The extremely thick layer of paste is really suspitious, the paste is meant to work as a thinnest layer between two bodies that touch each other. It is not meant as a gap filler, for that there are thermal PADS or even the graphene sheet.
If used as a gap filler, the paste is misused and you really cannot draw any conclusions about a product used in an unintended way.
I think you got my statement wrong.
The scenario in which I explained pump-out acceleration DOES NOT imply if paste is used as gap filler.
The scenario I described refers to the unexpected creation of micro-gaps due to the difference in pliability of thermal pad material differences between user applied pads and factory applied pads, therefore accelerating the pump-out effect due to having too much of a gap between the die and heat sink contact.
And if we are being needlessly pendatic, thermal paste IS used as gap filler as there are microscopic imperfections between the die and the heat sink contact area that will prevent effective heat transfer due to not being able to contact the surface area of each other.
GPU dies are not as susceptible to paste overfill issues leading to gaps compared to other direct die TIM applications (such as laptop heat sink mounts) as GPU heat sink assemblies have significantly stronger mounting pressure that can and will push out excess paste (Unless there is a significant curving on the GPU heat sink contact, but that is another issue).
I overfilled Thermalright TFX application my Colorful RTX 3070, and it had no adverse effects on the specified heat sink gap tolerances other than consuming too much thermal paste from the syringe. I can provide temperatures as proof, and this is at an area where 35 degrees Celsius is the standard average daytime ambient temperature.
I see no issue with the amount of paste that the original user posted, not unless the mounting pressure applied by the heat sink mounting screws is highly unequal in nature (which is very, very unlikely).
And to the original poster, Honeywell PTM 7950 (and similar products under other brands) will work well on that application. I used it on my Sapphire Nitro RX 5700 XT last year and it still does not show any signs of significant temperature deviations between core to junction temperatures. I can run a stress test later to show proof, if needed.
