Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Generalized Fermat Prime Search :
Errors since updating
Author |
Message |
|
Since I allowed a windows update of the Cuda software yesterday I have had nothing but errors.
I am at work and will reload from the NVidia site when I get home and let you know if it improves. In the interim I will disable CUDA work.
Has anyone else had similar problems?
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 | |
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 14045 ID: 53948 Credit: 482,681,090 RAC: 578,366
                               
|
Since I allowed a windows update of the Cuda software yesterday I have had nothing but errors.
I am at work and will reload from the NVidia site when I get home and let you know if it improves. In the interim I will disable CUDA work.
Has anyone else had similar problems?
It's your video driver that's the problem. The 295 or 296 windows drivers have a SERIOUS bug in them.
Either upgrade to a more recent version of the driver, or downgrade to the 285 driver. Doing either will solve your problem.
You can also avoid the problem by changing your power setting such that your screen or screens are never put into sleep mode, but it's better to upgrade your driver.
____________
My lucky number is 75898524288+1 | |
|
|
Nope, I've been running 295 since I got the new laptop about 3 months ago with no problems. I am not sure but I think the 630 is OK on 295, but I know that others have had problems so your advice is the best general tip.
I finish in an hour and will get the latest from NVidia when I get home, whatever the latest is now.
I had 1 critical windows update and five reccomended ones yesterday, I refused the remote desktop one and got the Cuda plus 2 security ones, I do not have the windows numbers for them. Since then it has had only failures whereas before it never has a single problem with GFN shorts.
I will only update from Nvidia in future!
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 | |
|
|
Ok I'm home now and found that :-
1. Windows was waiting for a restart to enable the updates
2. The CUDA update had failed to load.
I am running 306.97 now, I tried to load 310.7 but it said it could not find the hardware.
I removed the windows update from the programs list and have restarted GFN cuda short.
Interestingly all the errors were within 5 seconds of an hour run time.
I will see if it is still giving errors an hour from now.
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 | |
|
|
Good, all working now with over 2 hours gone.
I'm not sure if updating the driver or deleting the failed windows attempt did the trick but it is normal now.
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 | |
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 14045 ID: 53948 Credit: 482,681,090 RAC: 578,366
                               
|
DaveB,
It only appears to be working because you were looking at the computer. There's a serious bug in the 295/296 drivers and I strongly recommend either upgrading or downgrading to a different driver.
With those drivers, if Windows powers off the screen -- and you're on a laptop so it's almost certainly set to do that -- the GPU becomes unavailable as a CUDA device. From that point on, at least until the monitor is powered on, any CUDA program (including BOINC tasks from any project, anywhere) will be unable to start. Programs that are already running are fine, but every program that starts to run will be unable to attach to the GPU and it will fail.
That is what the errors were showing.
Unless you want to configure your laptop to never power off the screen (and I can't imagine why you'd want to do that), the only way to prevent this problem is to change to a different driver.
The reason all those tasks took almost exactly 1 hour of time to error is because GeneferCUDA is specifically programmed to wait one hour before reporting the error. This prevents computers configured such as yours from flooding the server with errors.
Good, all working now with over 2 hours gone.
Actually, you are mistaken. It works just fine when you're looking at it -- because the screen is powered on. But once your leave the computer and the screen powers off, the computer is unable to start new GPU tasks and they all fail. Any GPU tasks that your computer attempts to start while the screen is powered off are going to fail.
____________
My lucky number is 75898524288+1 | |
|
Message boards :
Generalized Fermat Prime Search :
Errors since updating |