Hotspot over 114°C

[cocoliche]

Hi everyone! Two weeks ago, I noticed a problem with my Sapphire AMD 7900 XT (MBA) graphics card. The problem was that after a long CS2 session, the hotspot reached a maximum of around 110°C, while the core stayed at 70°C. My GPU is 3 years old. I'm not sure if the issue has always been there, but to be fair, I bought it refurbished. If I had to guess, I’d say it didn't use to have this problem, because that would have caused thermal throttling which I should have noticed while gaming, and that wasn't happening.

So, I decided to repaste it. I bought Arctic MX-7 thermal paste and UPSIRON U6 PRO thermal putty. After doing the maintenance, running a stress test caused my hotspot temperatures to shoot up to around 114°C in less than 10 seconds, with the core hitting 86°C. I immediately stopped the stress test as a precaution, because if I had let it run, both the hotspot and core temperatures would probably have gone even higher.

At first, I would have said it was a mounting issue and that the heatsink wasn't making good contact with the PCB, since the whole unit was hitting very high temperatures and the symptoms got much worse. So, I opened the card several times to reduce the amount of thermal putty, eventually using just the bare minimum needed to squeeze the putty and cover the chips, but this didn't fix the problem. Every time I opened and closed the GPU, I made sure to tighten the X-bracket in a cross pattern (1, 3, 2, 4). I also checked for a proper thermal paste imprint when separating the heatsink from the die, and it was definitely there.

Many people recommended using PTM to prevent the pump-out effect, saying it could be ideal for what I was experiencing. I bought the PTM, it arrived today, I installed it, and the problems are exactly the same. I'm completely at a loss as to what to do next.

 

[pendragon1]

retry with the paste but use more on the die. if its still like that, it could be failed heatpipes... any chance you have a spare aio you could zip tie to it?

 

[SticKx911]

retry with the paste but use more on the die. if its still like that, it could be failed heatpipes... any chance you have a spare aio you could zip tie to it?

I'd also suspect the heat pipe problem. A small crack and the fluid vaporized or something like that.

 

[cocoliche]

retry with the paste but use more on the die. if its still like that, it could be failed heatpipes... any chance you have a spare aio you could zip tie to it?

Yes, I've already tried adding more thermal paste but it didn't fix it. No, unfortunately I don't have a spare AIO. If it were a heat pipe failure, shouldn't the heatsink be cold in certain areas and hot in others? Because it seems like the heat is spreading throughout the entire heatsink

 

[cocoliche]

Yes, I've already tried adding more thermal paste but it didn't fix it. No, unfortunately I don't have a spare AIO. If it were a heat pipe failure, shouldn't the heatsink be cold in certain areas and hot in others? Because it seems like the heat is spreading throughout the entire heatsink

Yes, but as I mentioned to pendragon1, If it were a heat pipe failure, shouldn't the heatsink be cold in some areas and hot in others? Because it seems like the heat is being transported throughout the entire heatsink

 

[pendragon1]

no, it wont be able to move the heat properly and eventually it will heatsoak the entire thing. got an extra tower cooler you can strap on it?
youre going to need to get creative, or move on to a new card.

 

[Dreamerbydesign]

Yes, but as I mentioned to pendragon1, If it were a heat pipe failure, shouldn't the heatsink be cold in some areas and hot in others? Because it seems like the heat is being transported throughout the entire heatsink

Actually a heat sink failure would be equally hot if the card was running for a few minutes. If the vapor chamber is dysfunctional and empty it will still heat soak the heatsink.

 

[hititnquitit]

How are your fans behaving? If you haven't already, eliminate your fan profile and crank them up to maximum speed and see what that does.
Are you ocing? If so, step everything back down to stock.
You can also try cleaning everything off of the heatsink and doing the flashlight test to see if your heatsink is making good contact. Point the flashlight through both the front and back side of the GPU/heatsink mating area(laid flat) and look for any unusual gaps where light shouldn't be coming through.
Can't think of anything else but maybe something above will trigger something.
GL!

 

[Starfalcon]

I have a 7900 XTX and mine gets no where near that hot. I would also check your adreneline settings also as it like to default to zero rpm mode and have the fans come on super late. Every time I update drivers I always turn zero rpm off and have the fans run all the time to keep the die cool.

 

[T4rd]

Sounds pretty much exactly like what happened to my son's old RX 570 a few years ago. One day it just started throttling on him while the fans were pretty much at full tilt. I just upgraded his card instead of wasting money on a new cooler for it since it had served him very well up until that point and could definitely use more power in some of the game's he was playing. I assumed the heat pipes leaked or failed somehow after trying to repaste it a couple times and ensuring solid contact with the GPU core.

 

[chameleoneel]

Hi everyone! Two weeks ago, I noticed a problem with my Sapphire AMD 7900 XT (MBA) graphics card. The problem was that after a long CS2 session, the hotspot reached a maximum of around 110°C, while the core stayed at 70°C. My GPU is 3 years old. I'm not sure if the issue has always been there, but to be fair, I bought it refurbished. If I had to guess, I’d say it didn't use to have this problem, because that would have caused thermal throttling which I should have noticed while gaming, and that wasn't happening.

So, I decided to repaste it. I bought Arctic MX-7 thermal paste and UPSIRON U6 PRO thermal putty. After doing the maintenance, running a stress test caused my hotspot temperatures to shoot up to around 114°C in less than 10 seconds, with the core hitting 86°C. I immediately stopped the stress test as a precaution, because if I had let it run, both the hotspot and core temperatures would probably have gone even higher.

At first, I would have said it was a mounting issue and that the heatsink wasn't making good contact with the PCB, since the whole unit was hitting very high temperatures and the symptoms got much worse. So, I opened the card several times to reduce the amount of thermal putty, eventually using just the bare minimum needed to squeeze the putty and cover the chips, but this didn't fix the problem. Every time I opened and closed the GPU, I made sure to tighten the X-bracket in a cross pattern (1, 3, 2, 4). I also checked for a proper thermal paste imprint when separating the heatsink from the die, and it was definitely there.

Many people recommended using PTM to prevent the pump-out effect, saying it could be ideal for what I was experiencing. I bought the PTM, it arrived today, I installed it, and the problems are exactly the same. I'm completely at a loss as to what to do next.

Which PTM did you get?

Generally, PTM takes 7 or more thermal cycles, before it will give optimal temps. And the first few uses, the temps will be pretty bad. I have used PTM many times. Several different brands. My favorite is Thermal Grizzly Phasesheet. But, it is a little bit expensive.
My favorite for price, is Joyjom:
https://www.amazon.com/gp/product/B0BX42N9SZ?smid=A1LSGV0IRJX84A&psc=1

Its good stuff. They have it in several different sizes/prices. I have Joyjom PTM in two laptops and a 4070 TI Super, for over two years, now.

Use your GPU a little bit each day (run a game at medium settings or something which isn't MAX LOAD), and report back in a week. Temps will hopefully be better, in a few days.


**you can also try Kryosheet
https://www.techpowerup.com/review/thermal-grizzly-kryosheet-amd-gpu/

Techpowerup used it on a 7900 XTX and found it performed really well.

 

[cocoliche]

Sounds pretty much exactly like what happened to my son's old RX 570 a few years ago. One day it just started throttling on him while the fans were pretty much at full tilt. I just upgraded his card instead of wasting money on a new cooler for it since it had served him very well up until that point and could definitely use more power in some of the game's he was playing. I assumed the heat pipes leaked or failed somehow after trying to repaste it a couple times and ensuring solid contact with the GPU core.

But the weird thing is that the problem wasn't this aggressive before. The GPU hotspot used to reach about 98°C during a 15-minute Deathmatch session in CS2. Then, in a competitive match or another game with a longer session, it would hit 110°C. But now it goes straight to 110°C immediately, literally within a minute it's already at that temperature. It's as if something has gotten worse.

 

[cocoliche]

Which PTM did you get?

Generally, PTM takes 7 or more thermal cycles, before it will give optimal temps. And the first few uses, the temps will be pretty bad. I have used PTM many times. Several different brands. My favorite is Thermal Grizzly Phasesheet. But, it is a little bit expensive.
My favorite for price, is Joyjom:
https://www.amazon.com/gp/product/B0BX42N9SZ?smid=A1LSGV0IRJX84A&psc=1

Its good stuff. They have it in several different sizes/prices. I have Joyjom PTM in two laptops and a 4070 TI Super, for over two years, now.

Use your GPU a little bit each day (run a game at medium settings or something which isn't MAX LOAD), and report back in a week. Temps will hopefully be better, in a few days.


**you can also try Kryosheet
https://www.techpowerup.com/review/thermal-grizzly-kryosheet-amd-gpu/

Techpowerup used it on a 7900 XTX and found it performed really well.

Yes, I bought the Thermal Grizzly PhaseSheet. I have to admit I tried gaming right after installing the PTM, but do you really think a few cycles can drastically improve this behavior? It seems too aggressive, and the PTM should manage to drop temperatures by about 15-20°C to keep them at decent levels

 

[cocoliche]

How are your fans behaving? If you haven't already, eliminate your fan profile and crank them up to maximum speed and see what that does.
Are you ocing? If so, step everything back down to stock.
You can also try cleaning everything off of the heatsink and doing the flashlight test to see if your heatsink is making good contact. Point the flashlight through both the front and back side of the GPU/heatsink mating area(laid flat) and look for any unusual gaps where light shouldn't be coming through.
Can't think of anything else but maybe something above will trigger something.
GL!

My fans are running fine, I tested them at 100% and the result is the same. I'm not overclocking the GPU. I cleaned everything up, and I couldn't do the flashlight test because this GPU's shroud design doesn't allow it, but I'd say the contact is good, though I'm not sure how good the mounting pressure is. I added a rice grain sized drop of thermal paste to check if that small amount spread well under the heatsink's pressure, and it seemed like it did

 

[chameleoneel]

Yes, I bought the Thermal Grizzly PhaseSheet. I have to admit I tried gaming right after installing the PTM, but do you really think a few cycles can drastically improve this behavior? It seems too aggressive, and the PTM should manage to drop temperatures by about 15-20°C to keep them at decent levels

Its common for fresh PTM to be a lot worse than average paste.

Use your GPU a little bit everyday. And tell us how the temps are, next week. Its important to completely shutdown your PC, when you aren't using it. The PTM has to thermal cycle. So, it needs to be used and heat up. And then it needs to fully cool off. And that needs to happen a few times, before the temps become optimal.

I have used PTM in Laptops, in GPUs, and on regular CPUs. Heatsinks and AIO's. Its almost always the same: Temps are kinda bad at first, and get better over a few days.

If it hasn't gotten better in a week, then something is wrong with the GPU.