Wes Newell said:
About the thickness of a piece of paper is more than plenty, if that
helps.
Which reminds me of a method I thought of that I want to try next time:
Get a hard-plastic or metal straight edge and put some paper-thin tape
from one side to the other, over the edge. (Wrap it over the edge.)
Do it again with a gap between the two a bit bigger than the CPU.
Then use that to spread the goop onto the heat sink bottom.
I've also wondered if it would be a good idea to also spread the goop
on the CPU and then wipe as much off it as possible with a plain
straight edge -- just to make sure that all pits are filled. (I've
noticed when gluing things with the glue-one-side-only method and then
separating them, that there will sometimes be dry spots, but of course
those were courser surfaces.)
BTW:
T = Q * z / ( k * A ), where (example)
T = temperature drop across goop = 20 deg_C [See note 2]
Q = power flowing through goop = 60 watts [some go higher]
z = thickness of goop = 0.0001 m [See note 3]
A = area of goop = (0.010 m)^2 [barton ?]
k = Thermal Conductivity of goop = 3.0 W/(m*degC) [See note 1]
Note 1: IIRC, 3.0 is about as good as it gets. Good quality plain old
"white stuff" is about 0.4.
Note 2: This means that the chip is at least 20 deg_C hotter than
ambient with a perfect heat sink.
Note 3: 500 sheets of typical 20 lb paper is about 50mm thick so one
sheet is about 0.1 mm (about 0.004 inch).
So "more than plenty" is probably right. Use thin paper.