Sign in with
Sign up | Sign in
Your question
Solved

GPU Temp 3 (VRAM?) --> 120C

Tags:
  • GPUs
  • Thermal Compound
  • Graphics
Last response: in Graphics & Displays
Share
September 11, 2014 9:49:34 PM

Hey all,

Attempting to fix extremely hot VRM temps (-->125C) on my decrepit HD5870 I popped the heat sink off and put some thermal paste on.. everything.

This seems to have fixed the VRM issue, but now GPU temp #3 which I think is the VRAM is approaching 120C under stress testing. Probably because thermal paste was a bad idea.
http://imgur.com/zcDyBgE
(Ignore the fan speed. Its not being controlled, because my current cooling solution looks like this http://imgur.com/zXjStgz)

My question is, how long is a card at these temperatures likely to last? Can I expect to get a few weeks out of it until Nvidias 900 series potentially arrives, or should I order some actual VRAM pads to fix my mistake?

By the way, what prompted this at first was the VRM portion of the board getting uncomfortably/painfully hot during memtest86 running. Does anybody know if it is normal for memtest86 to to stress/heat the gpu in this way? It became far hotter then it was when idling on desktop.

Thanks!

More about : gpu temp vram 120c

September 11, 2014 9:51:28 PM

Sounds like you might have missed some thermal pads on the RAM.
m
0
l
September 11, 2014 9:54:16 PM

I didn't have any thermal pads lying around; I just left the beaten up ones that were on there, and added some thermalright white goop thermal paste on top of the VRAM to fill in any gaps. (VRM and GPU I used arctic silver).

What is the shelf life of thermal paste? This stuff has been in the tube for >4 years, could it be no longer thermally conducting properly?
m
0
l
Related resources

Best solution

September 11, 2014 9:57:26 PM

Thermal grease sometimes dissolves thermal pads. I would get some new ones...

http://www.performance-pcs.com/catalog/index.php?main_p...
Share
September 11, 2014 10:00:33 PM

Thanks for the link.

Any idea on the longevity of the card if I don't replace the thermal pads? Am I looking at likely days... or likely weeks?
m
0
l
September 11, 2014 10:29:33 PM

It's not a good idea to test with Furmark/Kombuster, these programs stress the gpu way beyond any gaming load.
Use Unigine Valley for a stress test as it better simulates a gaming load.
If you are using this card for crypo mining or something i would avoid doing that until you get new thermal pads.

That ghetto cooling, i love it.
m
0
l
September 13, 2014 9:07:12 AM

Thanks for the tip Shady. I got around to trying that test (Unigine Heavan though, not valley. I assume its equivalent).
Scored 614 ( 24.4FPS ) on 1080p, ultra, 2x AA, normal Tessellation.

GPU temp never budged above 75C, VRAM temp stayed below 80, and VRM temps didn't touch 60C. I think I am safe for gaming!

It is possible there was some burn in required for the thermal paste as well, as I have re-tested furmark, and my GPU is taking longer to throttle, throttling less, and the VRAM temperatures (while still absurd) are 5 degrees lower.
m
0
l
September 13, 2014 6:06:48 PM

The only thermal paste, that i know of, that has a burn-in time is Arctic Silver 5.
Burn in time is 50-200hours depending on use.
m
0
l
!