Re-Applying Thermal Paste Works Wonders

Discussion in 'Videocards - AMD Radeon' started by Eastcoasthandle, Dec 31, 2020.

  1. Eastcoasthandle

    Eastcoasthandle Ancient Guru

    Messages:
    2,773
    Likes Received:
    452
    GPU:
    Nitro 5700 XT
    After about a year using the 5700xt I noticed that I was losing FPS. Temps seem fine. But I noticed that the fan speed was ramping up a lot higher. But temps were still fine.

    Well, having done this 1000x already I knew I had to re-apply paste but I thought it was odd that I needed to do so this early. Well, when I opened it up this is what I found:

    [​IMG]


    [​IMG]

    That's a 1st! Saying I needed to re-apply paste is an understatement. But if you look closer you will see that thermal paste was applied in small dot grid. And I wondered if that causes some of my OC issues from a month ago.

    However, after re-applying thermal compound the fan is at 0 RPM at desktop consistently now. I almost forgot it could do that. :eek:
     
    Last edited: Jan 7, 2021
    GSDragoon, OnnA, ZXRaziel and 2 others like this.
  2. MerolaC

    MerolaC Ancient Guru

    Messages:
    3,438
    Likes Received:
    300
    GPU:
    MSI 5600XT G. MX
    That is called thermal pump out.
    The paste becomes liquid like, like it's seen there after it has suffered thermal cycles.

    Few commercials pastes are usable on for a direct die application.

    Thermal Grizzly Hydronaut is one of them, which doesn't suffer pump out.
    Other I've been using for my clients on their laptops is Gelid GC-Extreme.
    Others will suffer pumpout.


    Watch this vid in full:
     
    ZXRaziel and Undying like this.
  3. Mufflore

    Mufflore Ancient Guru

    Messages:
    12,502
    Likes Received:
    1,162
    GPU:
    Aorus 3090 Xtreme
    I found changing the paste to Gelid Extreme on any gfx card I owned, even when brand new, can help a lot.
    I havent noticed a pump out effect on a GPU since using Gelid extreme paste, I apply it very thin.
    ie my 1080ti FE has an Arctice cooling Extreme III cooler fitted since new and temps are still as good after 4 years.
    However the pump out effect still happens on a delidded 6700K CPU core, some cores jump up to 10C difference over time, that I havent solved.
     
    ZXRaziel likes this.
  4. anticupidon

    anticupidon Ancient Guru

    Messages:
    5,605
    Likes Received:
    2,058
    GPU:
    Polaris/Vega/Navi
    Well, didn't know about that pump-out. Guess one never stops learning, and that's a good thing.
    Gonna repaste all my gear in 2021.
    New year resolution already!
    Now, that's a reason to restock on thermal paste.
     
    ZXRaziel likes this.

  5. ZXRaziel

    ZXRaziel Master Guru

    Messages:
    411
    Likes Received:
    124
    GPU:
    Nvidia
    I have never seen thermal paste behaving like that , it was usually dried out for me .
    From physics Point of view you would think that the liquid should help the heat transfer, clearly there is more to it , would be interesting to know how thermal pastes are formulated and how do they work in depth .
    Interesting.
     
    anticupidon likes this.
  6. MerolaC

    MerolaC Ancient Guru

    Messages:
    3,438
    Likes Received:
    300
    GPU:
    MSI 5600XT G. MX
    More on this if anyone is interested.





    I did TONS AND TONS of research on this because of my work.
    MX-2 and 4 for example are EXTREMLY affected by Pumpout. I know because I have tested. The more silicone they have the more they pumpout.
    I always cringe when people say I changed my GPU paste to MX4. They say: the previous one was dried like rock.
    No, that's because that way it won't pump out. That's how it comes out of factory.
    Remember the dies on Intel CPUS? That look like rock? Yeah, it's by design, those pastes last YEARS before they actually become bad.
     
    ZXRaziel likes this.
  7. Astyanax

    Astyanax Ancient Guru

    Messages:
    9,531
    Likes Received:
    3,328
    GPU:
    GTX 1080ti
    This is a myth created to get you to buy more of the crap.

    The transfer material is the suspended particles in the paste, not the liquid itself.
     
  8. MerolaC

    MerolaC Ancient Guru

    Messages:
    3,438
    Likes Received:
    300
    GPU:
    MSI 5600XT G. MX
    LOL
    Myth.
    Sure, man.

    If you do you own testing, you wouldn't be spouting bullshit like this.
     
  9. Espionage724

    Espionage724 Active Member

    Messages:
    99
    Likes Received:
    38
    GPU:
    Radeon RX 580 8G
    What about using a graphite thermal pad? I switched to one for my CPU a while back and it seems to work well enough, but I like the idea of never having to worry about it once it's installed (won't dry out or degrade). I'm thinking about using one with my RX 580 at some point.
     
  10. MerolaC

    MerolaC Ancient Guru

    Messages:
    3,438
    Likes Received:
    300
    GPU:
    MSI 5600XT G. MX
    For a die application, don't do it.
    You need a HIGH amount of pressure to achieve good contact between the die and the heat sink.
    Stick to paste or LM.
     

  11. Passus

    Passus Ancient Guru

    Messages:
    1,654
    Likes Received:
    468
    GPU:
    RX 5700 Mech OC
    I use mx-4 on my 5700 and I also replaced the thermal pads

    easy 10c reduction on core and around 20c on junction
     
  12. Chastity

    Chastity Ancient Guru

    Messages:
    2,649
    Likes Received:
    868
    GPU:
    Nitro 5700 XT
    I'm still a big fan of ICD-7. I repasted laptops and never needed to redo it, ever. Haven't needed to redo my paste on my 5700XT, but now I'm curious to watch fan speeds.
     
  13. INSTG8R

    INSTG8R Ancient Guru

    Messages:
    1,659
    Likes Received:
    94
    GPU:
    Nitro+ 5700XT
    I just automatically repose my cards probably within a week of purchase p, I just can’t fight the urge. Usually use MX4 but last few I’ve been using Kryonaught.
     
  14. Eastcoasthandle

    Eastcoasthandle Ancient Guru

    Messages:
    2,773
    Likes Received:
    452
    GPU:
    Nitro 5700 XT
    If anyone is wondering about the thermal pad thickness of the vrm's/mosfet area of the Nitro+ it looks to be 1.5mm thick. Already compared OEM pad to 1.5mm one and looks to be a match.

    I would recommend Fujiploy Extreme Ultimate 17 W/mK


    But from watching the video any liquidie thermal paste should be replaced once a year.
     
  15. Eastcoasthandle

    Eastcoasthandle Ancient Guru

    Messages:
    2,773
    Likes Received:
    452
    GPU:
    Nitro 5700 XT
    Update: On CPU

    https://********/1sFtbDJ/paste1.jpg
    https://********/KzmcjpR/paste2.jpg

    Sorry about the blurriness I did some quick snaps with the phone as I had the loop in one hand and phone in the other.
    But as you can still see 100% pump out using thermal grizzly kryonaut. After about 2 years use. CPU temps were fine. .
    Which is strange to me. I would expect higher temps. But I was only get mid 50C running Cinebench. That is what I always used to determine if I need to reapply paste. At desktop I was around 20C. After re-applying it temps were still around 20C at desktop. Nothing changed temp wise using kryonaut

     

  16. MerolaC

    MerolaC Ancient Guru

    Messages:
    3,438
    Likes Received:
    300
    GPU:
    MSI 5600XT G. MX
    Thermal pumpout is only a problem on direct die applications, on IHS it won't matter at all.
     
    Eastcoasthandle likes this.
  17. anticupidon

    anticupidon Ancient Guru

    Messages:
    5,605
    Likes Received:
    2,058
    GPU:
    Polaris/Vega/Navi
    Care to elaborate?
    I'm curious. The above video made me think.
    I am intrigued.
    I didn't know about this until now.
     
  18. MerolaC

    MerolaC Ancient Guru

    Messages:
    3,438
    Likes Received:
    300
    GPU:
    MSI 5600XT G. MX
    Basically, on a direct die application, the tim end up squishing out of the die because of the thermal cycles and high temperatures (hot cold hot cold, that causes expansion and contraction of the hetsink) and you end up with poor coverage between the cold plate and the die.
    The temperature of the IHS is lower than that of a die, so, pump out is less likely to happen.
     
    anticupidon and Goiur like this.
  19. anticupidon

    anticupidon Ancient Guru

    Messages:
    5,605
    Likes Received:
    2,058
    GPU:
    Polaris/Vega/Navi
    But the thickness doesn't make a difference?
    I am asking because I bought some Shin Etsu TIM and was a pain to spread.
    But have the best temps always.
    We are talking 2012 time frame.
    Lots of brands and lots of improvements since then.
     
  20. Mufflore

    Mufflore Ancient Guru

    Messages:
    12,502
    Likes Received:
    1,162
    GPU:
    Aorus 3090 Xtreme
    Thicker helps but doesnt prevent it.

    I used MX4 or something like that on my 6700K delid until I had chosen something better.
    I think my CPUs lid is not flat on the inside because of how severe the issue becomes.
    At first all cores were within a few degrees C of each other after applying new paste, but after a few days some cores were up to 15C higher temp than others. Its always the same cores unless I reverse the lid, then its different cores, so I suspect the lid as well as the paste.
    Then with Gelid Extreme paste (very thick) it took a few weeks for the same to happen but isnt quite as bad, around 10 to 12C difference even after a year.
    I really want to get all cores near each other permanently though because it affects the max overclock quite badly.
    Not that it matters so much now cos I have a 10700K gaming machine.
    But dammit, this shouldnt be rocket science to fix!

    I actually bought one of those Kickstarter direct die water blocks a few years ago that ended up being severely delayed due to mfr'ing problems and companies not doing as asked.
    So the guy built his own CNC machine and they are apparently shipping out now, only mine was supposed to be sent a month ago and no word since then.
    Its been a strange experience.
     
    Last edited: Dec 31, 2020

Share This Page