CPU Lapping Tests

Add Your Comments

I have been trying some different things with lapping CPUs to see what makes lapping so effective. I had a hunch that exposing the copper underneath the nickel (I am guessing it’s nickel) cap of Intel’s CPUs is as important as lapping the CPU flat.

To test this out, I took my trusty old C266 SL2QG and proceeded to try two things:

One, minimally lap it flat exposing no copper, and
Two, grind off the “nickel” to expose the copper.

Now one thing that bugged me is how to determine when the CPU cap is flat? I tried a bunch of approaches and hit on one that is very simple, cheap and works real well. I take a permanent pen marker and paint the CPU cap with it. I then use 600 grit paper and sand it till I can not see any signs of the marker (It is very important that when you sand, you mount the sandpaper on an absolutely flat surface – best is a piece of glass). Stands to reason – if you can see marks left by the pen, then it is not perfectly flat yet.

What I found is what I expected – the CPU caps are really very flat to start with. I only needed minimal (about 2 minutes) sanding to get it flat. I then ran the CPU with Prime 95 on my BX6-2 and recorded thermal diode temps for 4 different bus speeds. Next I then took the CPU to my ginder and proceeded to grind off the “nickel” cap to get to the copper base. I then polished this up to 600 grit and tested it with the marker pen for flatness until I had no spots left. I then re-ran the CPU and recorded CPU temps again – the table below details what I found:

SL2QG 266    
Speed CPU Temp Nickel CPU Temp Copper
266 35.0 C 35.0 C
400 41.5 C 40.0 C
448 42.0 C 41.5 C
468 42.5 C 42.0 C

No doubt that the CPU runs cooler after exposing the copper, and no doubt that minimally lapping a CPU is a prudent appoach as well. Grinding off the nickel cap absolutely voids Intel’s warranty, so if you want a lot of lapping’s benefits but are not ready to go “whole hog”, then just smoothing out the cap can get you there. (BTW: Don’t forget the heatsink base as well – cheapo heatsinks are not all that flat!)

I would appreciate any additional experiences with lapping. You will note I stopped at 600 grit – I think going beyond 600 is not worth the effort. I believe this to be the case because you have to use thermal grease anyway, and its thickness mitigates any benefits from a mirror-like finish. I stand ready to be contradicted.

PS: In terms of technique, I wrapped plastic wrap around the CPU to keep the copper dust away from it. I cut a hole in the plastic wrap to expose the CPU cap and then used masking tape to ensure that no copper dust migrated to areas where it would do no good. I used a power grinder for the first sanding to speed things up although I don’t recommend this unless you are really familiar with such tools. Worked fine.

PPS: Lee Schneider has a real good description of lapping technique:

“To get a flat surface, shops use surface grinders. These are large,
floor mounted machine tools with stationary grinding wheels and the work piece mounted on a moving bed. This is obviously not a practical solution for the average home handyman.

There is however another method which works very well:

Modern windows glass is made by floating molten glass on a bed of liquid tin. The result is that a piece of cheap, quarter inch glass makes a perfect lapping bed. Put your 600 grit paper with the abrasive side up on the glass, hold the CPU or heat sink in your hand and carefully sand the piece until flat (and the nickel has been removed). The only precaution is to not rock the piece while sanding because it is possible to achieve a crown if the piece is not held in a proper fashion.

If the piece is relatively flat to start with, this is a much lower risk, as the initial flat surface will help prevent rocking during the sanding motion. A CPU can be sanded by simply putting your thumb, index or middle finger on the bottom inside the rows of pins, and then gently sliding the CPU back and forth against the 600 grit paper.

Heat sinks are a bit harder to do – the higher they are the easier it is to accidentally rock them during the sanding motion. With a bit of care and practice however, it is possible to get a very flat and smooth surface this way.”

Leave a Reply

Your email address will not be published. Required fields are marked *