STILL! Re: [GLLUG] desktop freezing??

[STeve Andre']

On Thursday 06 December 2007 19:37:05 Benjamin Cathey wrote:
> >->> Try posting copies of the complete /var/log/dmesg and the section of
> >->> /var/log/messages from about the last 10 minutes before the freeze
> > until ->> you shut it down or reboot.  If you have logins and things in
> > these ->> logs, you might want to blank out IPs, usernames, etc.  This
> > will end up ->> in public archives after all, so it's a good idea to
> > review them.
>
> I am not sure when it is happening or what to be looking for - everything
> seemed fine after the new power supply and now it is happening again.  The
> only thing I can figure is it MUST be overheating because it was running
> fine with the new power supply until I put the case cover back on.
>
> I turned the fans up to 'medium' speed - that SHOULD be enough.
>
> Where can I check to see if an overheat caused this??

You are probably onto something.  I would run the computer for several
days with the cover off, to make sure that it is indeed a temperature
related problem.  If you can then see freezes after putting the cover
back on you'll have a good idea, but only if the computer is rock solid.
I'd use a week to determine this.

If this is the case, you have a part which is partly OK.  As long as its
cold it's happy.  These things are a bitch to debug.  It might not even
be the motherboard, though I'd  bet that it is.  At this point I think I'd
try to find an identical motherboard and swap it out.  You can easily
spend 50 hours tracking something like this down.  An infra-red heat
imaging camera is useful for this.  I once had a friend who's company
had one and I got permission to take a problematic radio there, and
found that a tiny resistor in the receiver was getting too hot and
causing distortion.  If I hadn't had the camera it would have taken a
long tme to figure that out.

Have you run something like memtest86 on the unit with the case on?
If it doesn't crash but finds errors that will say something.  If you're
running with multiple sodimms or whatever, try running with  just one.

--STeve Andre'