What's the culprit?

[PsharkJF]

Copy/pasted from another forum, figure more eyes looking at it would help!

Generalities:
Core2Duo 2.66/1333 FSB
Asus P5N-E SLI
PNY GeForce 7900GS (256meg) PCI-E (Box says it requires 20A on +12v line)
500w PSU from AGI: States max of 18a on +12v1 and 17a on +12v2. Strange.

My new stuff's being 'randomly unstable', just started today:
Main symptoms:
1) Screen goes black, as if just pulling out monitor cable
2) Some disk access, but usually none.
3) Tower still has power but does not restart nor does it respond to three-finger salute

Things I've done to fix it:
1) Did a chkdsk. Did not find any errors in read-only mode.
2) Checked cabling. Reseated the four-pin power molex that goes next to the traditional 20-pin ATX power molex. All good. No fix.
3) Checked heat on CPU, chipset, video. All warm but not hot. No fix. (Later installed an Asus program that showed me chipset/CPU temps and both were at 40 deg C. Vcore voltage was 1.28, it said optimal was 1.31.. something inside of me says if that .03 V mattered, it'd be more unstable, earlier!)
4) I had recently installed new nVidia video drivers. Rolled them back, and at the restart prompt, it 'unstabled' itself but seemingly I was able to restart the machine by clicking and hitting enter. (Hinting that the prompt was still 'on screen' and I had just hit 'ok'...)
5) Reseated the graphics card in the PCI-E slot but did NOT reseat the 6-pin add'l power connector.
5a) Completed one full test of Memtest86 with no errors.
5b) My 7900GS has two digital video outs. Moved the monitor cable and VGA/digital converter to the other video out.

6) The event viewer logs all these random hangs as regular shutdowns.
7) Multimeter readings, idle:
(20+4 pin main power):
+3.3v - 3.37-3.38v
+5v - 5.09-5.10v
+12v - 12.15v
Load:
+3.3v - 3.38v
+5v - 5.09v
+12v - 12.14-12.18v

8) I ran memtest all night. It crashed sometime in the middle of the night, but the very odd thing about that is when I hit escape, the system went right up and booted, no problems. It's as if Memtest was still running in the background but the graphics card had shut itself off...
If it was the motherboard, you'd think the system wouldn't be responsive to keyboard commands and Memtest would cease to run. If it was the PSU, you'd think the voltages would fluctuate more under load.

So which is it? Graphics card failing, motherboard failure, or crappy PSU?

 

[NoelS]

PsharkJF,

Well, seems like you've done quite a bit of testing... hard to say which it could be, but at this point the best route might be to first use (or maybe borrow) a different graphics card to see if that's the culprit. If it's OK, you can buy a PSU tester for about $20 at Radio Shack or on-line. That may be cheaper than just going out and buying a new PSU right off the bat...

If both tries fail, then call Asus and RMA the mobo - if you have to do that, ask for a cross-ship. Much faster than sending your board, them testing, repairing and returning it (cross-ship requires you give your credit card, they won't charge it unless you don't send your old board back with 10 days)

Hope something here works,

Noel.

 

[Err0r404]

1. check your bios, switch around the power off options, and sleep options. Sometime there is a option that say "suspend to RAM" etc. try different setting if you already have that.
2. Vista? or XP? check microsoft kb if you are running vista for updates/hot patches. especially SLI patch and sleep issues.
3. Able to run your computer fine while u playing games? does this problem occure only when it's idle?
4. not enough power or power spike/dip = graphic card going off...however it doesn't make sense and easy to figure out, just swap PSU.
5. Most likely your computer going into sleep mode? and can't recover from it....or similar power profile issues. > Slove this by disabling sleep mode, and leave it as POWER USER, never shutdown/standby. See if this goes away.
6. Not sure about the power consumption on a SLI board, is this machine overclocked? check heat issues plz.
7. confirm this isn't monitor problem. (just process of elimination)

 

[PsharkJF]

UPDATE:
Got in touch with a friend of mine who had a spare (460w) PSU and spare 6600GT (PCI-e video card) and he's letting me borrow them to test with.
Last night, used his PSU with my 7900GS and it failed after approximately five hours. After that, ran Memtest with the same hardware until this morning. Completed 16 error-free tests over 9:10.
As of right now, I'm using my PSU with his 6600GT and no crashes as of yet.
I'd put it at 97% chance of a bad 7900GS. [Methinks it's a bad VGA BIOS.] I'm still under warranty, but since I lost the recipt I have to wait '12 days' for Fry's to mail me the recipt of my purchase again, then another two weeks according to PNY for the actual RMA. D'oh ;[

EDIT:
Noel:
My problem was I had nothing to test with. My brother's machine still uses AGP and a 20 pin PSU - hence, I didn't have a spare 20+4 pin PSU and a PCI-e video card with which to play around with - heck, makes me wish we had still kept the Voodoo3 2000 PCI
I had three components which could be the problem and no way to prove which one it was - and I really didn't wanna spend any money on new components without conclusively enough proving that I had pinned down the problem component.

Err0r:
1) Double-checked the BIOS options, there weren't any power-conservation options enabled that popped out at me.
2) XP box, but it fails both inside and outside of XP. The incomplete list of 'failure locations': pre-POST, Memtest, XP boot screen, XP safe mode, XP login screen, idle in XP, in a game in XP.
3) See above, but it's not consistent in where it fails.
4) Switching out the PSU did not fix it.
5) Again, nothing in the BIOS pops out at me... and considering it fails outside of XP as well, it wouldn't be anything in Windows that's causing the problem.
6) Not overclocked. Heat isn't an issue as it will fail when idle as well as under load.
7) Not a monitor problem as I'm able to display with my friend's card no problem.

 

[NoelS]

PsharkJF,

Not having spares to test with can always be a problem. I'm glad you seem to have found the problem. I hope that continues to work.

Frequently when you don't get any beep, the GPU can be a good suspect... But, unfortunately, not always

Noel