Suggestion in how to measure gap and clearance on CPU heat sinks

If you have a CPU to heat sink or delidded CPU that you are considering sanding down the lid thickness to make closer contact with the die I have a solution that you can use to determine the clearance.

There is a product that machinists use to determine clearance inside of engine bearings when the caps are put back on the crankshaft. It looks like a really skinny piece of plastic string and you snip off a tiny piece and put it between the two objects. When you compress the assembly the plastic string gets smooshed flatter and the width of the string corresponds to the clearance.

This is called "Plastigauge" and costs just a few dollars for enough gauging materials to last you a lifetime. Here is the manufacturer's web site;

http://www.plastigaugeusa.com/how.html

You can buy it online, at some auto parts stores and even hardware stores (Ace hardware carried it the one time I needed some, several years ago).

The plastigauge comes in different "diameters" and the squish amount can work for really tiny clearances or really wide clearances (for the fat stuff). Also when you disassemble and look at the squished plastic string you can tell if the clearance is different in one corner of the heat sink or the other (it will squish down to a different thickness) so that will tell you if you have an alignment issue or a flatness issue that needs to be smoothed out.

As we know, the mechanical clearances between heat sink and CPU lid or between the CPU die and the CPU lid on a de-lidded chip should be as small as possible. You do not want an "interference" fit on the die and the lid (where there is less than 0 gap) as it will put pressure directly on the die.

Excessive clearance will mean that the gap needs to be filled with thermal compound and every thermal material has poorer thermal conductivity than the heat sink material. Here are some rough comparative numbers;


Gelid 8.5 W/mK
Arctic silver 8.9 W/mK
Galinstan (galium/tin/indium eutetic alloy) 16.5 W/mK
Galium 29 W/mK
Steel 12-50 W/mK
Nickel 90 W/mK
Magnesium 156 W/mK
Aluminum 205- 250 W/mK
Copper 385 W/mK (best grade)
Silver 406 W/mK
Diamond 1000 W/mK

In a perfect galaxy the CPU, lid and heat sink would all be made of diamond so heat would conduct nearly ideally. In reality there are some aspects of cost, weight and machinability that limit our choices.

Every boundary between the CPU components and the (air or water) introduces a slightly different insulating/conducting barrier. You want to keep the barriers as thin as possible with the highest possible numbers.

Here is an example.

The CPU die uses a thermal paste to make contact with the lid of the CPU. That lid is nickel plated but the plating is a few microns thick. The core of the lid is copper that has pretty good thermal performance.

De-lidding replaces the thermal paste with galinstan that is much better than a paste in conducting heat. BTW, you probably do not want to use gallium; it has a freezing point slightly under 30 C and when it cools it expands. You could have a computer that does not have a good connection when hot (because the thermal compound contracts) but then when it is shut off and cools the gallium expands and cracks the CPU die against the immovable lid. Galinstan does not freeze until -20 C so it is less of a risk.

Additionally pure gallium is pretty reactive to certain metals and will eat aluminum alive (think of a hundred years of corrosion in like an hour, you can actually watch it disintegrate aluminum, that's why it is illegal to transport on an airplane (like mercury).

Gallium and galinstan will even pit copper (in testing on a possible cooling fluid for a fusion reactor) but it is a much slower reaction.

Ideally you would want something like silver or nickel (has to do with electrochemical potentials too) that are a bit more inert.

While I will de-lid a CPU and use galinstan as a thermal compound (because the inside of the CPU package lid is nickel plated) I probably would be less thrilled to use it between nickel and copper (the CPU lid and the heat sink) because of a slow corrosion potential (more so on a vapor chamber copper cooler that is a bit more customized and expensive.

I would never-ever use liquid metal as a thermal compound on anything aluminum. That will be a science experiment gone bad and your machine would die probably before you put all of the case screws back in.

Watch:



I do want to use liquid metal (galinstan) but I will do something to protect the copper; For that little postage stamp sized area that comes in contact with the CPU lid I will nickel plate just that area. I can boundary that off with tape and create a little modeling clay retaining wall around the "square" and fill it with a few cc's of plating electrolyte and spend the 10-20 minutes with a strip of nickel attached through a few D cell batteries to put down a nickel place in that little square. Then when it is done I can take off the modeling clay, remove the tape and give the heat sink a scrubbing down. It will leave a little corrosion-resistant nickel plated square that is right on top on the copper heat sink (very minimal corrosion risk then).

 

Do you plan on reattaching the IHS for your CPU? You plan on lapping your laptop's heat sink? Please post back what sanding material you use and your process. Do you feel it is a bit excessive to do your own nickel plating? The Intel IHS itself seems to be pretty well constructed.

Please post back your results with tons of images.

 

Why not just use a feeler gauge? You can get it for <$10, it's made of metal, and doesn't get used up over time.

In 8 years of machine shop/machinist work, I've never used plastigauge, FWIW.