Mobo/CPU failure advice

[TNorthover]

Just after a little advice really, on the off-chance that someone has had the same problem.

When I first turn on the computer (S478 P4) in the morning, it often fails to boot (kernel panic halfway through -- different point each time). When rebooted it seems to work fine.

I've run memtest, and if run very first thing in the morning it reports memory errors (at different places each time) then crashes, after that things seem flawless.

Now, because it clears up after a while, I suspect it's a heat problem (contacts getting better when heated and all that rubbish). Both CPU and mobo are overclocked (I know what's happening now is to be expected with overclocking, but I've got 3 good years out of it and it's heading towards time to replace anyway, so I'm not especially bothered -- calculated risk and all that); the heatsink fell off the motherboard a while back, but on the other hand the CPU has always run pretty hot.

Anyway, my thinking is that it's not the memory (errors show up at different point each time). That really leaves either mobo or CPU; It could be a memory bus error on the motherboard, perhaps a timing issue with the CPU, or the CPU itself.

So, anyone got any experience with these exact problems, or know a good way to distinguish the different cases in software (I was thinking of writing a tiny boot program to stress the CPU without touching main memory (small enough to fit in cache and twiddle registers) and see if that fails; if it works it would distinguish between "motherboard memory bus" and "other" problems. Anyone know if that's likely to work?).

Any advice appreciated.

[radoslawc]

Hi
It could be many things that cause this bahavior not only heat but broken capacitors in motherboard power system (near MOSFETs they are constantly heated causing electrolit to dry), broken PSU, memory/cpu connects covered with metal oxides (heat and humidity) - if you have another mobo/mem/cpu set (or your friend or someone) try to replace each of them and test system if it works to find broken part. For stress testing use mprime from www.mersenne.org - - its really good has option to set FFT size to test CPU only or CPU + RAM -- see what fails first.
Good luck
_________________
“You know all those extra lines of code that we sometimes refer to as bloatware. It's actually very important and efficient stuff. There is some very substantial functionality built into the operating system.” Laura DiDio

[TNorthover]

I would have thought most of those things would be persistant, rather than clearing up after a couple of minutes. Annoyingly I don't have another set handy, that'd be much easier. I'll look into the mprime, but unless I can convert it into a boot program I don't see much hope (if the computer boots up, it works fine until the next day).

Oh well, thanks for the suggestions.