Re-paste my 2080Ti, idle and load temp significantly down

[xkm1948]

So my 2080Ti has been overclocked mildly most of the time. As it is approaching 2yrs old I decided to check the thermal paste as I have been seeing some weird fan behavior. Basically when core temp gets to around 80C under load the fan will suddenly spike to 100% no matter what fan curve I set it to.

Some previous discussion here:

Resolved - Fan 1 Sudden High RPM at 80C - RTX 2080 Ti XC - EVGA Forums

I have the EVGA RTX 2080 Ti XC and am having a weird issue where Fan 1 will suddenly go to 4100+ RPM. It seems a little random but happens around 80C. I have tried setting my fans to a fixed % and Fan 1 will still raise to 4100+ RPM. I've attached a screenshot of Precisio...

forums.evga.com

Finally had enough of this so I tear it down. And this is what I find:

So yeah, quite a large portion of the die is not actually getting much TIM contact. No wonder it is doing that.

So wiped off all the thermal paste and put some fresh Thermalgrizzly Kyronaut on there. Cool and quiet again. Exact same overlock, same room temperature and same fan profile. I used to get 82~83C under full load in super position 1080p extreme. Now I am only getting 73C under full load. Most importantly, the noise level is waaaay done.


Cannot say I am happy with EVGA's TIM application. But oh well.

 

[nguyen]

That looks to be the worst stock TIM application LOL. I bet you also experienced lots of stuttering in-game, you just didn't notice them. It won't affect benchmark score but if you actively look for 1% low fps you can easily spot them.

 

[xkm1948]

That looks to be the worst stock TIM application LOL. I bet you also experienced lots of stuttering in-game, you just didn't notice them. It won't affect benchmark score but if you actively look for 1% low fps you can easily spot them.

Not really. Most of gaming sessions are VR so if there are ANY stuttering I would instantly vomit to death.

I was running a very close to stock OC, +10MHz on Core and +100 on VRAM. I need to use it for lots of CUDA work as well so cannot bear to have it overclocked too much.

 

[looniam]

lol, yeah its a good idea to have a big bottle of 91% ISO for re-pasting an evga card.

 

[micropage7]

clean it to the edge, don't leave any old paste and residue there

 

[Sloeri007]

Bit of an old thread already, but had exactly the same with my Aorus 2080 Ti....
After 30 secs of playing anything, fans ramped up to 4000+ rpm (fail safe mode)
So for some period i undervolted the card, to keep it within normal temperature's. (0.825 volt for 1500Mhz....)
not really you want to do with a 1400€ card. but at least it stayed below 75°C that way.

i even was in contact with gigabyte, had send logs and a movie, they said it was "normal"....

so decided to take the plunge, and repaste it myself.

i was baffled when i opened it up, there was nearly no thermal paste and it was bone dry. (the GPU was covered properly though, have to give em that)

redid it with Grizzly Kryonaut.
Result : 12 degrees lower temps, and clock is now stable at 1860 at approx 68°C with fans at 85%

Back a happy gamer now.....

 

[xorbe]

Yeah I *just* finished this same thing minutes ago. My 2080Ti FE looked just like OP's pic, but even less paste!!! All towards one corner. Fans were going hard core every 5-7 seconds for 2-3 seconds. I never even overclocked this gpu, it's had an easy life. It also made it obvious that my case fans were set up incorrectly, so I fixed that too.

 

[Deleted member 193596]

Here is my 2080 Ti AMP Extreme (1650€ MSRP) Application....

repasting with Kryonaut dropped temps by almost 20°C.. (81° to 63°C)

 

[biffzinker]

My MSI RTX 2060 had a similar thermal paste spread, the only difference was just a tiny bit of the outer die wasn’t covered. Looked better than @xkm1948 photo shows.

 

[maxfly]

I repaste anything i can can get my grubbies under. Laptops, gpus, mbs its the second thing i do. After making sure they work of course. Ive seen some beauts over the years.

 

[biffzinker]

I forgot to mention re-pasting the GPU dropped the temperature down into the low 60-70C, and the sudden fan spin-up fluctuations to 100% i was having stopped. Used Arctic MX-4

 

[xkm1948]

My MSI RTX 2060 had a similar thermal paste spread, the only difference was just a tiny bit of the outer die wasn’t covered. Looked better than @xkm1948 photo shows.

I also had the sudden spin up to 100% under load, especially during workload that pull on CUDA and Tensor cores together.

 

[Cranky5150]

Clean and re apply would be the safe bet mate..paste is cheap, however application is not...

 

[kiriakost]

8 centimeter clearance distance from what even is near to VGA DC fans, this will deliver higher benefits than any thermal paste.

 

[xorbe]

My 2080Ti was never this quiet from the day I got it, it must've been pasted wrong the entire time.

 

[Vayra86]

Yeah... I need to go for it as well on my 1080. Just too lazy but its really becoming time.

I'm down to 90% power target now, which is actually also the best 24/7 setting for this particular one, but I see my clocks decrease as ambient goes up during summer, and it seems to be getting worse. A year back I held 1911 mhz at throttle target, now I get 1790 something mhz lol

Still amazes me how versatile and well GPU Boost works with all these changes in conditions.

 

[kayjay010101]

And that's why the first thing I do with a GPU is tear it apart, repaste, and put it back together. I've seen too many issues with stock applicated pastes, and it takes 15 minutes to do. Case in point, my 1070 has had massive issues the last 2 months after I sold my 2080 Ti. I never remembered it being bad at all, but when I got it I got constant crashes and thermal problems (idle at 50C, load at 85C), not to mention the incredibly annoying fans ramping up to 100%. Had to set a very aggressive fan curve to keep thermals in check, a -200MHz underclock on both VRAM and core, and a much reduced power limit to even get it to not crash under any form of load.

After a couple weeks I decided enough is enough and I opened it up to find a terrible paste spread. Fixed it up with some NT-H1 and bam, good as new. Actually surprised at how good the 1070 holds up 4 years on, what a great value card. I guess the paste was just good enough for the 3 years I used it, and then when it went into storage for a year the paste dried up and when used again it wasn't covering the die anymore. Or maybe it was always bad and I just never noticed because I had it undervolted for the longest time.

8 centimeter clearance distance from what even is near to VGA DC fans, this will deliver higher benefits than any thermal paste.

I'm sorry to seem mean here, but I have no idea what this sentence is supposed to mean and it's piqued my interest. Are you trying to say a good fan is better than thermal paste at thermal transfer?

 

[droopyRO]

Yeah... I need to go for it as well on my 1080. Just too lazy but its really becoming time.

I'm down to 90% power target now, which is actually also the best 24/7 setting for this particular one, but I see my clocks decrease as ambient goes up during summer, and it seems to be getting worse. A year back I held 1911 mhz at throttle target, now I get 1790 something mhz lol

Still amazes me how versatile and well GPU Boost works with all these changes in conditions.

Why don't you undervolt it ? My 1070Ti dropped about 5ºC with undervolt.

 

[Vayra86]

Why don't you undervolt it ? My 1070Ti dropped about 5ºC with undervolt.

That's a lower power target right? Not sure how else you would undervolt it?

 

[droopyRO]

That's a lower power target right? Not sure how else you would undervolt it?

Nope, check the latter part of this video.

Got mine to 0.931V with 1860Mhz, about 4-5ºC drop from 70º to 65-66ºC.

 

[edechamps]

I just had a similar issue with my 2-year-old ASUS ROG Strix GeForce RTX 2080 Ti OC: GPU hovering at 80°C at full load, memory temps through the roof (106°C), fan panicking and going 100% every few seconds. A Google search led me to this thread.

As suggested, I changed the thermal paste on the GPU and voila! Problem gone. GPU hovering near 75°C and memory temps around 90°C (even though I didn't touch the memory thermal pads!). Fan now operates at normal speed. So I just wanted to say thank you for the great suggestion!

 

[Deleted member 205776]

My GPU sits at ~65C on stock and ~61C undervolted. I think MSI did a good job with the TIM.

 

[Ferrum Master]

You are digging in the wrong way.

Looking at the fact that the thermal paste has spread past the corners and formed a rectangle, the paste was more than enough.

The real issue relies how straight that vapor chamber bottom is. I bet it is sunken in just like the paste spread out. The die gets unevenly pressured and the actual spots where the paste is missing, where the pressure points are.

 

[X71200]

A guy I know always rips apart cards and puts higher quality TIM and pads. It always drops down his temps, but I personally don't do it. New / good GPUs tend to not need it.

 

[Shrek]

That is so weird, one can see how the paste was evenly pushed out the edges but then withdrew.