Arctic Silver 5

[Admiral]

Originally posted by Kurt
You could try using a mix of tin and lead, by warming up the heatsink enough you might attain a reasonable state of heatsink-cpu fusion. No more air bubbles! Even beats toothpaste!!!

These guys seem to be on to soldering/fusion.

[Jon P. Inghram]

I wonder if you could actually polish the headspreader and bottom of a heatsink well enough to properly cold weld them in a vacuum? If you could that'd be hard to beat unless you plan on removing the heatsink afterwards of course.

[Greebe]

personally I wouldn't bother to try anything else and just wait for Diamond substrates to be utilised in future procssors. Temps then can be upto 500f before breakdown and have somewhere around 5x the thermal conductivity of Silicon (currently under developement for use in space where ambient temps typically destroy todays tech without additional mass of all the neccessary cooling equipement

carbon nanotube HS's have a future here too

[Brian R.]

As far as metals foils go, I think gold would be best because of its maleability. A liquid metal may not wet the silicon surface - unknown problems there.

Forget exotic materials for a $200 cpu. Their utility is pure speculation.

[spadnos]

On the subject of foils: I'm not sure the malleability of the metals matters much. It's true that you can put a dent in a gold bar if you hit it with a hammer, and you won't dent a steel bar - but the forces involved are still huge compared to what a human can generate without tools. What heatsink clip attached to a motherboard can withstand that force, let alone the actual chip you're attaching the heatsink to.

There's also a difference between "deforming" and "conforming". Look at the amount of force used to mint coins, and you'll see what it takes to conform metals to a shape. That's a surface with large features (easier to conform to, because the bend radius - for lack of a better term - is larger), we're talking about microscopic features on a more-or-less smooth CPU and heatsink.

I'd say pastes or liquids are still the best bet.

Now - for the radical idea: Instead of sealing up a chip with solder between the heatsink and the core, how about just making a water seal, and directly pump that water around as your coolant? Basically, get rid of the water block, and glue the pipe directly to the CPU...

(of course, I'm not going to try that on any machine of mine )

- Steve

[az]

Actually, you can dent gold with your fingernail, and your watercooling idea has been done already

AZ

[Brian R.]

24K gold is very soft. May or may not work, but I will try it someday when I am working on a non-critical cpu

[cjolley]

It's hard to inagine it would actualy flow as it would need to do, but, you never know.
Or at least, I never know...
chuck

[Brian R.]

If it can touch 10% of the surfaces and offer 10x the heat transfer per sq in, it is fine.

[Marshmallowman]

I have read/seen a web site where some guy actualy did use gold leaf as the interface material.
He actually used 2 gold leafs, one on the core and one on the heatsink and then press them together, the hope being the gold would fill the irregularities in the surface but conduct better than any kind of metal suspension(HS goop)

As I vaguly recall it worked reasonably welll, but the performance was not worth the effort I think AS or ceramic based ones are as good as you need now.

The difference between a reasonable and a good thermal goop is only 1% to 2%.

We have actually had this discusssion before, and the general consensus was ...we would all like a diamond heatsink, but that is not going to happen any time soon..

But since diamond manufacturering has leaped forward, the possibility of a CPU made direclty on to a diamond heatsink, (or diamond thermal stub, for mounting onto a heatsink) is actualy looking like a real possibilty.

[Brian R.]

I guess I wasn't looking for great, just easy and neat. If the foil doesn't stick to the core, it might be reusable too.

[Greebe]

Gold leaf is bloody fragile. Used some in the past (assisted a friend who does art restoration one weekend). Statically charge the gold so it would cling to the surface while sliding the leaf off it's carrier (called guilding). Then ultra lightly tamping down with a fine natural fiber brush. Any abnormality would cause the leaf to flake into the air and tiny bits which had to be dusted away. I can't possibly see how it could be reused (other than possibly recycled). It wouldn't be competitive to compounds, but it sure would look nice... 25 leaf 23k books run somewhere around $30 (~2 square feet)

[Brian R.]

Originally posted by Marshmallowman
I have read/seen a web site where some guy actualy did use gold leaf as the interface material.
He actually used 2 gold leafs, one on the core and one on the heatsink and then press them together, the hope being the gold would fill the irregularities in the surface but conduct better than any kind of metal suspension(HS goop)

As I vaguly recall it worked reasonably welll.

Sounds like fun anyway. Wonder why they didn't use just one layer...

As far as reuse of the resulting gold layer, I guess you just have to do the experiment and see. Speculation is useless. There is a large difference between gold foil and gold leaf. I initially thought of foil, but leaf is interesting too. At some thickness, the gold layer may be reuseable, at least from a fragmentation standpoint.

[cjolley]

I've worked with gold leaf in the past.
It would not be resuable.
Foil might, but leaf - definately not.
It would be like trying to reuse paint.
chuck

[Brian R.]

Foil is definitely the way to go if resusablilty is importent. I agree. What is the exposed surface area of one of these new Intel cpus?

Another interesting (to me) possibility is plating a gold surface on the heat sink. Probably not reuseable, but it might be very efficient from a heat transfer standpoint. You could mask off the areas of the heat sink on which you don't need deposition.

It is also easy to deposit a silver mirror, either chemically or plated electochemically...