View Single Post
Old 01-15-11, 08:23 PM   #23
Gentoo User
Join Date: Jan 2011
Location: Portland, Oregon
Posts: 14
Default Re: NVRM: os_schedule: Attempted to yield the CPU while in atomic[...] [260.19.21 x86

Originally Posted by voegel View Post
Unfortunately the 195.xx version is unsuitable for me, because I then have problems with HDMI audio. This only works flawlessly on my hardware with 256.xx and up.

Interestingly though, a downgrade to 256.xx doesn't solve the problem, whereas in Mythbuntu 10.04 I didn't have any trouble with that version.
I understand the HDMI audio issue. The point isn't necessarily to keep that version indefinitely but use it as part of diagnostic methodology. If the driver is otherwise stable (minus HDMI audio), that tells you the kernel + hardware is fine.

As an interesting point of reference, I have been running for hours now without an issue with the change in CPU/motherboard in Linux. Mind you, the previous combination worked fine in windows. Thus, this tells me the issue is with the nvidia drivers and that CPU/GPU combination (my previous post with the hardware/software matrix is still valid in my tests).

Update: 01/18/2011. I ran all day Saturday and Sunday with the change in CPU/motherboard just fine. I'm unclear as to what within the driver could cause the issue with Opteron 2378s under Linux but works fine under Windows. I am relatively convinced that it is either something within the kernel module itself (most likely by far) with a lowering probability that the proprietary OpenGL implementation is the cause.

Perhaps we should be asking a different question: How many other folks have multiple CPUs (I'm not talking multiple cores here. So if you only have 1 CPU chip that has 2+ cores, that isn't what I mean.)? e.g. CPU chip 1 and CPU chip 2.

Maybe the proprietary kernel module cannot properly schedule time if there are multiple physical CPUs vs multiple logical CPUs.

Last edited by jpi110; 01-18-11 at 11:43 AM.
jpi110 is offline   Reply With Quote