Graphite as a thermal interface material?

[Cathar]

Maybe not quite the right sub-forum, but many of Pro-cooling's best minds gather here and I wanted this to reach the right audience, if only to tell me that I'm being silly.

How would graphite go as a thermal interface material?

Would grabbing a 6B pencil and scribbling all over the CPU core until a substantial even layer was produced act as a semi-effective thermal interface material?

Some links to ponder on:

http://www.electronics-cooling.com/h...techbrief.html

http://www.e-insite.net/semiconducto...cleid=CA164440

[hara]

I suck at explaining this scientifically: I think that for graphite to conduct heat it has to be processed in such a way that each layer is in the same direction. Something that has to do with the arrangement of free electrons. Also, it has to be solid and processed in a special way to conduct heat. Scribbling will not do the job.

At least that's what I've been told.

[Since87]

Before you take the pencil to the CPU...

Pyrolytic graphite sheet is available from Digi-Key. The good thermal conductivity is in the plane of the sheet, rather than through the thickness though.

There is information on the bottom of this Digi-Key catalog page.

Digi-Key sent me the attached file as additional information from the manufacturer.

[thetruememphis]

Graphite is a great conductor, but the heat that the graphite would be a problem, maybe mix a small amount of graphite with something else.

[Cossey3]

the top link Cathar posted shows why it wouldnt work, the thermal conductivity of copper is 400W/m/k (higher is better) for layers of graphite it is 2000 along the layers and 10 across them. putting layers on with a pencil will result in layers parallel to the cpu so it will be acting as a insulator rawther than a conductor. to make it work you would need layers coming up vertically from the cpu which would need to be grown on the cpu as it would be too difficult to place them on it. you could try getting hold of some silver foil and using that compressed between the cpu and block.

[Since87]

Originally posted by Cossey3
the top link Cathar posted shows why it wouldnt work, the thermal conductivity of copper is 400W/m/k (higher is better) for layers of graphite it is 2000 along the layers and 10 across them. putting layers on with a pencil will result in layers parallel to the cpu so it will be acting as a insulator rawther than a conductor. to make it work you would need layers coming up vertically from the cpu which would need to be grown on the cpu as it would be too difficult to place them on it. you could try getting hold of some silver foil and using that compressed between the cpu and block.

I presume there must be some value to the pyrolytic graphite sheets that Panasonic/Matsushita are selling through Digi-Key for $30 for a 60mm X 90mm sheet.

A thermal conductivity of 10 W/m/k puts the material in the same ballpark as high end thermal paste for CPU to WB heat conduction. Since CPU's tend to have hotspots, the high lateral thermal conductivity of the material may give a significant advantage as a heat spreader. I would guess it can't fill surface irregularities as well as paste though.

I'd be curious to find out what kind of results someone got with this stuff on an Athlon. (I would guess it's much less useful with a P4 due to the IHS.)

[Since87]

More data on the Pyrolytic Graphite sheets.

[Cossey3]

if you have 2 very lapped sufaces it would be worth it but even with a highly lapped waterblock there will probably be too many irregularities in the cou to make if work better than as3. if your prepared to lap the cpu top to get it very flat then its worth considering

[bigben2k]

Here's Panasonic/Maco spec sheet, for this PGS.

[edit]It's rated at 0.4 C/W, so for an 80 Watt source, the temp gradient would be 32 deg C ???

Digikey stocks it, but p/n EYG-S060910, which is 60 * 90mm (smallest available) costs $29.88 .

[Since87]

Perhaps this material could be thermal epoxied to the top of a CPU, and then this material lapped. (Possibly important to look at differences in coefficient of thermal expansion for PGS and silicon.)

I'm thinking that this stuff, might be able to improve overclocking by spreading heat laterally from hotspots, even if the ondie temperature measurement remains the same, or even increased. (Somewhere I saw a depiction of the thermal gradients across a die, and the hotspots were fairly localized.)

[Since87]

Originally posted by bigben2k
Here's Panasonic/Maco spec sheet, for this PGS.

[edit]It's rated at 0.4 C/W, so for an 80 Watt source, the temp gradient would be 32 deg C ???

Digikey stocks it, but p/n EYG-S060910, which is 60 * 90mm (smallest available) costs $29.88 .

Yeah, I've had a hard time coming up with a practical use for this stuff. The 'heat pipe' type usage shown in one of the documents I linked might be worthwhile in some really space constrained situations, (laptops) but I haven't figured out any way to make use of it for desktop CPU cooling.

[bigben2k]

Maybe it's just me, but I don't see adding a 32 deg C gradient as a good thing. Then again, as long as the CPU temp is low...

Another thread on the subject of TIM joint.

Seems to me like thermal epoxying this stuff would add another 16 deg C to the gradient, which puts us up to 48 (?).

[Cossey3]

thats 0.4 along the sheet isnt it so its over a much larger distance so it wont be anything like 48oc across the sheet.

[bigben2k]

Originally posted by Cossey3
thats 0.4 along the sheet isnt it so its over a much larger distance so it wont be anything like 48oc across the sheet.

Are you sure?

There's another rating: 600 to 800 W/m*K, but I can't figure out the temp gradient with that:shrug: The film is 100 um (0.1 mm) thick.

[Groth]

Originally posted by bigben2k
It's rated at 0.4 C/W, so for an 80 Watt source, the temp gradient would be 32 deg C ???

You need to look at how they can up with that number. Thermal resistance of PGS

°C/W numbers apply to a specified system Note that they use a transistor - small contact area. Note that they use a piece of PGS only as big as the transistor. Note that they include passive cooling by a heatsink as part of their result.

What we need to know is the thermal conductivity perpendicular to the sheet. Then we can derive a °C/W for a TIM joint.

[leejsmith]

this is a little out of my league but what about diamond paste
isnt diamond the best conductor

some details i found

Synthetic diamond compounds or slurry can be used for rapid stock removal (90micron) to superfinishing (1/15 micron).
The formulation uses De Beers™ synthetic diamonds micronised into an octohedral shape for consistant long lasting cutting performance. All compounds are oil and water miscible and can be lubricated with RS diamond lubricant.
Suitable for use on Ferrous metals, Tungsten Carbide, Steel, Stainless steel, Glass, Synthetic and natural Precious stones.
All Syringes contain 5g of synthetic diamond.
Diamond compounds and Syringe labels are colour coded to enable grade recognition, the coarse grade being white (90 micron).
selection guide


would the 0- 1/15 micron one be ok ?

it's £11 for 5g thats not too bad

[Groth]

Nevermind. I went back and re-read that datasheet.

They weren't including the heatsink in their calculations. Assuming that they used a standard TO-220 sized transistor, the PGS would have a thermal conductivity of around 1.67 W/mK perpendicular to the sheet. That result jibes with other figures I've seen - the product of the planar and perpendicular conductivities in the 1000-1200 region.

That puts the thermal impedance across a T-bred die at 0.6 °C/W. You'd need a helluva good scheme for the lateral heat spreading ability to make up for that.

[Gareth]

What about using gold leaf?

i know its very fragile but if u was careful and used a layer, its very soft so should 'mould' to the surfaces of the cpu and heatsink/waterblock nicely

[Cossey3]

0.078125w/m/c across the sheet so 6.25oC for 80w over 128mm2

[HighFlowRod]

Didn't see anyone metion this: how about graphite lube (the stuff that lubricates keylocks). Usually found in any automotive section of any store. Maybe worth a try.

[Cathar]

Originally posted by HighFlowRod
Didn't see anyone metion this: how about graphite lube (the stuff that lubricates keylocks). Usually found in any automotive section of any store. Maybe worth a try.

I mentioned that at the AMDMB forum thread where this post came from. I thought perhaps mixing it in with a small amount of synthetic oil to form a paste.

It would be interesting to see if it was possible to cheaply "roll your own" thermal paste using a bottle of graphite lube and some grease, and have it match or outperform some of the thermal pastes on the market.

That's basically where this idea was coming from. The micronised graphite lube should fill the pits and gaps pretty well.

[utabintarbo]

Originally posted by Cathar
I mentioned that at the AMDMB forum thread where this post came from. I thought perhaps mixing it in with a small amount of synthetic oil to form a paste.

It would be interesting to see if it was possible to cheaply "roll your own" thermal paste using a bottle of graphite lube and some grease, and have it match or outperform some of the thermal pastes on the market.

That's basically where this idea was coming from. The micronised graphite lube should fill the pits and gaps pretty well.

No way to "line up" the crystals. See above.

Bob

[garbage!!!]

Grafite cunducts heat by fonons( Vibrations of the whole atomic net, not only eletrons) that is why it only conducts heat well in one direction( diamonds do not have this problem because it has a proporcional unit cell).I think it would not be a good ideia to use pencil or grafite lube as thermal interface since the planes are all going to be mixed up.
Atentiously
Guilherme Cunha Prada

[HighFlowRod]

Originally posted by Cathar
I mentioned that at the AMDMB forum thread where this post came from. I thought perhaps mixing it in with a small amount of synthetic oil to form a paste.

It would be interesting to see if it was possible to cheaply "roll your own" thermal paste using a bottle of graphite lube and some grease, and have it match or outperform some of the thermal pastes on the market.

That's basically where this idea was coming from. The micronised graphite lube should fill the pits and gaps pretty well.

You could crush the graphite structure with a pestle and mortar and mix it with some conductive gel or liquid. Maybe?

[Cathar]

Originally posted by utabintarbo
No way to "line up" the crystals. See above.

Bob

Yes, I know that this is the case, and it won't be anything like 2000WmK, more like 10-20WmK, but the question remains, "is that good enough"?

Would the resultant paste match, say, Arctic Silver for performance?

One way to find out. Heading off to buy some graphite lube and give it a burl.


burl - Aussie euphamism for trying to do something. Also an Aussie euphamism for driving a car as in "I'm goin' for a burl in me new car!"