Thermal paste pump effect

[tp4tissue]

Pretty severe on modern gpus.

Just repasted 9 months ago.  Only watched movies, so didn't notice performance issue.

But flipped on a game today, 85 Celcius, repasted, same load 55 Celcius.

The more/ larger the thermal cycle, the faster it pushes out.

[Leslieann]

I have doubts...
Not that it pumps, it can to a degree which is why you get so much suction on older past, but them showing a massive air cavity (or any cavity) is B.S.

Where did the air come from? If it's pasted all around, there's nowhere for air to get in.
Yes, some bubbles can/could exist and expand but this is why we use the paste methods we do, to reduce this problem, for the most part if it's done right it pumps out paste then sucks most of it back in and there shouldn't really be a chance for air to enter.  There is some initial pump out on first high temps, that's it self leveling and spreading.

All of this is even more important on a GPU as you should even paste the sides of the core because that too can lead to hot spots on the die. So again, where would this mythical air bubble come from when the die is swimming in paste?


If it pumped out as shown, liquid metal would never work.

[tp4tissue]

I believe the -Voiding is primarily due to the non-flatness of the interface as it bends.  The paste does not form an air tight seal, especially given surface roughness on both the die and heatsink/ IHS.

I've read that the Liquid metal oxidizes slightly on the edges which is why it is especially good at preventing pump out, it's also small enough such that it fills in more voids of the surface imperfections. Liquid metal DOES pump out though. just takes longer.  From personal experience I do see a slight crust where the die touches below the IHS, this is happens in about 3 years.

The video is exaggerated to show what's happening, but it's quite clear when Tp4 took the heatsink off, there was very little material in the middle of interface.    Tp4 is sure he mounted correctly the last time, because he's tested after each repaste, and the return of performance is consistent. 

[Darthbaggins]

What paste was used?  Kinda surprised having to repaste in less than a year. . .

[Leslieann]

The purpose of the paste is to fill all of the imperfections and create a perfect contact, if your heatsink is seeing enough pumpout to create an air bubble that heat sink (or IHS) is absolute garbage.

[Darthbaggins]

I haven't repasted my machine yet in the past 5yrs and seen only a variation of, at most, a 1c shift (most of the time it's due to dust build up in the rad). I will be repasting once i swap over cases since I have to reconfigure my loop. 

[tp4tissue]

What paste was used?  Kinda surprised having to repaste in less than a year. . .

Since Tp4 has to clean it so often, used the bundled pastes, Thermalright TF7

Pump out probably has very little to do with paste though, more so the number of hot cold cycles.  Since tp4 only watches movies, get alot of cycles.

[Darthbaggins]

Considering I haven't had to repaste in years running Thermal Grizzly Kryonaut as it has retained its interface ability over the years - only other pastes I run are MX-5 and ShinEtsu (ShinEtsu is very inexpensive and has been a good use case for when I have had to swap components alot).

[tp4tissue]

Been thinking about switching to those phase shift pads, but they're pretty much ruined if you have to take off the sink to clean the fins, so long run, it's too expensive. For laptop I think it makes alot of sense as long as the laptop's fans can come out without removing the whole sink.

[Leslieann]

My 8700k has never been repasted since new and it's had some serious heat cycles being in an ITX toaster oven. Same for my 1070.
I tend to use better paste (Noctua, MX5, etc..) but meh, never really had an issue.

I love the idea of the graphite pads and such, but why?
Paste is cheap, often free. I don't think I've ever repasted anything a second time and some of my junk is pretty old.

[tp4tissue]

My 8700k has never been repasted since new and it's had some serious heat cycles being in an ITX toaster oven. Same for my 1070.
I tend to use better paste (Noctua, MX5, etc..) but meh, never really had an issue.

I love the idea of the graphite pads and such, but why?
Paste is cheap, often free. I don't think I've ever repasted anything a second time and some of my junk is pretty old.

Tp4 now suspects LLann may not actually c0mput3r that... much..

[Sniping]

My 8700k has never been repasted since new and it's had some serious heat cycles being in an ITX toaster oven. Same for my 1070.
I tend to use better paste (Noctua, MX5, etc..) but meh, never really had an issue.

I love the idea of the graphite pads and such, but why?
Paste is cheap, often free. I don't think I've ever repasted anything a second time and some of my junk is pretty old.

Tp4 now suspects LLann may not actually c0mput3r that... much..

I feel like reality is more that people just get their computer set up right once, and pretty much don't touch it after. I assume 80% of enthusiasts are in this boat. temps will slowly rise over time but not enough to cause any noticeable issues, and by this point, people are already thinking about upgrading again. I also haven't repasted a system ever, only did it if I had to go in and swap the CPU or similar.

[Leslieann]

I feel like reality is more that people just get their computer set up right once, and pretty much don't touch it after. I assume 80% of enthusiasts are in this boat. temps will slowly rise over time but not enough to cause any noticeable issues, and by this point, people are already thinking about upgrading again. I also haven't repasted a system ever, only did it if I had to go in and swap the CPU or similar.

Exactly.

Old stuff is repasted as soon as I get it (usually) and my good stuff is done well the first time then only gets messed with when I upgrade. Are the temps higher after a few years, probably, but these days we have better coolers and chips that can handle going over 60c. That slight increase might matter if you overclock a lot but overclocking hardly nets anything these days.

[tp4tissue]

Are the temps higher after a few years, probably, but these days we have better coolers and chips that can handle going over 60c. That slight increase might matter if you overclock a lot but overclocking hardly nets anything these days.

This is what Tp4 is worried about though, there's average temp, and there's "Hot Spot" and not just where the hot spot sensor might be, the "anomalous" hotspot could be anywhere depending on the flatness of the interface, for example, maybe atop the video codec accelerator, and that section gets borked in the long run.

[Leslieann]

Are the temps higher after a few years, probably, but these days we have better coolers and chips that can handle going over 60c. That slight increase might matter if you overclock a lot but overclocking hardly nets anything these days.

This is what Tp4 is worried about though, there's average temp, and there's "Hot Spot" and not just where the hot spot sensor might be, the "anomalous" hotspot could be anywhere depending on the flatness of the interface, for example, maybe atop the video codec accelerator, and that section gets borked in the long run.

It's not like there's no compound to spread the heat even with pumping and anything modern *should* be able to detect that and throttle even if you can't read the temp. It would need to be one hell of a localized hot spot and garbage paste application/horrible heat sink design.

Could it happen in theory, maybe, is it common, absolutely not, and certainly going to be rare if you use good parts/compound/application.
I've blown up more hardware watercooling than I ever will by simply burning something up.