The GPU VRAM chips need special handling as the gap to the heat sink is greater than the distance the GPU is. This gap is the issue! Standard CPU/GPU chip TIM is designed to fill the micro surfaces of the metal to improve the heat transmission. It is not designed to span a gap!
Think of it like water pressed between two sheets of glass so only a very thin layer is present between.
The VRAM chips have about 1.5mm of gap between the top of the chips to the heat sink. Apple started off with heat pads and then transitioned to special gapping TIM. Sadly, it was a failure! As it just can’t effectively move the heat across the distance. I only use thermal pads on the VRAM chips.
OK, moving back to the CPU & GPU chips …
The differences between the top five thermal pastes are so little it really doesn’t matter! Unless you are pushing your system to the wall every day for long periods. The best way to think this is do you need a tricked out car to get to work every day or are you in the middle of the indy500 race?
I personally use both Thermal Grizzly Kryonaut (12.5 W / mk conductivity) and Arctic Silver 5 (9.5 W / mk conductivity). I use Kryonaut on my pro’s gear and I use Silver on most of my standard users. The most important aspects is being non-conductive or capacitive as that will mess up your chip if it spreads past the area it should be placed.
What is important here is the volume needed is very little! You only want enough to coat the black silicon chip in a laptop or the metal heat spreader on a desktop class chip. You don’t want it to oozing onto the green chip base as it will interfere with the small SMT devices scattered on it. This is why you need to be careful! To much and you’ll chase your tail looking for a non-problem!
Going back to the thermal pads for a sec…
To be safe I dry fit the heat sink on and using a feeler gage (or a few strips of card stock) I measure the gap between the VRAM chips and the bottom of the heat sink. The pad height needs to be this or slightly higher.