Announcement

Collapse
No announcement yet.

Need input on weird sudden reboot bug

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Need input on weird sudden reboot bug

    Hi Folks,

    so I've got this weird problem but I'd need some input/ideas to pin it down enough for a proper bug report, so I'm asking you for help.
    First I need to introduce you to my admittedly non standard harware configuration which probably plays a big role in the bug since it happned on it since the beginning (so every Kernel Version from pre 3.0 to 3.3-rc3).

    I've got a Tyan Dual G34 Motherboard with 2 8-Core Magny Cours Opterons
    with 8*2 GB RAM one stick for each memory channel of the Opterons.
    Added to that I've got a Radeon HD 5770.

    On the software side I've been running Arch Linux and Windows on this system and am running 3.3-rc3 at the moment.

    So here is the different situations which trigger a sudden reboot of the system:
    - Running OpenCL programs with Catalyst (only ones with higher memort traffic trigger it) on Linux, it works on Windows
    - Sometimes when using (with the Open Source driver) Gnome 3.x or KDE 4.x it just happens out of the blue
    e.g. when maximizing a Youtube Video or even just clicking on a link, it seems to be related to the amount of change on the screen
    - In earlier versions I could trigger it quite reliably by scrolling through my photo collection in F-Spot or Shotwell but the thing is only when using Gnome or KDE it works without problems in Xfce 4.8 even though I use Xfwm with compositing

    The bug is triggered only in Gnome and KDE not when using Xfce 4.8 even though I use compositing in Xfwm.
    So I've been running Xfce for most of the time and the system is rock stable with it not a single crash in the past few months.
    It seems that some obscure GL command or something used by both KDE and Gnome makes the system unstable.

    On the other hand I can play Games on Windows without problems though I've done that a lot less then using Linux so there is some chance it just didn't trigger but on the other hand the machine made it through one game night without crashes on Windows.

    As the title says the bug looks like a sudden reboot and there is absolutely nothing in the logs.
    The screen just freezes for 2-3 seconds and sometimes the sound stutters and then the system just reboots.

    All this is much to weird and unclear for a proper bug report so I'm asking here if anyone got an idea.

  • #2
    Make sure you have a large enough power supply.

    Comment


    • #3
      I've got a 1000 Watt power supply but I'll test it again with maximum load on the CPUs and GPU on Windows, though I already know that there is no problem even with X-Plane 10 which hogs quite a bit of CPU und GPU.
      Also note that when the crashes happened there was barely any CPU load.
      On the ATI OpenCL forums a while back I saw a post from a guy who also experienced sudden reboots with OpenCL programs and high PCI-E bandwidth he could fix it by turning on some RAM access barriers in the BIOS but I couldn't find anything equivalent with my board and suspect that Opterons are different in this regard compared with the desktop phenom he was using...

      Comment


      • #4
        Ok, so after coming home I just tried running LuxMark 2.0 which is an OpenCL benchmark using both GPU and CPU and it runs completely stable in Windows so it seems even at full load the power supply handles the system just fine...

        Comment


        • #5
          Hi guys, though it seems no one seems to have any real idea what could be the problem I've kept investigating.
          I have been running with a friends Radeon HD 4670 and no crash nor any other malfunction has happened since then.
          Any idea how to further investigate whats happening with my other card and why it's not playing well with my other hardware?
          The only thing I can think of is trying my card in a friends computer for a while to see whether it's really the combination of hardware that's causing problems.

          Comment

          Working...
          X