• Happy holidays, folks! Thanks to each and every one of you for being part of the Tom's Hardware community!

[SOLVED] Thermal paste not staying on GPU die

dantekb43

Honorable
Nov 21, 2014
33
0
10,540
Hi everyone!

Can you please help me with this thermal paste application issue that I'm having? My GPU model is a XFX 5700 XT THICC II

SHORT VERSION

I've tried repasting my GPU for the 5th time in the last 4 months, and in every single one of the attempts the GPU got good temps, but then 2-3 weeks later Tjunction would skyrocket to 110C and when i take off the cooler to see what's happening
the thermal paste is squished to the edges of the die and barely anything stayed on the die, leading to huge temps.
Do you guys have any idea why this is happening or how i can fix it?

(Check photos below to see what I'm talking about, and yes, the photo was from my last attempt where i tried to put A LOT of thermal paste to see if it would end up fixing it, but then all of the paste got accumulated on the sides)

LONG VERSION

I am a "proud" owner of a XFX 5700 XT THICC II (yes, it's the bad model that Steve from Gamers Nexus talked about, I know this, but it's the only one I could get my hands on at the time)

About a year and a half ago, I took off the upper shroud plastic and the backplate off the card (Go watch the Gamers Nexus video about "fixing the THICC II" to see why i did this).
After i took them off and repasted the card, temps were amazing, no issues whatsoever and i moved on with my life (all of this is using Noctua NT-H1 TP).

About 4 months ago, Tjunction was hitting 100+C, so I disassembled the card to repaste it, no big deal.
About 2-3 weeks later, Tjunction was back at 100+C, and i repeated the process.... then it happened again....and again...

In my 4th attempt, i tried applying A LOT of thermal paste to see if it would fix the issue, but as you can see from the pictures below the TP squishes out the side completely (at least the high Tjunction is explained).

I've tried applying less TP, more TP, increasing the pressure on the back screws, decreasing pressure.... nothing appears to permanently fix it.
One of the tests I'm currently doing while writing this is to put back the backplate/shroud that I've removed from the card (to see if it has anything to do with deformations on the PCB because of weight or whatever)
but now temps suck because this cooler design sucks, hence why i removed the shroud in the first place.

The only "solution" that i can think of is to use those IC Graphite thermal pads that are replacements for TP and since they're solid they're not gonna squish out the sides.

Does anyone of you guys have ANY idea on why this MIGHT be happening? I'm at a loss here because after years of repasting several GPUs I've never dealt with this bizarre issue.

Thank you all in advance for the help 😊

IMG-20210721-121932.jpg


IMG-20210721-121925.jpg
 
Solution
I'm gonna write this down and mark this as "Best Answer". However while it technically solves the problem it does not explain the cause of this issue, it's just a bypass so that i can finally use this GPU normally.

The "solution" was to use IC Graphite as a substitute for thermal paste. [!!!WARNING!!! IC GRAPHITE IS ELECTRICALLY CONDUCTIVE, DO THIS AT YOUR OWN RISK AND PAY ATTENTION FOR IT NOT TO TOUCH OTHER COMPONENTS OTHER THAN THE GPU DIE]

What I did was to buy one of the 30x30mm IC Graphite pads, measured the die and then I cut the pad to the exact size of the die...
then i applied the TINIEST amount of Noctua NT-H1 TP to the die, not to cover it, but to use the TP as some sort of "glue" to keep the IC Graphite in...
Hi everyone!

Can you please help me with this thermal paste application issue that I'm having? My GPU model is a XFX 5700 XT THICC II

SHORT VERSION

I've tried repasting my GPU for the 5th time in the last 4 months, and in every single one of the attempts the GPU got good temps, but then 2-3 weeks later Tjunction would skyrocket to 110C and when i take off the cooler to see what's happening
the thermal paste is squished to the edges of the die and barely anything stayed on the die, leading to huge temps.
Do you guys have any idea why this is happening or how i can fix it?

(Check photos below to see what I'm talking about, and yes, the photo was from my last attempt where i tried to put A LOT of thermal paste to see if it would end up fixing it, but then all of the paste got accumulated on the sides)

LONG VERSION

I am a "proud" owner of a XFX 5700 XT THICC II (yes, it's the bad model that Steve from Gamers Nexus talked about, I know this, but it's the only one I could get my hands on at the time)

About a year and a half ago, I took off the upper shroud plastic and the backplate off the card (Go watch the Gamers Nexus video about "fixing the THICC II" to see why i did this).
After i took them off and repasted the card, temps were amazing, no issues whatsoever and i moved on with my life (all of this is using Noctua NT-H1 TP).

About 4 months ago, Tjunction was hitting 100+C, so I disassembled the card to repaste it, no big deal.
About 2-3 weeks later, Tjunction was back at 100+C, and i repeated the process.... then it happened again....and again...

In my 4th attempt, i tried applying A LOT of thermal paste to see if it would fix the issue, but as you can see from the pictures below the TP squishes out the side completely (at least the high Tjunction is explained).

I've tried applying less TP, more TP, increasing the pressure on the back screws, decreasing pressure.... nothing appears to permanently fix it.
One of the tests I'm currently doing while writing this is to put back the backplate/shroud that I've removed from the card (to see if it has anything to do with deformations on the PCB because of weight or whatever)
but now temps suck because this cooler design sucks, hence why i removed the shroud in the first place.

The only "solution" that i can think of is to use those IC Graphite thermal pads that are replacements for TP and since they're solid they're not gonna squish out the sides.

Does anyone of you guys have ANY idea on why this MIGHT be happening? I'm at a loss here because after years of repasting several GPUs I've never dealt with this bizarre issue.

Thank you all in advance for the help 😊

IMG-20210721-121932.jpg


IMG-20210721-121925.jpg
you're doing everything fine, the thermal paste worked well. TJunction is the max safe temp. Look at this:
unknown.png
 
you're doing everything fine, the thermal paste worked well. TJunction is the max safe temp. Look at this:
unknown.png

@Koekieezz I'm sorry, but it's not fine for two main reasons

1) The GPU is constantly crashing when this happens, so even if i was OK with the temps the fact that it's crashing doesn't let me, and the only way to stop it is to either repaste the card or MASSIVELY downclock/undervolt the GPU.

2) In normal operations, the Tedge would stay @ 70C and Tjunction @85C max... now edge stays at 70C and junction skyrockets to 110C
which by the way is ABSOLUTE MAXIMUM for safe operations, after this the card downclocks so that it doesnt kill itself.

I'm still at a loss on what to do...
 
Too much TIM is almost as bad as too little.
The main downsides to too much non-electrically-conductive paste are nasty clean-up and wasted paste. Ease of clean-up is the main reason I try not to over-do paste.

As long as the hold-down force is adequate and the paste has low to moderate viscosity which most aftermarket stuff does for ease of application, most of the excess will ooze out and take care of itself. Linus and others have reviewed "paste application methods" including scenarios with heaping amounts and the delta from best to worst was still only a few degrees excluding the too-little-paste scenarios.
 
Too much TIM is almost as bad as too little.You just need a thin film to fill any miniscule gaps between the heatsink and processor. Have you changed the thermal pads on all the other hot bits?

Like I said, Just one of my tests was to apply A LOT of TP just to confirm that it was not a lack of TP for whatever reason (and of course, it was not).

I didn't change any thermal pads because they were neither defective nor they were causing the crashes.
 
I'm gonna write this down and mark this as "Best Answer". However while it technically solves the problem it does not explain the cause of this issue, it's just a bypass so that i can finally use this GPU normally.

The "solution" was to use IC Graphite as a substitute for thermal paste. [!!!WARNING!!! IC GRAPHITE IS ELECTRICALLY CONDUCTIVE, DO THIS AT YOUR OWN RISK AND PAY ATTENTION FOR IT NOT TO TOUCH OTHER COMPONENTS OTHER THAN THE GPU DIE]

What I did was to buy one of the 30x30mm IC Graphite pads, measured the die and then I cut the pad to the exact size of the die...
then i applied the TINIEST amount of Noctua NT-H1 TP to the die, not to cover it, but to use the TP as some sort of "glue" to keep the IC Graphite in place while i mounted the cooler back on.

Done, problem "solved".
It does perform slightly worse than NT-H1 (i would get Tedge 70C and Tjunction 85C, now with IC Graphite I get Tedge 73C and Tjunction 89C)
but now i can use this GPU without having to repaste it every 2 weeks or so.
 
Solution
IMO, too much paste is not as bad as too little - it is wasteful though.
The cooler mounting pressure is going to squeeze off excess paste. Use too little paste, and you're left with hotspots.


That said, NT-H1 isn't a very good paste for gpu applications, as it's viscosity isn't very high. Throw it all over cpu applications if you want though.
The gpu dies are smaller, as well as having a smoother surface than the IHS on a cpu. This does have an effect on which pastes work well or not.
NT-H1 is more likely to 'slide' off.
For gpu applications look to Kingpin KPx, Gelid GC-Extreme, Cooler Master Mastergel Maker Nano, Prolimatech PK-3, or Noctua NT-H2(has higher viscosity than the original).
 
IMO, too much paste is not as bad as too little - it is wasteful though.
The cooler mounting pressure is going to squeeze off excess paste. Use too little paste, and you're left with hotspots.


That said, NT-H1 isn't a very good paste for gpu applications, as it's viscosity isn't very high. Throw it all over cpu applications if you want though.
The gpu dies are smaller, as well as having a smoother surface than the IHS on a cpu. This does have an effect on which pastes work well or not.
NT-H1 is more likely to 'slide' off.
For gpu applications look to Kingpin KPx, Gelid GC-Extreme, Cooler Master Mastergel Maker Nano, Prolimatech PK-3, or Noctua NT-H2(has higher viscosity than the original).

You might be right, however I've repasted many GPUs with both NT-H1 and MX-4 (which has an even lower viscosity than the Noctua one) and I've never had any issues for years.
Hell, this GPU itself was repasted with NT-H1 and it didn't present issues for 1.5 years.

I might try one of these later as an experiment, I just can't test them now because in the country I live in most of these are not sold here, I have to import them (I might find the Cooler Master one).
I had to even import the IC Graphite (which ended up being expensive as heck because of fees), but at least i was gonna import something with a high chance of success.

Thanks for the idea! If one day I end up trying it (and if I still remember that I made this post haha) I will update it.
 
Like I said, Just one of my tests was to apply A LOT of TP just to confirm that it was not a lack of TP for whatever reason (and of course, it was not).

I didn't change any thermal pads because they were neither defective nor they were causing the crashes.
Changing the pads is a good idea for optimal cooling. Pads get deformed when pressed against components and you may not get full contact after re-using.

Regarding too much TIM, if non-conductive material is used, too much is not as much a concern. If conductive, it is a potential disaster.