help: oc'ed system intermittently shutting off

[potatoBBQ]

i have a p4 2.26 oc'ed to 2.55 @ 1.55v that's been running stable for nearly a year and just recently it started shutting off intermittently. i'd be just browsing and it would just shut off abruptly and i wouldn't be able to turn it back on for a few minutes. the motherboard leds are still on when this happens, but pressing the power button does nothing so i end up manually turning off the ps from the back. my mobo temp is around 38C and cpu temp is at 42C (it's been pretty hotter than usual lately, the temps aren't really that much lower on cool days). i don't have any spare parts so it's hard for me to troubleshoot... i'm thinking it's the motherboard or cpu, so any ideas would be greatly appreciated, thanks

p4 2.26 @ 2.55 @ 1.55v w/ zalman cnps7000-alcu heatsink
enermax noisetaker 420w ps
asus p4b533-v motherboard
512mb corsair pc3200

 

[Navid]

I would clean up the CPU heat sink and remove any dust bunnies from it. Be careful not to damage anything by ESD. Keep touching the ground of the PC (PSU case) to keep yourself discharged when you are working inside the case.

Are you saying that the CPU temperature is 42 at idle?

To test the stability of the system, you can run prime95.
http://www.mersenne.org/freesoft.htm
You should be able ti run it overnight with no errors.
If you get errors, you should either improve cooling or reduce overclock rate until you get no errors.
If you do get errors, you should expect freezes and reboots since you have an unstable system.

Edit:
You should first run memtest in order to see if it is the memory that is limiting.
http://www.memtest86.com/
You should be able to run it overnight with no errors. If you get errors, you should either relax your memory timings, or change the mem:cpu ratio, or reduce the overclock rate.

After, you get to run memtest with no errors overnight, you should then run prime95 ans described before.

 

[Fern]

To the good advice above, I'll add that you may want to d/l and install motherboard monitor 5. You can set it up to monitor your system (volate and temps) at intervals of say 10 seconds.

After a shutdown, go check your logs to see if the temps got too high or if voltage fluctuaitions are shutting you down.

fern

 

[potatoBBQ]

i've tried running prime95 but it runs w/o finding any errors for like an hour... until it just completely shuts off. while running prime95, the mobo temps reach 40C at times which seems unusually high, but i don't see anything physically wrong w/ the board. this is w/ 2 intake fans, 1 exhaust fan, and the side panel removed on a lian li pc-65 case. i guess i'll take the machine apart today to see if there's anything visibly wrong and then try running mbm5. i was sort of leaning towards it being a mobo problem because i figured if it was the cpu, i would at least experience some sort of freezing/instability at some point but i don't. the machine is either working 100% stable or it just completely shuts off, nothing in between.

 

[Mullzy]

When my voltage isn't high enough Prime95 will reboot my PC without throwing a normal error (like the rounding one we've all seen!)

Here's what my 2.8C did to me once I started getting above 230 FSB

1) Increase FSB, Prime Reboots PC (random time, 2-30 minutes).
2) Increase voltage, Prime Fails with rounding error (random time 2-400 minutes)
3) Increase votage again... Prime solid

It may be that you temps have increased due to dust build-up on the Zalman. Mine grabs dust like crazy. Try spraying it out with some compressed air and see if that gets you 2-5 degrees on your CPU and stability again.

 

[Mav3N]

i have a similar issue

i have a 2.4c that overclocked or not will reboot if the system remains idle.
If i run anything cpu intensive games or 3d apps etc my system will run like a horse for days on end.
But if i let the cpu go idle and cools after a while the system just shuts Off. And i cant turn it on with the power button. i have to flick the switch on the back of the ps for a few seconds then turn the switch back and then hit the power button on the front of my case.

ive run the prime95 without (like i said as long as its using the cpu its great)
my temps never get above 41c
and idles close to 30c.

two sticks of ocz 3200 ram 512mb running in dual channel mode
mb is an Abit IS7-E
420w ps
2 hd and 1 cd drive

 

[Mav3N]

update: my problems appeared to be solved.
i reseated my processor and hsf

i cleaned off the current thermal compound and then reapplied
reseated the processor then placed the hsf back in place.

i took special care to make sure everything was as snug as possible.
been running fine now without reboots even when idle

 

[Concillian]

Prime95 is great for stability testing. When testing CPU I use the middle test (large FFTs I believe). I can have a system that works Prime 95 for 4 hours fine but is still sketchy in some games BF1942 and BFV seem especially prone to crash to desktop on a less than TOTALLY stable CPU overclock). You want to be able to withstand 24 hours of Prime 95, if running on an HT P4, you want to run TWO instances (and set priority of the first to CPU0 and the second to CPU1)

I disagree with running Prime95 overnight. It's cool overnight, you are not stressing the system to the max. Prime95 on a warm summer afternoon in a warm room is a much better stability test.

I actually lower my OC (or increase my voltage) until I can run Prime95 24 hours, then I lower the OC just a little bit more to add a little bit more stability. You give up 20-30 MHz, but I feel the stability is WELL worth it.

In all likelihood you just need to lower your OC a little.