However, I'm not going to be making any changes. The NOELIA tasks may sometimes currently crash the drivers, but I my system needs to work properly for other scenarios. I hope GPUGrid fixes their problem, as it seems relatively easy to reproduce it.

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 30066 - Posted: 18 May 2013 | 16:17:40 UTC
Last modified: 18 May 2013 | 19:47:38 UTC

Trying SKs suggestion with a timeout of 10.
Edit: well.. now upon suspending a Noelia WU my display hang for an estimated 20 - 25 seconds before the driver reset happened.

MrS
____________
Scanning for our furry friends since Jan 2002

Post to thread

Message boards : Number crunching : Recommended driver for GF660?

//