Only checked the OP of the Op-er, scanning through some other posts quickly.
First, on the TR-Ultra 120 Extreme or the Ultima 90, you need a primitive jig (as in jigs that are featured for your woodworking router in "American Woodworker Magazine.") I spoke to tech-support at SVC.com, concerning their approach to custom-lapped U-120-extremes. They confirmed my guess that they'd designed a jig so they could knock out the custom-lappings like an M-16 spits lead.
The reason you need a jig is as follows. TR deliberately designed a convex ridge that runs evenly across the middle of the heatsink base. Therefore, you need to start the lapping to begin grinding a flat surface evenly in that ridge-- as opposed to grinding it down on one side or the other.
I used two pieces of white-pine, although oak would've been better, cut 1/2" x 1/2" x 6" length. [and the reason I used white pine: I already had a length of 1/2" x 1/2" that was 12" long.]
I have a little six-inch quick-release clamp that came in a set I bought at an auto-parts store (although you'd find this sort of thing useful for woodworking.) I slipped the clamp through the heatpipes, put the wood slats on either side of the base (and obviously, none of these items come in contact with the pipes.) I aligned them with the glass coffee table I use for lapping, so that they were also flush with the base. I tightend the clamp.
Then I started lapping. When the flat area on the convex ridge had widened to 1/2" or more, I pulled off the wood and clamps, and carefully continued, attentive to keeping the flat surface flush with the sandpaper.
You need to check flatness with a straightedge -- a razor-blade for instance. You do this by holding the surface up to a strong light and turning it so that the razor-blade spans various points on the "compass."
This holds equally for the IHS of the processor, although you don't need a jig for that. You're also advised to grind off all the nickel-plating, because nickel has a significantly lower thermal conductivity and higher thermal resistance (opposite sides -- same coin) -- than copper.
These things can be carefully measured at controlled room-ambients, and I am telling you that I did it. The lapped version of the U-120-Extreme was worth at least an extra 3 to 4C improvement over the unlapped version -- some people reporting more. Lapping the processor cap as I did this week means between 4C and 5C improvement. Diamond thermal paste, given tests I performed last summer with both my own homemade "mix" and IC Diamond -- is worth between 2 and 4C degrees improvement.
Amen.
Go in peace.
PS. Apply powers of deduction to this story:
Temperature accounting 101
Facebook
Twitter