Anyone ever try Gold instead of thermal paste?

[sitka]

Just saying,
Somebody might have a some gilding around their place. The stuff craft people use to cover statues and picture frames. I imagine there is the real gold stuff and a fake kind. It's only a couple molecules thick but when burnished onto a surface is very close to electroplating. With a little effort about 20 layers should give a good base.

 

[johncar]

Sitka,
As a matter of fact yes, Jim Mitchell tried it about 3 years ago after we had tested Circuit Works' Conductive Silver Grease for "running", (it doesn't run). Jim was relating the then new application of silver grease as a better interface material to Ivan, Russian icon painter in former career...who suggested the gold leaf you describe.

Not exactly sure if Jim got real gold thing or hobby store substitute, since he only paid about $12 for a few small sheets. But he did roll 2 sheets, one on Cyrix chip, other on hs surface, then assembled.

Guaged improvement over the usual zinc-oxide grease grease by feeling the hs and noting that it was slightly warmer. This fits theory since
a lower thermal resistance interface will cause a higher proportion of cpu power dissipation/heat to flow thru the path that has had its thermal res reduced...ie the hs path as opposed to all other secondary heat flow paths. Then this slightly higher hs heat flow causes the hs temp to rise slightly. (Note, chip gets cooler, but hs gets warmer with better interface...if you don't believe it..go to http://members.nbci.com/candjac/index.htm socket thermistor article).

Jim never quantified the improvement further, and no one ever pushed the gold leaf as well...several issues. First, cost...apparently the 2 layers sort of fuse so the material is not likely to be re-usable. And secondly, the approximately 11 fold improvement of silver greases such as current idealized Arctic Silver over zinc-oxide types reduced the interface temp drop to about 1 or 2C for Socket7 chips, which would have made gold leaf subject to the "law of diminishing returns"...possibly better but not much "improvement" in terms of temp/degreesC...(might be viable as power dissipations climb and chip areas get smaller...sure that interface temp drops are not 1 or 2C on Tbirds).

Think gold leaf theory is like this... rolling the very soft gold leaf gets rid of most air and flows some material into microscopic surface cavities, but you won't get 100% surface contact. However, even if you got only about 15% or so, (over the normal 1 to 2%), the thermal resistance of gold is so very much lower than any grease compound, including silver greases, that an improvement is possible. However it remains to be accurately measured, and as above, might suffer from "diminshing returns" except where power dissipation is very high and chip area very small.

Thanks for the opportunity to relate Jim's experience.
John C.

 

[sitka]

Great story, I'm going to look for some at an art store. Hope they have a very small pack. I've heard that the reason it sticks so well is it actually starts sharing electrons (theoretical sharing based on probability distribution for valence particles) with the base material.

 

[PMFleXXX]

Wow, this sounds crazy enough to work

 

[TonyG]

I too am going to see if I can find some of the gold leafing. I think there might be some around the house. I doubt it will make much difference what thermal compound I use on my non-overclocked K6-2 450MHz.


Shucks, my mom is already in bed so I can't ask her if we have any.