[wplug] Another OT buggy-hardware post

[Poyner, Brandon]

I had a similar problem on a home built Athlon 1.3Ghz.  Specifically
memtest86+ would throw memory errors if I clocked it appropriately in
the BIOS, leading to system lockups and reboots.  If I under clocked to
1.25Ghz the memory errors went away in memtest86+ and the system was
perfectly stable.  At first I suspected the memory was bad, and Mushkin
replaced the memory with hardly any questions asked.  It wasn't the
memory.  I removed everything but memory, pci video card, floppy drive,
motherboard, cpu, and keyboard and it exhibited the same behavior.
Later on I replaced the motherboard and that didn't help.  Even later on
I replaced the CPU and I was able to clock it appropriately.  I never
did exactly nail it down, but I suspect the CPU was faulty.

So I would suggest running memtest86+ at the proper clock speed and
under clocked and see what happens.

Brandon Poyner
Network Engineer III
CCAC - College Office
412-237-3086
 
 

> -----Original Message-----
> From: wplug-bounces+bpoyner=ccac.edu at wplug.org 
> [mailto:wplug-bounces+bpoyner=ccac.edu at wplug.org] On Behalf 
> Of Brandon Kuczenski
> Sent: Tuesday, March 14, 2006 1:56 PM
> To: WPLUG
> Subject: [wplug] Another OT buggy-hardware post
> 
> I'm looking to tap into the wisdom of WPLUG for advice in 
> minimizing the
> financial cost of fixing my desktop computer, which crashes 
> frequently 
> (MTBF ~36 hours) under heavy CPU load.  I'm running Debian testing, 
> 2.6.11-k7 on an Athlon XP 2000+ and an Asus A7N-266 mb.
> 
> There are three types of crashes.
> 
> One, if I set my motherboard's jumpers properly (i.e. my 
> 2000+ runs at 
> 1667 MHz) then the computer is guaranteed to crash within 
> five minutes of 
> boot.  The computer only runs at its reduced clockspeed of 1250MHz.
> 
> Two, if I run 'lame' on about 30 minutes of audio, there's a 
> 90% chance 
> that the computer will crash (behavior: spontaneously powers 
> down).  This 
> also happens in certain other situations with high CPU load.
> 
> Three, sometimes the computer will randomly lock up -- screen 
> freezes, 
> mouse / keyboard unresponsive, sshd doesn't answer connection 
> requests.
> 
> After considering the situation, there are four different 
> possibilities 
> that I think are likely.  Any one will cost about $80-100 to 
> fix, so I'd 
> like to get it right.
>    a. Temperature.  I have lm-sensors installed; mobo temperature is 
> 40-43C and CPU temperature is 49-51C in the steady state.  
> These don't 
> seem to be abnormally high... fix is to buy a better case+heat sink.
>    b. Faulty processor needs replaced (because of the 'type 1 crash')
>    c. Faulty motherboard needs replaced (alternative diagnosis of 'b')
>    d. It's possible that the 'type 3 crashes' are caused by a 
> bad video 
> card.
> 
> I haven't really considered the memory because the problem 
> has not always 
> occurred (though I can't remember when it started occurring). 
> But I'll 
> install and run memtest86+ while I wait for list responses to 
> trickle in.
> 
> I built the box myself with parts (presumed new) from a 
> discount on-line 
> parts house.  The Video Card is from eBay.
> 
> The case has really awful cooling, which is why I thought temperature 
> might be the culprit.  I'm also running a cheap aluminum 
> heat-sink instead 
> of a copper one.  It's seated so tight that I couldn't get it off the 
> processor to check the conductive paste (which I applied when 
> I installed 
> the heatsink).  I would have to probably take the whole thing 
> apart in 
> order to be able to get at the heatsink to remove it.
> 
> If i were better off, I'd just buy a new Asus board, new 
> processor, new 
> case, and new heat sink, but I won't be able to swing that.  
> As it is, I 
> am in a constant state of panic that my computer will shut down.
> 
> Any advice?
> 
> -Brandon
> 
> _______________________________________________
> wplug mailing list
> wplug at wplug.org
> http://www.wplug.org/mailman/listinfo/wplug
>