Soak testing a PC

[kernow]

So, I built a pc for my flatmate probably mm about 15 months ago, its had a new power supply since then as the cheapo case one blew. It was absolutely rock solid for a year, no problems at all. I'd probably built 3 of these pc's the same spec for various people based on:

Core 2 duo 3GHz
Asus P5K/R motherboard
Random Nvidia card whatever they can afford (GTX260 this time)
2/4GB RAM
WD 500GB HD

So erm as I said it was running fine for a year, just like the other two I'd built. The last say 5 months its been all over the place, blue screening, etc. I've formatted it since then in the past and it ran ok for a while, but then started with the instability once more.

I've run memtest on it before but as he's away this week I said I'd give it a proper test. I've currently done:

*12hr memtestx86 - no errors
*WD hard disk diagnnostics - no errors after performing extensive test

I guess if anyone has any suggestions for good nvidia gpu testing tools, or CPU burn in or whatever let me know. I don't think its overheating because sometimes it'll BSOD in the most lightweight usage situations. I haven't recorded the error code yet but I should have done by now really. Sometimes I guess they are so nondescript anyway.

I'm pretty sure its not the memory now, doubt its the hard disk although he said he did install something the other day, turned on the next day and it wasn't there anymore but that could have been a user error or something.

I've updated the drivers, bios, etc. It keeps losing its dual monitor setting sometimes also, the other day the whole nvidia control panel tools vanished and had to be reinstalled, its got an LG 22" for a main display and he's taking HDMI to a 32" generic LCD TV for watching movies etc.

Any ideas? I plan to test the CPU next probably with a 12hr prime95 or something I dunno, maybe install some temp monitors. I have hirens bootCD on a USB key I'm just booting off at present and trying various things.

Its certainly not a windows install issue. Although it is running on SATA and I remember that can sometimes be flaky on XP depending on chipset support.

What about testing the GPU ? 3DMark or something, and then just compare it to other machines of a similar spec? Its been literally 10yrs since I've used that.

Let me know what steps/tools you use to diagnose a blue screening PC.

Cheers

[speedlolita]

I google the error code and that usually solves the problem. :P

[kernow]

Yeah, I did that before in the past. Its passed the memory check and now hard disk too. I guess the next step is a GPU or CPU check of some kind. blech.

[speedlolita]

Heh, random blue screens are never fun.

Worst though I've found is no video output, always a bitch to diagnose.

Furmark would do to test the GPU.

Is the PSU up to the GPU? The GTX260 requires quite a beefy one - obviously from a reputable brand.

[kernow]

I'll have to check the new PSU actually, last one was a 550w but I am pretty sure this is a 700, not the best brand however. bugger I hope its not something like that. It has both power connectors plugged in from PSU to gfx card.

Thing is why would it have been fine for a year beforehand.

[abigsmurf]

Power supplies lose capacity over time, if it was only just supplying enough power it could easily cause the system to become unstable later on

[kernow]

Well I've been reading and Nvidia recommends a 500w PSU with 36amps on the 12 volt rail. The old one was 550, and this new one I'm pretty sure is a 700, and it was still BSOD'ing.

I'm going to run a cpu/gpu burn in today. After that, memory done, hd done, whats really left?

[abigsmurf]

What's the error? They're usually a lot more telling than they appear.

Also, get speedfan and look up your CPU/GPU temps. Hot weather could be pushing the system over the edge.

[kernow]

Yeah I'll get some temp monitoring software on there. To be honest he's usually rebooted it and then comes down and tells me it's bsod'd again. I might just leave it on today and see whats happening. Don't think its the hot weather because it's been doing it since January pretty much.

I think its too hot today to be attempting burn in tests so I might start tonight instead. Will give speedfan a go though cheers, and if it blue screens on me personally I'll make sure to record the error this time.

[Daragon]

It still works after you've submerged it in water? I'm not sure PCs are like watches in that respect mate

*Ba dum tsh*

I'll fetch my coat shall I?

[Hohum]

Check the voltage readings in BIOS perhaps, although I doubt the PSU would be failing already unless it's a garbage unit. Other than that, try the following apps:

http://hcidesign.com/memtest/ < more stressful than memtest86+, so more likely to find any RAM fault. Runs under windows so don't allocate too much RAM for it to check unless you want to thrash the hard drive to pieces

http://www.mersenne.org/freesoft/ < prime95 - run the stress test built into that

[kernow]

will do

[kernow]

I've run furmark on it for a bit before I got bored, got the GPU to 101C, was fine. was averaging 68-80ish FPS.

Its now running prime95 for 30mins and obviously the CPU temp is rising but its doing fine also. The various voltages look fine.

I can't get this thing to bsod at all, I dont know how he manages it

I know I'm not running the tests for very long, but still.

Turned off 'automatically reboot' so next time if it does blue screen I'll get it noted down and search for the code.

[kernow]

I'm honestly puzzled, I could leave the tests running longer but a 12hr memtest, full hd scan, and a bit of a burn in, I can't get it to crash. He said he's just been on the desktop before, or ripping a DVD and its bsod'd.

Puzzles me especially as it was like a rock for a year, and every time I use it, its fine also.

[Hohum]

What O/S?