Ok so I have 2 GTX 980's in SLI that have been working great for over a year now; both bought used. Unfortunately since last week one of the GPU's started showing absurd temperatures. Like 58C when idle (with fans @ 100%) when it usually never went above 41C at idle (when the fans were literally off). When I launch a game; even if it is just in the main menu the temperature goes up to 95C, the fan goes to full blast but oddly enough it does not thermal throttle. Games still run perfect without any issues. Now I would have thought that the thermal paste had gone bad but this change happened literally in terms of minutes. I was playing Overwatch for quite some time just fine and all of a sudden this issue came out of nowhere in the middle of a game. I tried using a fresh copy of windows, switching the PCI lane and flashing the motherboard BIOS but nothing seemed to help. I reached out to EVGA who agreed to have it sent for RMA but I just wanna check if I can fix it myself instead of going through the trouble of sending it and all that. I believe the temperature sensor on the card simply went bad so I don't think I'm going to be able to fix it; but I may be wrong... Any ideas of past experiences are greatly appreciated.
 
Thanks

But do you think it's possible that the paste goes bad in literally minutes ? I'd have though it would degrade over time not have an immediate change :/

It's a long shot, but it would never hurt to also wipe all drivers with Display Driver Uninstaller, and also remove whichever monitoring software you are using(Afterburner or Precision I assume) and be sure to remove user settings as well when prompted. I've dealt with some very odd bugs with both Afterburner & Precision that were fixed by re-installing the program. But l agree with Sajin, re-thermal paste it, and see where that get's you. Perhaps somehow your screws backed out a little overtime causing the core of the GPU itself & the heatsink to not be making good contact? especially given they are used cards, who knows if the prior owner had taken the shroud and heatsink off, and didn't properly torque down the screws when putting it all back together. Other then that it's very possible the temp sensor has/is failing, just haven't really seen much of that since the ABIT Motherboard days lol. At least if you end up having to RMA, you still have a 980 to hold you over 

Time to send it in.

Looks like a bug in precision.

Wow.

Try going back to a different driver version after cleaning out your old drivers with ddu.

Just finished installing an older driver (30th November) and the same issue is present though Idle temps are far better at 38C. Though I did notice that the GPU clock and memory clock are now far lower on idle than they were before; but whenever I launch furmark or a game it rushes up to 95.

Just installed 378.66 and the same issue keeps persisting. Idle temps are now great but load temps still reach 95C and ramp the fans up to 100%...

No it is not hot at all when reporting the 95C temperature. It's just as hot as the other card. 

never know if the card heats up it may start top warp  , then with that  contact between the  cooler and the card may ''break''   and heating issues occur ??
 
like that top card in sli  heat rises  and it takes some heat from the lower card on top of its own   and the bottom card can help block air flow to the top card to add to that ??
 
 
page 1 bottom pic
 
 http://www.legitreviews.com/sapphire-radeon-hd-7950-3gb-oc-edition-video-card-review_1862
 
or
 
https://linustechtips.com/main/topic/25577-warp-pcbgpu-overheat/
 
may try to check yours when cold and after it heated up to see ??
 
sometimes card sagging may cause this  ?
 
 
I got stuff every bit  10 years ols in use toady and never had thermal paste on them fail unless I break it loose    once you apply  correctly and set it its there for the long haul  and should not ever need to be replaced 
 
like here
 
'' compound thins out to enhance the filling of the microscopic valleys and ensure the best physical contact between the heatsink and the CPU core. Then the compound thickens slightly over the next 50 to 200 hours of  use to its final consistency designed for long-term stability''
 
[set and forget]
 
 http://www.arcticsilver.com/as5.htm

you know  could be the cards fans ?/      maybe there misreading there speeds and just really not spinning up  to properly cool the card  ? I've seen that  before  where the reported speed was showing 80-100% and the fans were looking like they were maybe 20%  at best ??  I took a fan from a known working old card and tried it in there place and all was back to right   , speeds temps and all  ???
 
I guess you need a good gpu fan in a junk box to test that theory  out  . 
 
 
well that evga bottom card looks warped from your pics???  looks like it starts to ride high at the psu  plugs area
 
funny is that gigabyte cards had out of the box warped issues  a lot
 
http://www.tomshardware.com/forum/id-2353388/solved-graphics-card-heatsink-slightly-bent-real-problem-gtx-980.html

well still could be a matter of bad fans ??    hard to get replacement evga fans   maybe find some on e-bay   and then there not cheap  there
 
maybe best to shrug your shoulders  and try a RMA  with evga   with that just cut the chase  get a working good card back in replacement 
 
theres only so much you can try and do   if you don't have stuff handy around the house  to test and try  with  

ya, if they are willing to work it out  maybe best  in the end   you atleast get a tested card back that should work as expected   
 
thing is  if you get back another  980 ??    it maybe one of them next best card things now a days on that   if there are no 980's  turned in for refurbishing  .
 
they don't show any 900 series at all
 
https://www.evga.com/products/productlist.aspx?type=0
 
being you use another brad card for your SLI  you cant claim that I assume   to keep sli  on that rma  ??
 
I guess that something you have to discuss with the evga rep your dealing with on your rma
 
 

no extra pay ,  it will be like the next best or as equal to 10 series    what ever 10 series performance wise is the same as that 980 you got  .  then you never know tthey may have a few 980's on the shelve  just for RMA use ??   you should  get a contact on that  on what they decided to do ahead of the return