Video card acting up
 

I have an nvidia card 9800gtx, Mandriva 2009 32bit. Everything works fine, the driver is installed, the card is recognized properly. Compiz works, and all seems to be fine, but every now and then the screen will just freak out, show like a grid over it, but the machine is still working. I can still access the machine over the network, and if I drop to a shell I can see what I type behind it, its just all screwed up. It isn't a heat issue with the card. The machine seems to be fine, I can leave it run all night, it seems to act up usually when playing flash video, but has done it just out of the blue before. Would you think that this is a bad card, or some driver issue?

If it's a CRT monitor you may lower your horizontal frequency and see if that helps. If it's a LCD and you are using digital input then I'd suspect the card has issues. Drivers normally cause the issues all the time and don't act up under different conditions like flash player.

If it happens in X and you do ctl+alt+backspace then the issue should go away if it's driver related or monitor related of course that's only going to last a minute if you are running in graphics mode or init 5 then it will restart X again.

Try starting the machine from power off, then go to console and type init 3 then startx and if it acts up try the ctl+alt+backspace and see if it clears up and drops you to a shell. You should not even have to type reset to clear the terminal it should be clean if it's driver or monitor settings issue.

"the driver is installed" ... what driver ? the version ? Anything in the logs ? Xorg.0.log ? syslog ? etc.

The driver is the nvidia 177.70. I don't know that it is the most up to date, but that is the one mandriva used by default and it seems to work.

The only thing I can find in the logs is this from syslog right before it dies(I think)

Apr 24 20:02:58 Cory-desktop kernel: NVRM: Xid (0002:00): 8, Channel 00000003
Apr 24 20:02:58 Cory-desktop kernel: NVRM: os_map_kernel_space: can't map 0xe0200000, invalid context!
Apr 24 20:02:58 Cory-desktop kernel: NVRM: os_map_kernel_space: can't map 0xe0200000, invalid context!
Apr 24 20:03:00 Cory-desktop kernel: NVRM: Xid (0002:00): 62, CMDre 00000000 0000085c 0100cb1a 00000007 00000000

I have no idea what this means, but I am quite certain that this is the last thing that it is logging. I found several instances of it in the log, and each was just prior to a restart.

I think it's probably related to the nvidia drivers, and with such a new card I would use the newest nvidia drivers:
http://www.nvidia.com/object/linux_d...32_180.51.html

I thought about that, but if it messes up the install to where I can't get into X, how do I roll it back to the old driver? Not that it wil, but on the off chance.

To use the regular driver edit '/etc/X11/xorg.conf' and change:

to

Or you could just run 'xorgsetup' as root.

Success. After coming back into X though, everything seems just the slightest bit choppy. I wonder if that is because I'm still using an old CRT.

Ah, it appears that I still have the same issue, although it seems less frequent. I can't find any info out there particular to the card though. I am pretty much out of ideas.

Does this happen when not using Compiz ?

If so, try different live CDs, if you have Window$ try that, if it still happen then it might be the card itself, or sometimes PSU or mobo. Are these new ? how old are they ?

I am relatively confident that it is not a hardware issue. Card and power supply are a week old, 850 watts. Mobo is maybe a month old. Cannot seem to reproduce the error using ubuntu or suse. I'm running gkrellm and the temperatures are fine.

I disabled compiz when I woke up today, and I am going to run it and see what happens.

EDIT: Didn't take long to see that the problem is the same with no desktop effects. I have an old 80gb hard drive laying around I am going to toss in here and dual boot windoze and see if anything happens there.

Yes after much research it is just a linux driver issue. No problems in windows, and after reading errors like this have been fairly common in linux with certain nvidia cards. So I spent my money for nothing :(

Complain to nvidia about it, they may fix it.