Certainly. I think there needs to be some experimentation to flesh out what is really going on. The material behavior that would be required for "pump-out" to be a thing just does not jive.
I've already seen this myself on my 3090 FE. This was with Thermalright TFX and Thermalright Odyssey 1.5mm pads.
When I had Thermalright Odyssey pads, temps rose steadily after a few weeks. First was the core to core hot spot delta increasing. Started at 10C, then ended up at 16C After three weeks.
Then the overall temp increasing under the same ambients (77C-->82C load).
Took off the heatsink and investigated.
There were small sections along the VERY EDGE of the core with no paste on the die at all. Welp, there's your hotspot problem. And there were sections where the paste seemed to have great contact and areas with it poor, in a concentric shape. Igor's lab explains all of this.
Just look at the pictures. You do know who Igor is, right? he used to write reviews for Tom's Hardware.
Of course, you’ll be amazed if, for example, three supposedly identical cards from one manufacturer have completely different temperatures or fan speeds and this is then reflected very similarly in…
www.igorslab.de
There's an obvious contact problem. Take a look at the plots and the high resolution scans. Normal people don't have access to such expensive equipment to measure this stuff.
This is what causes your "pump out" which is accelerated by low PSI pressure.
Heatsink contacts paste, thermal cycling shifts the paste around, areas with low contact end up getting even lower contact, some sections get pushed out. Heatsink expands/contracts, paste gets shifted around, until the end user notices something isn't right.
When I tested Kryonaut Extreme on my 3090 FE with the Thermalright pads, the problem was even worse.
Kryonaut Extreme was about 1C worse than TFX starting out (as expected on direct dies, just like shown here):
Thank you for visiting the TechnologyGuide network. Unfortunately, these forums are no longer active. We extend a heartfelt thank you to the entire community for their steadfast support—it is really you, our readers, that drove
forum.notebookreview.com
But after just ONE week, temps had worsened by 5C. FIVE C. Is massive. And it just took one week of gaming and looping Heaven benchmark for it to happen.
I took off the heatsink and was horrified by what I saw.
While the entire die had "paste" on it, the KE was in a really 'swimmy" type pattern, absolutely nothing like what it looked like when I had tested and removed it on my CPU or my other video cards. It looked very similar to what the "soupy MX4" looked like, except the paste was still fully intact (probably was reusable too but I never re-use). The paste was simply too thin for the terrible contact pressure that was happening, but was naturally thicker at first since it was spread thick and had to compress.
Note: I saw this exact same 'swimmy' pattern on my dead Vega 64, when I applied Kryonaut to it and mounted the PCB to the heatsink, but didn't screw in the X-bracket around the die. Left it there for a month, took it off, and saw that same weird "wavy" pattern. And without the X-bracket, mounting pressure was extremely low, so there you go (I noticed this long after I fixed my 3090 FE deltas; I was preparing the dead Vega 64 for selling on ebay as parts).
The cause:
The TR Odyssey 1.5mm pads. 1.0mm were too thin (there would be no contact with the VRM's or VRAM) but the Thermalright pads had a too high shore rating for the leaf spring (unlike AMD cards, only the leaf spring screws helped with mounting pressure!) to compress the pads enough and to keep proper contact with the core! A quick test with "Fujifilm Ultra Low Prescale" contact paper determined my worst fears--there wasn't 'even a mark left on the contact paper!
So yes, extremely low contact pressure also causes pump out (paste flattens, no longer makes proper 'complete' contact due to convex cores, problem exasperated by too hard thermal pads).
Solution:
Stuck with Thermalright TFX, switched out the Thermalright Odyssey 1.5mm pads for Gelid Extreme 1.5mm pads. These pads are much softer and thus compress more easily, giving the leaf spring less work to do.
I didn't take a picture, but a test with Fujifilm Ultra Low Prescale (25-85 PSI rating) showed some nice bright red marks fully around the die, showing MUCH better contact was being made. While it wasn't completely uniform everywhere, at least it was mostly in the pattern of the die, there were actual red marks around the edges, and remember--the first test showed literally no firm contact at all (<25 PSI obviously), so that means changing to softer pads made a gigantic difference. In fact the pressure was showing up as better on the 3090 FE now, than on my (sanded die + IHS relidded) 9900k + NH-D15!
End result:
10-11C Core temp->Core hotspot deltas after 24 days.
I think that's a big win and shows that pump out is a real thing, and is worse with runny paste on imperfect dies and heatsinks.
This was not cheap for me to test. I spent a LOT of money buying different thermal pastes and pads, and I never would have solved this without reading other people's comments about 3080 FE's and 2mm Gelid Extreme pads 'working' vs 2mm Thermalright pads 'having terrible core contact', and eventually put 2 and 2 together (3090 FE's use thinner pads, but it was the same thing re Odyssey hardness vs Gelid Extreme softness).
