Hard crash during 3D games

[vainglorious11]

When I play Mass Effect and Saints' Row 2 my system inevitably crashes after a while. After 15 mins to several hours of play the screen fills with colored vertical pinstripes and the audio buzzes. I never have problems during non-3D use and older 3D games like L4D are fine too. I've tried a lot of different things to fix the problem and nothing seems to help.

My system - everything is dialed back to stock timings:
GPU: XFX Radeon 4850 512MB
CPU: Phenom II X3 720BE
Mobo: ASUS M4A78T-E 790GX
RAM: 2x2GB OCZ Reaper DDR3 1600 @ 7-7-7-24
HD: 1TB Seagate Barracuda 7200.12
PSU: Antec TruePower Trio 650W (TP3-650)
OS: Windows 7 Release Candidate 64-bit (build 7100)

Things I've tried:
-Cleared GPU driver with DriverSweeper and installed latest copy - no change
-Ran each stick of RAM on its own - still crashed
-Logged temps during gaming with Everest - nothing's overheating
-Stress-tested with Everest Stability Test+GPUTool, Prime95, FurMark, 3DMark - no crashes during stress tests
-Ran MemTest, Windows Memory Diagnostic, several hard drive checks - everything passed
EDIT:-relaxed the RAM timings to 8-8-8-24 - no difference
EDIT:-relaxed RAM speed to 1333 MHz - no difference
EDIT:-logged voltages during stress test - fluctuation was pretty minimal

I suspect that my PSU, RAM or GPU are crapping out on me, but I don't know how to isolate the problem apart from what I've already done. The only thing I know I haven't tried is switching out components, because I don't have any spare parts that I can use.

 

[EXT64]

As the calibration of the voltage sensors varies, I wouldn't worry too much about the value, but rather how much it varies from no load. If you want to check the ram, try reducing the speed/timings and see if it still crashes. You can also try underclocking the GPU a little and see if that helps. Is anything overclocked? I'd try running the ram at 1333MHz.

 

[vainglorious11]

Thanks EXT,
My 12V shows a voltage of 12.1 with no load - how much variation is acceptable?
My GPU is at stock (650MHz?), which is as low as it will go using the Catalyst tuning utility.
I did try relaxing the RAM timings to 8-8-8-24, but I didn't think to lower the speed - I'll try running it at 1333MHz as soon as I get a chance.
Nothing is overclocked atm, I set everything back to stock when I started troubleshooting this.

 

[jsc]

PSU voltage tolerance is +/- 5%. So 11.4 - 12.6 volts is good.

 

[EXT64]

Yep, that is most likely ok. I have seen worse on some of my systems and they have not crashed.

 

[vainglorious11]

OK I tried relaxing my RAM speed to 1333 MHz, but it still crashed.

 

[vainglorious11]

Any other suggestions to pinpoint the problem?

 

[JackNaylorPE]

Run OCCT

http://www.ocbase.com/perestroika_en/index.php

Look at voltage graphs when done.

can also run the OCCT PSU torture test but be warned .... it will cause lesser PSU's to melt down.

 

[vainglorious11]

will try that, thanks

 

[vainglorious11]

Hmm, I ran the OCCT Linpack test for an hour and the 12V only fluctuated .08V the whole time. (12.05V-12.12)
Looking back over my Everest logs it actually only got as low as 11.92, not sure why I thought it was 11.8.

So if low fluctuation is a sign of a healthy PSU, mine is doing pretty good. It's a couple years old but it is an Antec.

 

[vainglorious11]

Other results in case they're useful:
3.3V fluctuation: .01 (3.44-3.45)
5V fluctuation: .03 (5-5.03)
VCore fluctuation: .03V (1.32-1.35)
CPU max temp: 56C

Looks pretty healthy to me. Once I figure this issue out I think i'll do some OC'ing

 

[vainglorious11]

Anybody else have ideas?

 

[vainglorious11]

Best answer selected by vainglorious11.