Catastrophic crashing on old PC EDIT: Overheating?

[Eightball]

God damnit!

Gonna preface this by saying that I am absurdly frustrated at my old PC. As long as I owned it I had trouble with it, and now I've got a new throng of bullshit now that I've rejiggered it and given it to Deb. If anyone can help suggest a test or a cure, I'd really appreciate it, as I currently have a $2000 crash-box.

Here's some background:

The computer is a 3-4 year old rig; i7-920, EVGA x58 motherboard, 3x2GB DDR3-1600, and 2x GTX 460s. I had it running for a few years, then decided to do a clean install of windows to give it to Deb. I used the upgrade disc that I originally had, and my license key was accepted. Installed all of our essential programs, then noticed a problem when we tried running Resident Evil 5. Here is that problem:

After a few minutes of play, the screen would black out, the keyboard would die (backlight goes dead), and the computer appears to stop responding to all input. This happens consistently. We thought it was an overheating issue at first; the computer was in an enclosure with poor ventilation. We moved it, and saw marginal improvements, but the same problem persisted. It happens rarely in other games as well, not just Resi5. I even went so far as to reinstall win7 a second time to fix this, but it still happens. Does anyone have a clue what this could be? I tried using different graphic drivers, but to no avail. I'm going to run some stress tests and monitor the temperature to see if that's really the problem.

More information: I have validate the cache and know that it's not a problem with the game. I will be trying without SLI enabled in a few minutes to rule out that possibility.
EDIT: Wow. Temperature test hit 103/93 CELSIUS before the crash happens. This sure as hell sounds like a temperature problem, and that the crash is the card killing itself before it gets hot enough to cause permanent damage...but this was run with fans full on and in a well ventilated area. How can I stop this, and why the hell didn't it happen before, when I was using it?

EDIT2: Opened one of the GPUs to see if it was clogged with dust. It wasn't, but its thermal paste was all fucked up. I think that the last time I opened it up, I ruined the layer of thermal paste, but didn't fix it. This would explain why I had crashing after the reinstall, but not before (I cleaned the cards around when I constructed Victor). Going to reapply thermal paste.

[Versus]

gl

[Eightball]

Replaced the thermal paste for both cards; running a torture test on one alone and it's holding steady at 98/99C. I guess all these months of me saying the temperature reading was wrong, I was wrong. Still, this is a very narrow margin. I might have to look into underclocking the cards. Is that a thing?

[Mission Difficult]

Replaced the thermal paste for both cards; running a torture test on one alone and it's holding steady at 98/99C. I guess all these months of me saying the temperature reading was wrong, I was wrong. Still, this is a very narrow margin. I might have to look into underclocking the cards. Is that a thing?

Yeah. I had a GTX260 that would shit itself on pretty much any source game for some reason. Used rivatuner and underclocked it around 100MHz and it was fine.

[Eightball]

[quote author=Eightball link=topic=6708.msg259813#msg259813 date=1357510301]
Replaced the thermal paste for both cards; running a torture test on one alone and it's holding steady at 98/99C. I guess all these months of me saying the temperature reading was wrong, I was wrong. Still, this is a very narrow margin. I might have to look into underclocking the cards. Is that a thing?

Yeah. I had a GTX260 that would shit itself on pretty much any source game for some reason. Used rivatuner and underclocked it around 100MHz and it was fine.
[/quote]

That's funny, I had a 260 myself before these 460s. I'm looking at a guide for Rivatuner now; if the torture test with SLI crashes me (or puts me above 100C, as 105C seems to be the limit), I will underclock. I wish I had taken a photo though, I'm almost certain what happened was me destroying the seal of thermal paste when I last tried to clean out the cards. Luckily I had some paste that came with my 212 EVO, I assume a heat block is a heat block. +1

[Eightball]

Well, Rivatuner wasn't working properly; missing dialogs and context menus :\. Did the underclock with EVGA PrecisionX though, dropped GPU clock from 675 to 575, and Memory clock from 1.8Ghz to 1.6Ghz. However, the changes didn't save after a restart, so I hope this isn't software that I have to have running all the time. Maybe by the time the thermal paste cures I won't have to bother with it anyways.

SLI survived OCCT stress test for 10min; one card maxed at 99C, the other at 91. Survivable, I guess, but still not ideal. Setting this to rest for now (until I see another crash).

[HeK]

That is still really fucking hot.

Dunno why it's running so hot..

[Eightball]

That is still really fucking hot.

Dunno why it's running so hot..

It got to that point entirely on its own; when I opened up MSI Afterburner (true story, EVGA Precision amounts to a reskin of afterburner), my GPU and memory clocks were still at the same as the 460 reference card. And it started pulling 90+ before I ever opened up the cards and potentially broke the thermal seal. Furthermore, it's in a full-tower case with huge (200mm+) intake fans, and I have cleared them of dust. So, no idea.

It's pretty funny when you consider that for maybe a year of gaming, I was a few degrees away from crashing. The whole time, lol.

EDIT: Let this also be a lesson for you system builders: dual cards can be the difference in heat production to ruin your day. Stick to one card!

[Evil Cheese]

What sort of PSU are you running in that box? Have you checked the voltages? Just a random hunch.

[Eightball]

What sort of PSU are you running in that box? Have you checked the voltages? Just a random hunch.

Cooler Master 1000W Silent Pro....definitely overkill, but to the best of my knowledge the voltages are fine. It may be worth mentioning that I dropped the voltage in Afterburner from .975V to .925V.

We got through a proper round of gaming and only hit 97C in Resi5. So, it's workable. When the paste cures hopefully that'll drop to 92 or lower.