GPU crashing randomly. Can't seem to find why

[da na]

A new old stock GTX 670 I recently bought has been crashing randomly the past few weeks. It's never really been used for 3D gaming, bought it to use in systems I was benchmarking.

After 5-20 minutes it will crash to a black, white, or gray screen. It's different every time but always a total system lockup. Could be mid benchmark, or could be 0% CPU/GPU load sitting on the desktop. It's seemingly random. 

I've tried in 4 different PCs: two different Xeons in C600 and X79 boards, a Q9650 in an Optiplex, and a Q9550 in an Asus board. Same symptoms on each.

Logs show it's not really caused by anything specific. No consistent data between each. Is it just a hardware failure? Seems strange that it would fail after a few months. 

(I've been properly plugging in power connectors. Displays are properly connected. Same thing on Windows XP, 7, and 10.)

Thanks in advance!

[BiotechBen]

Are you running old drivers? They might fare better given the age of the car.

[da na]

3 minutes ago, BiotechBen said:

Are you running old drivers? They might fare better given the age of the car.

I'm running the latest one. This didn't used to happen, perhaps I'll try an older version. 

[Hybris5112]

2 minutes ago, Mel0nMan said:

I'm running the latest one. This didn't used to happen, perhaps I'll try an older version. 

I just checked Nvidia's site and it does show under the 670 the latest drivers from January as being available. I would do a close visual inspection of the card and look for any sort of corrosion, burned spots, popped caps, etc. If you are up to it and the card is worth it to you you could take the cooler off and get a even better look and maybe do a repaste while you are at it.

[DoovidToonet]

I had a similar issue happen with my old RX 570. I know NVIDIA is different than AMD, but my solution might work (though it was just a minor patch.) I went in and underclocked the GPU by about 30%, and that seemed to add enough stability to do basic web browsing tasks and play most games, though some games would still have it lock up and I could never figure out why,

[TexasBulldog74]

Like mentioned before, Id roll back the driver to previous version and give it a try.  Newer doesn't mean better sometimes.

[Chiyawa]

A few things might be the culprit:

 

1st, temperature. Since it is a new old stock, it might be possible that the thermal paste or thermal compound has been hardened and no longer be able to transfer heat effectively.

2nd: VRAM issue. it could be VRAM is not in good condition, or it could be VRAM overheat as well.

3rd: GPU chip issue. It could be your unit is defective, but it is likely that the heat is the culprit.

[c00face]

8 hours ago, Mel0nMan said:

A new old stock GTX 670 I recently bought has been crashing randomly the past few weeks. It's never really been used for 3D gaming, bought it to use in systems I was benchmarking.

After 5-20 minutes it will crash to a black, white, or gray screen. It's different every time but always a total system lockup. Could be mid benchmark, or could be 0% CPU/GPU load sitting on the desktop. It's seemingly random. 

I've tried in 4 different PCs: two different Xeons in C600 and X79 boards, a Q9650 in an Optiplex, and a Q9550 in an Asus board. Same symptoms on each.

Logs show it's not really caused by anything specific. No consistent data between each. Is it just a hardware failure? Seems strange that it would fail after a few months. 

(I've been properly plugging in power connectors. Displays are properly connected. Same thing on Windows XP, 7, and 10.)

Thanks in advance!

this is likely a driver issue. it's just an old card, and i can bet my money that nVidia doesn't care too much about their older models and making it combability with all the updates of Windows or Mac.