Does lapping work?
A friend of mine just told me that lapping is the greatest thing.
I'm skeptical. I figure that the thermal paste can do its job.
How many degrees Celsius will my temps drop if I lap my 2500k and my heatsink?
Lapping works if the surfaces are not
matched to begin with.
Two flat surfaces are a match, flat on flat is a match.
Likewise one convex surface
nested onto a concave surface of equal radius of curvature will be a match. Neither is flat but their surfaces are still geometrically matched.
And yet a third criterion comes into play - flex. Or to be more scientific and specific - strain.
With this you match two intentionally convex surfaces which are designed to "flatten out" (elastic deformation under strain) under the mount pressure (stress, i.e. load) when you tighten the bolts on your block.
Generally lapping only needs to be done when you have mismatched
concave surfaces, valleys that end up being filled full of TIM that has less than optimal heat conduction.
My NH-D14 was convex, the center was a high point. This HSF did not need to be lapped as the convex center would just force the IHS to deform and conform to itself as a matching surface under the load provided by the mount.
However, my H100 was extremely concave, the center bowed inwards (away from the IHS) as you can see here when the hash-marks on the surface would only polish away at the edges of the surface:
But remember, this kind of works like addition and trying to get an even number as the result - two even number will always make an even number, and two odds numbers added together always make an even number, but an even number plus and odd number always leaves you with an odd number.
You can have a convex or a concave HSF surface and be fine without lapping so long as the IHS surface is convex
enough to fill in the gap of the concave surface when all the bending and flexing is factored in from the mount pressure. What you can't have is two concave surfaces, that is when you really need to lap.
In my case, my specific 2600K had a convex IHS surface, so lapping for flatness sake was unnecessary (but I did it anyway just for that extra degree or two of cooling that comes from minimizing total surface area of the mated surfaces):