[mythtv-users] OT. Howto diagnose machine dying/crashing

[Stephen Boddy]

On Friday 15 September 2006 17:39, Henri Paasovaara wrote:
> Have you tried to test your memory chips? You could try it with memtest86+
> program/package.
> I had chips which didn't work with motherboard I had. So I had to change
> back to old mb.

Hmm. I've had a misbehaving system for a while now (quite similar to the OP in 
specs and symptoms) that hard locked particularly on heavy I/O where CPU 
usage was significant. Absolute git to figure out. I actually added more 
memory as well which actually exacerbated the problem. It cycled through 
memtest 8 times without a single error with both sticks, but start a stress 
test script I wrote and the thing hard locked immediately even though it 
could boot the system OK.

Lots of component juggling later I've concluded the following. For some reason 
if I run the memory at 400MHz whilst the CPU FSB is locked to 333MHz, it 
causes hard locks even though the board should run OK with the disparity. By 
slowing the memory down to match the CPU FSB I cured the really easy to 
reproduce hard lock. I've also seen no more of the rarer hard locks I saw 
beforehand either, although I'm still getting everything back to normal.

Moral: Memtest is useful, but in the end patient drudge work, process of 
elimination and experimentation will find the answer.
-- 
Steve Boddy