PDA

View Full Version : Hep diagnosing an intermitent, potential PSU problem


CestusGW
02-04-2007, 03:24 PM
The rig:

AM2 X2 3800 (stock clocks)
Asus M2N32-SLI Deluxe
XFX 7900GT (stock clocks)
2x512Mb DDR-2 667 (4-4-4-12)
LG DVD-RW drive
250Gb SATA HDD
Floppy Drive
Coolermaster Praetorian Case (4x80mm fans)
550W Modular PSU (Ultra X-2)

The problem:

Over the past few weeks, I've been having application crashes, and sometimes full hangups with increasing frequency (from once every few days to once or more per day). I pulled my system apart, checked all the connections, cleaned all the dust off everything (there wasn't much) and then things seemed to be going alright.

The crisis:

Then I had a system reboot and an error pop up in BIOS: The 12V source on my PSU was only providing 9.4V. I powered the system off, unplugged the DVD-RW drive and the floppy drive and all the fans, left the side panel off and powered the system back on. BIOS then showed 11.94V. Power off, add the DVD-RW drive back to the PSU, check again: 11.90V. Power off, add the case fans back to the PSU, check again: 11.90V. Finally, power off, add the floppy drive back to the PSU, check again: 11.84V. Ran a pass of memtest86 like this, no problems.

Next step:

Here's where I'm stuck. I've been reluctant to keep pushing my system (posting this from my laptop), but I'm not sure how to track down the problem from here. The way I see things, the problem could be:
A) PSU is flaking out, but not doing it consistently
B) Floppy drive, DVD-RW drive, fans are shorting, but not doing it consistently
C) Motherboard isn't reading voltage correctly, but not doing it consistently

How can I go about isolating the component in my system that's responsible for this?

Bbq
02-04-2007, 07:00 PM
the bios is wrong. If it was at 9v, your computer wouldn't even work.

I'm guessing your issue has to do with memory. Memtest86+ it, maybe set it to 2T command rate, up the voltage a bit.

CestusGW
02-04-2007, 11:13 PM
I ran memtest86 two nights back for ~8 hours, no problems there (it was the first place I looked). So the problem doesn't lie there. I guess the next step I'll take is to run some SuperPI and 3DMark stress tests to see if anything wonky pops up doing that.