Copper-based thermal paste

[Loknar]

Graphite is used to reduce thermal conductivity usually, as it is highly athermanous - much like any carbon-based substance. But then again I don't know what you mean by "suspended", Necrotech.

 

[silverpig]

Graphite's thermal conductivity depends on orientation. It's basically a bunch of 2D layers. The conductivity is great along the layers (ie in x and y), but conduction through the layers (ie in z) is terrible.

 

[Loknar]

Yes but...

You may consider the anistropic nature of graphite in both thermal and electric properties. The production of perfectly aligned carbon planes is nearly impossible. Synthetic graphite is highly porous - thus we are unable to find a consecutive plane longer than 1mm is impossible. Using natural graphite, I still guess it is difficult to engineer a heatsink using this material - any mistake would greatly decrease heatsink performance;

Why are engineers ignoring graphite? Perhaps the real-world effectiveness of metals is better than graphite in the chip-cooling scenario (low temperatures (graphite is usually used as a conductor in furnaces >2000"C where is it considered a good conductor since metals would have molten at that temperature), uneven heatsource, nature of media (PCB, air))... your thoughts?

Its still interesting to imagine how an ideal graphite heatsink would perform.

 

[mmonnin03]

Ideal graphite (carbon) would be diamond. Now if we were able to create chips from diamond (unfeasible), we would not nead a heatsink at all.

 

[Loknar]

Diamonds, most definatly! It can be synthetic diamonds, easier to shape to design specification. I certaintly hope they will be embedded in the chip design soon. I guess it would dissipate heat better and operate at higher temperatures.

We have to say conducting the heat is one thing; getting rid of the heat is another. This is why liquids, heat pipes or phase change materials are interesting in spite of lower thermal conductivity mesurement. Annoyingly, we still need airflow to cool any heatsink. I would love to see PCs that use a big water tank for cooling all components - I'm so bored with noisy fans. I'd dunk my motherboard in an aquarium but I feel lazy to unscrew it from my case.

SLIGHTLY BACK TO TOPIC: I hope diamonds will be used in the near future but also, that the CPU chips become embedded in their heatsinks (every layer of material slows the heat dissipation). This will eliminate the need for icky heat paste in the first place - creating all sorts of compatibility/standardisation issues but - with performance in mind!

 

[AmberClad]

So, I think the general consensus then is that silver is a better thermal conductor. But I'm still not sure of the actual silver content in thermal pastes like AS5. My understanding was that only a tiny proportion of the stuff is silver, with the rest being aluminum and other materials.

But in any case, this brings up another question -- if silver is so popular in thermal pastes, why don't we see HSFs with silver-plated bases?

 

[mmonnin03]

Easy...cost. Look at how much silver jewelry costs. Its not going to happen.

 

[silverpig]

Originally posted by: mmonnin03
Easy...cost. Look at how much silver jewelry costs. Its not going to happen.

Silver is pretty cheap. It wouldn't cost that much, especially when people already shell out $80 for a hsf combo, or hundreds for water/phase change.

Besides, there was a silver based heatsink on the market a while ago called the silverado I believe. It had a large nearly pure silver disk which contacted the cpu.

 

[Nathelion]

My impression was that the bottleneck in modern cooling was exchanging the heat with the environment, not getting it from the processor to the heatsink. If that's correct, then that silver bottom plate would be pretty useless unless all the fins were silver as well?

 

[Loknar]

Originally posted by: Nathelion
My impression was that the bottleneck in modern cooling was exchanging the heat with the environment, not getting it from the processor to the heatsink. If that's correct, then that silver bottom plate would be pretty useless unless all the fins were silver as well?

Correct, adding a plating of silver does not improve thermal conductivity - making it worse probably. Its quite expensive (no- I'm not paying 10$ per ounce) but I think the main reason is that its not as strong as copper or Aluminium; Its highly malleable which can be a design flaw (read: liability for a customer's CPU failure). Gold has the same problem but worse (more expensive, more soft).

You have to appreciate the good balance of properties in Alluminium and Copper for HSF designs. As seen in Anandtech's benchmarks, the heatsink material matters less than the overall design/shape.

 

[silverpig]

Originally posted by: Loknar

Originally posted by: Nathelion
My impression was that the bottleneck in modern cooling was exchanging the heat with the environment, not getting it from the processor to the heatsink. If that's correct, then that silver bottom plate would be pretty useless unless all the fins were silver as well?

Correct, adding a plating of silver does not improve thermal conductivity - making it worse probably. Its quite expensive (no- I'm not paying 10$ per ounce) but I think the main reason is that its not as strong as copper or Aluminium; Its highly malleable which can be a design flaw (read: liability for a customer's CPU failure). Gold has the same problem but worse (more expensive, more soft).

You have to appreciate the good balance of properties in Alluminium and Copper for HSF designs. As seen in Anandtech's benchmarks, the heatsink material matters less than the overall design/shape.

No, it definitely helps. There are two main bottlenecks in heat transfer. The first is cpu to heatsink, and the second is heatsink to the air.

The silver is very soft and deforms to fit the cpu surface well.

This is something that is very important as any space between the two surfaces, no matter how small will kill the transfer of heat. Just look at how well lapping your hsf works.

 

[Biftheunderstudy]

I've been wondering about this lapping business. From a thermodynamics point of view, having the surfaces perfectly flat and no grease would be optimal. But since perfect flatness is not really attainable in practice, wouldn't a rough surface with a very fine conductive material in between work better? I mean that a really rough surface has a way higher surface area than a flat one and as long as the grease has a higher conductivity than the heatsink rough>flat.

I know that in general this is true, matte surfaces dissipate heat faster than shiny ones for this very reason--all you would have to do is exaggerate it a bit.

This may be true but things don't always happen like theory says they should, just thought I'd throw it out there since I'm curious and all.

 

[mmonnin03]

Having a rough surface area only works if the opposing material is in contact with ALL of the 'rough' surface. Best heat transfer is from 1 atom straight across to another atom on the heatsink, no transfering to thermal paste or air. A rough surface area only works better if both surfaces have the same roughness, they match exactly. Rough works with a metal and air since air will form exactly to the roughness of the metal. Think of fins in a heatsink as being the extreme roughness, so rough there are holes in it for air to pass through. Thing is with having two rough surfaces is that passing heat across a thermal paste lowers the efficiendy a lot, probably enough to negate the extra surface area from being rough.

 

[Nathelion]

I'm not sure a "rough" surface carries with it any benefit at all unless the medium it makes contact with convects.

 

[Biftheunderstudy]

In reality, rough is not better than smooth. Apparently the thermal conductivity of a joint goes something like k='RMS roughness'/LA or some such, where L is the distance between the joints and A is the surface area. I'm not sure how the rms roughness is calculated, but there have been models done. (With thermal grease, this changes somewhat)

In theory however, since the surface area does depend on the roughness of the surface, if the rms roughness remained low and the surface area due to roughness went up, the thermal conductivity would go down. This may be possible I think if the roughness were regular(maybe).

Not too sure though, I'd have to think on it some more.

 

[Nathelion]

I did give this quite a bit of thought before posting above, and concluded that I essentially don't know how to analyze this. That being said I can make a couple of hopefully useful points.
First and foremost, if one assumes that air does not convect, then fins are fairly useless so long as the heatsink is volume-limited, as opposed to mass limited. To illustrate this, imagine a sphere filled with fins and a sphere filled with solid metal. Let's assume the heat source is a point source at the center of this sphere. Now air is a very bad thermal conductor, so the lion's share of the heat will pass through the metal fins. Now since conductivity is proportional to surface area, a spherical cross section of the solid sphere at any given depth is going to be higher than a corresponding cross section of the sphere with fins.
However, the above heatsinks use vastly different quantities of metal. Constructing an array of fins only requires a fraction of the material that the solid sphere does. The real question here is rather: Given the same amount of metal, does it pay of to construct a smaller solid sphere, or is it better to build a larger array of fins? In this case the conceptual sphere would be made out of IHS material, while the surrounding medium would be copper or aluminum (as opposed to air). I don't really know how to answer that.

 

[Loknar]

A very wise post, Nathelion.

You illustrated the importance of dissipation very well. In spite of theoritical "heat conductivity" properties, the practical facts lead to our common two/three-stage heatsinks. The amount of metal, shape, thickness of fins, space between fins and airflow power are factors that need to be balanced to create a good cooling solution. So many factors in play lead us to an extreme variety of HSF on the market.

But because dissipation is more important than conductivity (eg: getting rid of the heat vs diamond-made heatsink) we see more and more heatpipe designs lately. And I would like to see more and cheaper liquid cooling solutions out there:

http://www.xoxide.com/thermalt...symphony-cl-w0040.html
Even though these liquids are much less conductive than metals, these cooling solutions are more efficient because the heat can be spread to a wider area faster. Please share your links anyone on affordable liquid or phase-change solution.

 

[Biftheunderstudy]

Thats pretty much it. In a HSF the most critical part is the rate of heat transfer, since the rate is linear in all of its terms raising any one of them by the same amount will produce the same results. I can't recall the exact formula, but its something like r=kvA/d with some other factors probably, but you can see the 'important' one here. k the thermal conductivity constant, d the distance between heat source and heat dissipation medim, v the air(or other medium) velocity, and A the surface area. One thing to note is that if you make high surface area to volume ratio conductors, d increases linearly but surface area goes like d^2 so your r rises--this is the whole idea behind HSF's.
So the whole roughness thing comes from this same idea, if you can raise the surface area of the joint, you can raise the rate of heat transfer across the joint so long as the rms roughness only increases linearly.

P.S.
You can also see that frying pans with copper bottoms and things like that are actually killing your heat transfer since d increases without increasing A. I think its for spreading heat equally though.

 

[mmonnin03]

First part is fine....but 'roughness' or more surface area only does you any good if you are making metal to metal contact along the entire curvature of CPU core to heatsink. No one is going to match grain for grain, but having 2 smooth surface is much easier at getting the most contact from CPU->HS which is the context of the original term of having a rough surface. Indeed having a more surface area, aka fins, will help with transfering the heat off of the HS and into the air.

 

[silverpig]

Originally posted by: Loknar
A very wise post, Nathelion.

You illustrated the importance of dissipation very well. In spite of theoritical "heat conductivity" properties, the practical facts lead to our common two/three-stage heatsinks. The amount of metal, shape, thickness of fins, space between fins and airflow power are factors that need to be balanced to create a good cooling solution. So many factors in play lead us to an extreme variety of HSF on the market.

But because dissipation is more important than conductivity (eg: getting rid of the heat vs diamond-made heatsink) we see more and more heatpipe designs lately. And I would like to see more and cheaper liquid cooling solutions out there:

http://www.xoxide.com/thermalt...symphony-cl-w0040.html
Even though these liquids are much less conductive than metals, these cooling solutions are more efficient because the heat can be spread to a wider area faster. Please share your links anyone on affordable liquid or phase-change solution.

Heatpipes aren't used for dissipation, they're used for conductivity. It's a mechanical rather than thermal mechanism, but all they do is transport thermal energy.

 

[AmberClad]

If surface area is the key to heat transfer from the heatsink, then aren't there even more effective ways of increasing surface area than fins? Layers of high density mesh/grid for example. That'd be harder to fabricate than stamped metal fins, I assume?

 

[mmonnin03]

Air must be able to flow freely enough between the fins or layers of mesh. And I would think mesh would create more noise, like a whistling effect.

 

[silverpig]

Originally posted by: mmonnin03
Air must be able to flow freely enough between the fins or layers of mesh. And I would think mesh would create more noise, like a whistling effect.

Well that, but you also have to get the heat efficiently out to the fins. It'll travel better down a solid rod than down thin mesh.

 

[mmonnin03]

I never mentioned moving heat across the medium but from CPU->HS and HS->Air.

But yes, a solid object with a cystalline structure will work best.