diamond thermal grease?

[ares350]

I was talking to my dad who about 3 years ago worked for a company who made diamond blades, and produced synthetic diamonds.

so I got to thinking; diamonds have the highest thermal conductivity known to man right?

so why not make a paste similair to artic silver which contains bits of solid silver. but use bits of diamond?

he said they have a cutting grease thats like a polishing compound type thing.

just got me curious if anyone has ever tried something like this as a thermal paste.

 

[Korndog]

this has been brought up a lot..
three points; the synthetic stuff doesn't have the same thermal conductivity as the reall stuff. Good luck trying to micronize(or whatever they call it ) it; its still damn expensive.

 

[MameXP]

Lets say :
Cpu: $150
HSF: $65
Thermal grease: $200

....... guh? something is not right here.

put a hole in your wallet : priceless

 

[ares350]

I didnt ask a price really... but thought not cheap, I was thinking 20$ an ounce. I have a handful of diamonds somewhere, ugly nasty little things. their not real high quality appearance wise but cheap.

I did ask about a huge heatsink block just for the sake of asking, he said the biggest they ever made was about 2carats. little ways off there.

 

[MassiveOverkill]

I was under the impression that Shin Etsu was composed of diamond material.

 

[man_utd]

There was that carbon black stuff a while back, that was made of something like compressed carbon (maybe not diamonds though?), but ended up barely beating artic silver origonal.

 

[PiSan314159]

They've gotten to the point where they can create chemically exact diamonds artificially. Hopefully in my lifetime they'll put that to use, I'm looking forward to diamond CPUs rather than diamond TIMs.

 

[Stupid Boy]

They'd probably scratch the heatsink and CPU.

 

[madcow235]

diamonds work so well because all the atoms are linked. They wouldnt work in tiny balls or whatever since they arent giant molecules they are just a bunch of tiny ones. If that made any sense i'd be suprised.
SAT's fried my brain

 

[Como]

why not a diamond heatsink base?

 

[sandman001]

Because diamond is actually incredibly rough, it has many many points like on an unfinished heatsink's base, just on a micron scale.

Simply put, they aren't flat....at all.

 

[Como]

...that too.

 

[kepten]

yes, a diamond coated heatsink base would be sufficient. and diamond coating is actually not a difficult job (depending on what material you want to coat).

that the diamond surface would be rough is not true. you can get diamond films on metals with a roughness of only tens of nanometers!

the diamond you would put down on the metal heatsink base however would have the thermal conductivity of a single crystal diamond but it would still be much, much better than any other metal including silver.

There is however the possibility to use what is called HOD (highly oriented diamond). It's just a film of diamond with the crystals all oriented in the same direction and it has a thermal conductivity that comes close to a singel crystal diamond.

The only difficulty is how to get the diamond film with a certain thickness on the metal. The main problem is that the thermal expansion of diamond is very, very small compared to a metal, especially the ones used for heatsinks like copper. So, if the diamond film on the copper is too thick it will peel off because of the stress in the diamond film.

Anyways, there are some methods how the stress can be released but I'm nit sure if anyone has really made a practical solution.

 

[sandman001]

that the diamond surface would be rough is not true. you can get diamond films on metals with a roughness of only tens of nanometers!

Really? I was always under the impression it would be insanely rough. So much so, that the added thermal grease would be enough to offset the thermal properties?

 

[kepten]

diamond films can be 'grown' on metal by what is called chemical vapor deposition and depending on your growth parameters you can get pretty smooth films.

what I would do is get the diamond film as flat as possible, even polish it so it's as smooth as a silicon wafer (like atomically flat) and also the cpu die ( I actually polished the rather rough die so it's completely flat) and then just press the heatsink on the die without any thermal paste.

If both surfaces are very flat you wouldn't need any thermal paste!

 

[sandman001]

I see, every interesting.

Wouldn't you still want some sort of thermal grease though?

I mean, one with a very liquidy consistency, I mean gettign it atomically flat is virtually impossible isn't it?

 

[I.M.O.G.]

Flat and smooth are two very different things... I'm wondering how flat the base and chip actually is. Mirror finish means nothing to flatness really, it just means it is smooth. That is a very smooth mirror finish it appears though. Depending on how you lapped the heatsink though, I wouldn't be suprised to find out it was convex if that were so.

 

[kepten]

true. maybe I didn't point out the difference between flat and smooth...

how flat the die really is depends on the silicon wafer used for making the cpu. As far as I know AMD uses 300mm wafers in their Fab.

Usually a 300mm wafer has a TTV (total thickness variation) of <= 2microns. But there are different 'grades' and you can get an order of magnitude smaller, like 0.2microns. But I have no idea what type of wafers AMD or intel are using for their processors...

 

[Xplos!ve]

ok, every1 stand up

now sit down if:

you HAVE NOT won the lottery
you HAVE NOT got 1 miilion dollars laying around somewhere
you are NOT a generally rich
you are NOT willing to sell everything you have

Congratulations of u are still standing you have just offered to buy the biggest and best diamond and see how well this theory works and post some results for us ASAP!!

 

[I.M.O.G.]

Theres a lot more variation in the base of the heatsink base though than there will ever be in the CPU die.