[SOLVED] Noctua Nt-h1 thermal paste on GPU, forced replacement every few months

[starkiller.lollo]

Hello everyone!
A couple years back I rerplaced the stock thermal paste on my sapphire rx480 with noctua's nt-h1, unfortunately getting only marginal gains in lower fan speeds. The card works great, my 1400core/2150mem mhz OC is stable and the fans manage to keep temperature under 78-80C with max 2300rpm. This is my issue: after 4-6 months it became unstable, I started noticing fans going to max allowed rpm even under mild loads and basically the cooler apparently just couldn't keep up (with consequent driver crashes and sometimes entire system shutdown). Thinking I'd done a poor application of the paste I reapplied it more carefully and using a little less of it, and all went back to normal again. But it just keeps happening, every few months I'm forced to reapply the thermal paste as if it lost thermal conductivity over time.

Has anyone ever faced this kind of behaviour? If it is caused by bad applications on my part, it would be weird considering there's been 4 of them by now and each time varying the amount of paste used (always used the grain-of-rice-in-the-middle method btw). Also I think it would be apparent immediately after rather than months later. The same I think would be true if I happened to get a tube from a faulty batch of grease. My only guess is that somehow Nt-h1 isn't made for working constatly at 75-80C under load, considering CPUs reach generally lower temps and might only peak at those temps occasionally, but this option seems sketchy as well to me. What do you think?

Thanks to everyone who will reply.

 

[jay32267]

I think there's a good chance of.....rather than the paste going bad (because as far as I know I've never had paste go bad)......I think there's a better chance of the screws coming loose.

 

[justin.m.beauvais]

Well, that is a strange one. Does the paste seem dried out when you change it? Sometimes in a dry environment the paste will just dry out and ruin the thermal conductivity.

Are you cleaning the surfaces properly with alcohol before you reapply then letting it dry? Sometimes having residue on the GPU or heatsink can cause a layer of poor thermal conductivity.

Have you checked the tightness of the heatsink mounting? If the screws are loosening over time you'll have contact issues

How about a different thermal material? The Noctua stuff isn't bad, but there might be a better compound for your application.

 

[tennis2]

Hmm. I just applied NT-H1 to my RX480 a month ago. Had great results. Will keep an eye out and report back if I can find this thread in 5 months.

 

[justin.m.beauvais]

Hmm. I just applied NT-H1 to my RX480 a month ago. Had great results. Will keep an eye out and report back if I can find this thread in 5 months.

When you load up the forums expand the forums tab and look for "Threads with your posts"

 

[tennis2]

I post a LOT. I added to my "watched" list though. That list is much shorter.

 

[mm333]

I have observe the same problem with Noctua NT-H1 on my GPU. Just after 4 months of repaste my 2070, the temp went up significantly higher than the stock 1yrs old paste used by Zotac. Each time I run Furmark load test, the temp will spike up from 34c to 60c in a second which indicate bad thermal paste.
While previously on my other system I've bad experience with MX-4 on direct die which have serious pump out effect. I guess this NT-H1 have the same issue too.

 

[Karadjgne]

NT-H1 doesn't dry out, and is good for over 8 years before even considering loosing any thermal conductivity. It sat on my i7-3770K for 6 years before cooler replacement caused a repaste and is still sitting on my gtx970 @ 124% OC, 7 years after initial treatment.

I generally have used Arctic MX-4 as it's slightly thicker for gpus, but was out at the time so used the Noctua paste instead.

One thing about paste, temps don't change radically, instantly. If paste is failing, it's over time, and temps will gradually rise over that period. It won't be 50 one day and 70 the next. That only happens with pastes that do dry out, like AS5, and the seal gets broken.

If the Noctua isn't to your liking, and MX-4 you find unreliable, try using Thermal Grizzly Kryonaut or Gelid GC Extreme.

If those pastes fail, you can be certain it's not a paste issue after 4 months, but something else entirely, more likely a windows update messing with drivers or some such.

 

[Karadjgne]

Each time I run Furmark load test, the temp will spike up from 34c to 60c in a second which indicate bad thermal paste

No, pastes are supposed to allow thermal transfer, the reason your temps spike from 34°C to 60°C is because that's what Furmark is doing. Generally you can expect Furmark to drive a gpu to 80°C or higher, so 60°C isn't bad at all.

If you are experiencing 'pump out' that's because you have used WAY too much paste, it's supposed to be hair-thin, fraction of a mm thickness between die and cooler/heatsink. And that's what pressure on the die from the heatsink will facilitate. By over using pastes, it's going to spill over the edges as pressure forces it out.

 

[mm333]

Did a repaste today, exactly the same problem as Arctic MX-4. From the screenshot below shows before was 34C spike to 60C+ in a second, after repaste it goes from 30C to 43C in a second upon load. You can see how the temperature drop after the load was removed.
MSI afterburner graph scale on both are the same, -5 to 97.

Here are the photos of how a thermal paste pump-out effect looks like, almost half of the GPU die do no have paste covered anymore. The thermal paste is still in perfect viscosity.
Just before you want to argue about the amount of thermal paste used, Zotac use way more paste than this and do not have such issue.

Currently re-paste with different brand of thermal paste, which I tested before that do not suffer from pump-out effect.

I have tried Thermal grizzly Hydronaut on my 1080ti, performance wasn't that good, it seems slightly worst than Zotac stock paste but so far did not notice any pump-out issue.

 

[BloodDivine]

Well, tennis2 how's the NT-H1 holding up 6 months later?

 

[tennis2]

Still goin' strong after 9 months!!
15+ minutes of FurMark = 72C @ 900rpm fans on my 1300MHz RX480 (MSI GamingX).

 

[BloodDivine]

Good to hear 😉. Which application method you used btw?

 

[tennis2]

Blob the size of half a grain of rice (one full couscous/barley), set the cooler on top, screw it down.

 

[thotcat]

Hello everyone!
A couple years back I rerplaced the stock thermal paste on my sapphire rx480 with noctua's nt-h1, unfortunately getting only marginal gains in lower fan speeds. The card works great, my 1400core/2150mem mhz OC is stable and the fans manage to keep temperature under 78-80C with max 2300rpm. This is my issue: after 4-6 months it became unstable, I started noticing fans going to max allowed rpm even under mild loads and basically the cooler apparently just couldn't keep up (with consequent driver crashes and sometimes entire system shutdown). Thinking I'd done a poor application of the paste I reapplied it more carefully and using a little less of it, and all went back to normal again. But it just keeps happening, every few months I'm forced to reapply the thermal paste as if it lost thermal conductivity over time.

Has anyone ever faced this kind of behaviour? If it is caused by bad applications on my part, it would be weird considering there's been 4 of them by now and each time varying the amount of paste used (always used the grain-of-rice-in-the-middle method btw). Also I think it would be apparent immediately after rather than months later. The same I think would be true if I happened to get a tube from a faulty batch of grease. My only guess is that somehow Nt-h1 isn't made for working constatly at 75-80C under load, considering CPUs reach generally lower temps and might only peak at those temps occasionally, but this option seems sketchy as well to me. What do you think?

Thanks to everyone who will reply.

I replaced the stock thermal paste in my laptop with this stuff and now the cpu thermal throttles, would not recommend

 

[Karadjgne]

Which is wild, I used NT-H1 on my i7-3770K for a solid 6 years at 4.9GHz (3.4GHz stock) and on my Asus Strix GTX970 at 124% OC, which I'm still using and is now 7 years old (both only a single application). And not a single thermal issue since the swap. For a long time it was the only paste I used in clients pc's, even though AS5 was cheaper, because it's simply better at every aspect from performance to cleaning to install to longetivity.

So either I got lucky, (if you knew me, that rarely ever happens) or ya'll didn't do it right and would have gotten the same results with any other paste.

From those pictures, that's way too much paste used, and you can see the ring of where the paste actually spread to. Looks like the heatsink was tightened down on one side first, then the other, not in an X pattern and not stepped.

 

[mm333]

Which is wild, I used NT-H1 on my i7-3770K for a solid 6 years at 4.9GHz (3.4GHz stock) and on my Asus Strix GTX970 at 124% OC, which I'm still using and is now 7 years old (both only a single application). And not a single thermal issue since the swap. For a long time it was the only paste I used in clients pc's, even though AS5 was cheaper, because it's simply better at every aspect from performance to cleaning to install to longetivity.

So either I got lucky, (if you knew me, that rarely ever happens) or ya'll didn't do it right and would have gotten the same results with any other paste.

From those pictures, that's way too much paste used, and you can see the ring of where the paste actually spread to. Looks like the heatsink was tightened down on one side first, then the other, not in an X pattern and not stepped.

It is supposed to form a layer of paste, but all of it has been pumped out that's why it looks a lot. Previously facing same issue with MX-4 on GPU and delided CPU die, tried all application method, same problem, but no issue when using it on CPU IHS.
The NT-H1 still running good on my AMD CPU after 1year while GPU already repaste twice.

I've did some research seems like many people are facing the same issue with NT-H1 on direct die application.

 

[starkiller.lollo]

Hello again, sorry for the long absence. I've come again into the same issue, just reapplied the paste to my rx480 and all temps problems went away instantly. Example: running The Witcher 3, the same scene would send my gpu to 82C, fan at 2800rpm, around 156w indicated; after changing the paste and let the scene set for a while temps are at 75C (as per temp goal in bios), fan at around 1800rpm, same amount of power used.

I will add that in the mean time i flashed a 580 bios (with custom 2800rpm fan max) onto the card, which has done magic in terms of voltage, power and general stability. Still after a year or so... you know the rest.

I honestly don't know what to think, I have the same (and I mean the same) nt-h1 on my 4770k, not a single problem. Temps have been very stable in the course of time. To reply to some posts, yes: all gpu screws are solid, there's full contact of the cooler with the die (from what I can tell when I take it apart). Yeah, the paste seems to be a little drier than new, but that's to be expected and it doesn't look different from a normal CPU application. Haven't had the chance to try mx-4 yet, maybe I will.

 

[mm333]

Hello again, sorry for the long absence. I've come again into the same issue, just reapplied the paste to my rx480 and all temps problems went away instantly. Example: running The Witcher 3, the same scene would send my gpu to 82C, fan at 2800rpm, around 156w indicated; after changing the paste and let the scene set for a while temps are at 75C (as per temp goal in bios), fan at around 1800rpm, same amount of power used.

I will add that in the mean time i flashed a 580 bios (with custom 2800rpm fan max) onto the card, which has done magic in terms of voltage, power and general stability. Still after a year or so... you know the rest.

I honestly don't know what to think, I have the same (and I mean the same) nt-h1 on my 4770k, not a single problem. Temps have been very stable in the course of time. To reply to some posts, yes: all gpu screws are solid, there's full contact of the cooler with the die (from what I can tell when I take it apart). Yeah, the paste seems to be a little drier than new, but that's to be expected and it doesn't look different from a normal CPU application. Haven't had the chance to try mx-4 yet, maybe I will.

Go thermal grizzly kryonaut, I've been using it on my GPU and CPU for couple of months now. No issue so far, temperature is lower than any paste I've tried. Don't use MX-4 on GPU, you'll face the same problem.
When you remove the heat sink, if you see the paste form a weird spot in middle of the die, it means it has pump out problem.

 

[tennis2]

My original application is still good. 14 months strong.

 

[starkiller.lollo]

Go thermal grizzly kryonaut, I've been using it on my GPU and CPU for couple of months now. No issue so far, temperature is lower than any paste I've tried. Don't use MX-4 on GPU, you'll face the same problem.
When you remove the heat sink, if you see the paste form a weird spot in middle of the die, it means it has pump out problem.

Yes, that's exactly the problem :\ I wanna try one last option, that is to reduce slightly the mounting pressure on the cooler. If that fails, or makes it even worse i'll have to try with thermal grizzly.