Are You Repasting Your GPU Correctly? - Testing Paste Applications (Thin, X, Line, Dot)

preview_player
Показать описание
It seems like no one can agree on how thermal paste should be applied, especially on graphics cards. So I set out to try 4 of the most common patterns that people recommend and see what works best.
#bestgputhermalpaste #gpu #thermalpaste #pcbuilding

Buy These Pastes:

Raw Data:

SSD Shootout Video:

---------------------------------------------------

Timestamps:
0:00 Intro
0:32 Improvements over previous video
1:30 Test setup
2:15 The testing
3:42 Results
6:41 Outro

---------------------------------------------------

Music (in order):
"VULF JAMS" - GARRISON
"If You Want To" - Me
Рекомендации по теме
Комментарии
Автор

Good to know that, as long as you apply enough of it, the specific method doesn't really matter. As for what might give better performance... I mean... I suppose liquid metal would the only comparable thing that could give you better heat transfer. But I don't think that'll really be worth it for most people. :P

thesilverydragon
Автор

I been around a while in the PC scene, I think a lot of this conversation has got muddled over time application was important when high end pastes tend to contain conductive materials.

Overspill when you used those could kill your stuff. This is less of an issue today since high end pastes are within 1C of each other in performance and are also non-conductive making overspill not a problem.

GPUs specifically are naked dies unlike CPUs today which have Integrated Heat Spreaders. It's less important to cover the entire CPU IHS so long as you paste the hot spots where the die is underneath it.

GPUs therefore require full coverage since every inch of that surface will get hot. Also it's a requirement to use non-conductive pastes with GPUs as the general application recommendations will cause some overspill inevitably.

AK
Автор

There was old article on an overclocking website that did like a dozen different application methods and did it over like a suite of 100 tests. It covered all your points but even through some more esoteric ones(spiral for bigger dies that some server manufacturers suggest, smiley face for the meme etc).

The tests were consistent across make and model they tested. Ironically the smiley face did the best - but even that was written off as well within variance.

long story short - doesn't matter. Its an argument for no reason.

varables
Автор

Just found out the channel, amazing and underrated content!

CodahWasTaken
Автор

The most important thing is the kind of paste. MX-4 is a good paste for CPU with an integrated heat spreader but it's a horrible choice for GPU which has an naked die. It will work for 2-3 months but eventually it will pump out because it's a very liquidy paste. For a GPU you need a very viscous paste that is more like a clay. Good examples of GPU pastes would be DowsilTC-5026, TC-5688, Shin-Etsu X7921, Prolimatech PK-3, IC diamond 24. Bad choices would be TG Kryonaut, MX-4 and 5, Noctua NT-H1 and 2 are all very liquidy and doesn't dry out but has trouble sticking to the naked die surface and Arctic Silver 5 because of its mostly oil properties which dries way too quickly. Quantity, Application method doesn't matter that much as long it's full coverage than it's good enough.

turuutmx
Автор

HH we NEED a youtube release of your theme song. it sounds so good!

spreadthegamer
Автор

Only danger is to use too little with the dot application, like way less than a pea-sized amount. When spreading, the paste might not reach the corners.
So better use a bit too much than too little :)

ProjectPhysX
Автор

I always do an X and have had 100% coverage on tge heat spreader when I've removed the cooler. Honestly, its really weird people gets so animated about this issue. I've seen nasty fights in comments of pc groups over thermal paste applications. On a gpu I just do a grain of rice blob in the middle.

adamadkins
Автор

pretty awesome
thank you for the video

i always use the X pattern then remove the heat sink to see if paste covers it all
then
re-apply

vanfly
Автор

Pretty cool video even after 2 years. Can't believe you got the exact same temps 8 times over.

Sapphire says to put 0.2 mm of thermal paste on their Pulse Vega 56 (~9 cm²). That would equate to 0.45 grams of MX-6 or 1/4 of a 2 g syringe.

mleise
Автор

I recommend to try phase-change material Honeywell PTM7950. It comes as a sheet and you don't have to mess around with any applying dots, dashes or crosses. Just cut out a piece the size of the IHS/core and stick it on. Also don't have to worry about whether the paste is too little or too much - it will be just right. In addition, it lasts longer, performs works better than traditional pastes and there is no pump-out effect.

goorthiss
Автор

After working with so many different GPU models over the years, I came to a conclusion that using spread method can be very questionable with some coolers.
To sum it up - Dot method is great for small dies. X for larger ones. X with tiny dots between lines for really big dies/IHS. I never spread because of accidents when coolers did not made perfect contact with die (which resulted in GPU extremely overheating).

OkiemElektroniki
Автор

Great video!
I think that this is quite a underrated channel for the amount of effort that goes into making these amazing videos. And one other thing, I have run into a bit of an interesting issue in your Discord server.

obsurdhotdog
Автор

Another great video with very thorough testing. It's sad though that we have to deal with people who seem to exist just to argue.

kyleallred
Автор

I personally like the idea of the thin spread method but from what I hear, theoretically, you could introduce tiny bubbles into the application doing it that way and in doing so, add a small layer of insulation between the die and the paste. But this is just heresay and I've not done enough testing (or have done an application poorly enough) to run into this issue myself.

sageosaka
Автор

I think you did very well with your testing method here. As always, data is key for clearing up these myths. Nice work!

NinjaISwill
Автор

Nice testing method. I have that very gpu. Your video makes me want to look at the temps the gpu is putting out.

JMassengill
Автор

Aye glad to see my new favourite channel show up again with another video

therealtsa
Автор

Ah! This is amazing, i was looking for a 1080 strix video and seeing a 1070 is perfect! Thank you man

valll
Автор

Bro we want that background music! We love it❤

djrockzz