Thank you amd, now I can't use Linux

[gradinaruvasile]

Are you using a distro that uses gcc 4.8? I've noticed a lot of problems with 4.8 related to variable sized array handling which we use a lot in the radeon driver. You may be hitting some issues related to that. Can you try a distro with an older version of gcc?

gcc 4.7 on Debian 64-bit Testing.

~$ apt-cache policy gcc
gcc:
  Installed: 4:4.7.2-1
  Candidate: 4:4.7.2-1
  Version table:
     4:4.8.1-3 0
          1 http://ftp.debian.org/debian/ unstable/main amd64 Packages
 *** 4:4.7.2-1 0
        500 http://ftp.debian.org/debian/ testing/main amd64 Packages
        100 /var/lib/dpkg/status

[agd5f]

You might still try this patch for 3.11:

http://people.freedesktop.org/~agd5f/0001-drm-radeon-atombios-fix-variable-sized-arrays-for.patch

[pandev92]

You might still try this patch for 3.11:
http://people.freedesktop.org/~agd5f...rays-for.patch

This is for me?

[agd5f]

Anyone with problematic behavior.

[gradinaruvasile]

Anyone with problematic behavior.

Hm. That can be interpreted...

Anyway, i applied the patch over the git kernel (there were a few whitespace warnings but that doesnt seem to be a problem).
I left the R600_DEBUG=nodma in /etc/environment, can that cause problems?

Lets see...

P.S. Are there any known specific kernel compile options that help (or ones that dont)?

[gradinaruvasile]

@agd5f:

Do you know when the shader opcodes for r600 will be back in mesa git's master branch? Cause a few days ago they were awol causing most games to display garbage textures (or nothing at all, in case of tf2). They work in the 9.2 branch.

[gradinaruvasile]

You might still try this patch for 3.11:
http://people.freedesktop.org/~agd5f...rays-for.patch

No luck. System frozen after a few hours when i was just browsing the net (the temperature monitor reported 15C, no heavy cpu or gpu activity).

Now i removed the nodma line from environment, added iommu=pt to command line.

PS the mobo is a Gigabyte F2A85X-D3H - if that has any relevance.

[gradinaruvasile]

So. No dice.
I tried all kinds of kernel command line options regarding iommu, disabled it in the bios etc - i even compiled the kernel with gcc 4.8, but all have the same result: hard freeze after 4-5 hours or after ~1 hour if previously the computer was used then left alone (and if it didnt freeze while not being used that is).

Other thing to note is that using xbmc (with or without vdpau) i get freezes after ~1 hour of playback. The sound replays the last ~1 sec or something once or twice then stops completely.

Playing opengl games is still stable and never got a freeze.

Most freezes occur when the computer is used lightly so i really dont think its about overheating.

Other thoughts:
i have 4 GB of RAM and the 64 bit kernel requires iommu for certain tasks and even if i disable it from BIOS, it will use its software variant. iommu=off on the command line results in tons of errors.
i tried radeon.msi=0 too, but that too resulted in tons of errors from AMD-Vi.

I observed that i get a lot of cpu speed drops/throttles - for example i compile the kernel with 4 threads, the ondemand governor asks for 1.8 GHz or the like speeds when it should keep 3.2 GHz.
I forced the performance governor, /proc/cpuinfo shows 3.2 GHz, but cpufreq-aperf shows ~1.9 GHz.
Temperature is around 48-50 C, so that isnt supposed be a problem. Does the radeon driver has its say in the actual cpu speeds? Or indirectly causes something that triggers speed throttling?

So, any other options? I can recompile the kernel if specific options are needed or at least monitor the driver - i have mounted debugfs to extract the clock speeds so if there are any other meaningful information sources in there, i could monitor/log them too.

[duby229]

So. No dice.
I tried all kinds of kernel command line options regarding iommu, disabled it in the bios etc - i even compiled the kernel with gcc 4.8, but all have the same result: hard freeze after 4-5 hours or after ~1 hour if previously the computer was used then left alone (and if it didnt freeze while not being used that is).

Other thing to note is that using xbmc (with or without vdpau) i get freezes after ~1 hour of playback. The sound replays the last ~1 sec or something once or twice then stops completely.

Playing opengl games is still stable and never got a freeze.

Most freezes occur when the computer is used lightly so i really dont think its about overheating.

Other thoughts:
i have 4 GB of RAM and the 64 bit kernel requires iommu for certain tasks and even if i disable it from BIOS, it will use its software variant. iommu=off on the command line results in tons of errors.
i tried radeon.msi=0 too, but that too resulted in tons of errors from AMD-Vi.

I observed that i get a lot of cpu speed drops/throttles - for example i compile the kernel with 4 threads, the ondemand governor asks for 1.8 GHz or the like speeds when it should keep 3.2 GHz.
I forced the performance governor, /proc/cpuinfo shows 3.2 GHz, but cpufreq-aperf shows ~1.9 GHz.
Temperature is around 48-50 C, so that isnt supposed be a problem. Does the radeon driver has its say in the actual cpu speeds? Or indirectly causes something that triggers speed throttling?

So, any other options? I can recompile the kernel if specific options are needed or at least monitor the driver - i have mounted debugfs to extract the clock speeds so if there are any other meaningful information sources in there, i could monitor/log them too.

50c is really hot for an idle temp. I have a PhII 955BE that it's load temp is 54c. Idle temp is 36-37c

[crshbndct]

Does it crash while using Chrome? I had the same problem, and it was a problem with Chrome. Disable all hardware acceleration for videos and stuff, reset all the chrome flags to default.

It sounds like it might be, since you mentioned chrome before. Also, check that the DPM is actually working, it isnt 100% there yet.