View Single Post
Old 05-03-12, 03:55 PM   #5
maro
Registered User
 
Join Date: Feb 2004
Location: Holy Roman Empire
Posts: 64
Default Re: Regression with 295.33 and GeForce GT 240

Quote:
Originally Posted by Licaon View Post
302.07 too?
yay, new drivers!

I installed the 302.07 driver on both machines with nvidia GPU, and with respect to the issue at hand the situation is unchanged: One machine with GTS 560 Ti works perfectly, the other one with the GT 240 has the issue.

While testing I noticed that the the "Atmospheric shaders" setting makes all the difference. Everything appeared OK without it, and locked up for seconds at a time with it. Which is sad because without it I have no shadows and - for reasons totally unclear to me - no antialiasing. Btw, I tried all combinations (eg. with atmo shaders without AA), but it really is the tick mark behind atmo shaders alone that controls the weirdness.

So I thought I had it sussed, but after closing the application the normal desktop would have these frozen seconds, too, so obviously this is not the full story.

Also noteworthy: one of the users plays stuff like quake 3 and serious sam without any problem on the machine. Of course those games date back to the stone ages before shaders were invented.

In the syslog I find only three entries for the period of my testing, which is a lot less than the actual frozen periods I experienced:

Code:
May  3 20:06:15 harold kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module  302.07  Fri Apr 27 11:50:07 PDT 2012
May  3 20:14:40 harold kernel: NVRM: Xid (0000:02:00): 8, Channel 00000005
May  3 20:14:42 harold kernel: NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
May  3 20:14:44 harold kernel: NVRM: Xid (0000:02:00): 6, PE0001 
May  3 20:14:52 harold kernel: NVRM: Xid (0000:02:00): 8, Channel 00000003
May  3 20:14:54 harold kernel: NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
May  3 20:14:56 harold kernel: NVRM: Xid (0000:02:00): 6, PE0001 
May  3 20:15:11 harold kernel: NVRM: Xid (0000:02:00): 8, Channel 00000003
May  3 20:15:13 harold kernel: NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
May  3 20:15:15 harold kernel: NVRM: Xid (0000:02:00): 6, PE0001 
May  3 20:15:41 harold kernel: NVRM: Xid (0000:02:00): 6, PE0001 
May  3 20:15:41 harold kernel: NVRM: Xid (0000:02:00): 13, 0001 00000000 00000000 00001130 ffb25353 0000e4e4
Another observation using nvidia-smi: the memory consumption on the card did not exceed 26% during the test. The card has 1024 MB.

If anyone has any further ideas, please come forward.
maro is offline   Reply With Quote