Synthetic diamond heatsinks

[ace971]

Hi everyone, 

 

I have been interested in the cooling side of things for quite some time now, and still feel that there's a lot of room for improvement in that regard, even though things have moved on a bit over the last 15 years.

 

I have been playing around with liquid metals, lapping and delidding CPU IHS etc.

 

I found a company, that is actually starting to sell synthetic diamond square blocks of which dimensions vary from 10mmX10mmX0.5mm to 10X10X0.3 , which is obviously too thin to replace an IHS as is.

 

However the interesting part is its thermal conductivity, the lowest is rated for >1000 Wm^-1 K^-1 @293 K

The best >2000 Wm^-1 K^-1 @293 K

 

Link is there: https://e6cvd.com/us/application/thermal.html

 

Question is: 

 

Does anyone know if there are - I have searched in vain so far - already manufacturers using these type of synthetic diamond based materials for IHS, and if so, or if someone was willing to try to work on these, what would be the thermal improvements observed over copper based IHS (as a reminder pure copper is rated at approx 385 Wm^-1 K^-1 

 

^Thanks 

 

^Mederic

[Fasauceome]

You can't call yourself a true gamer until your rig is cooled by literal diamonds.

[Jurrunio]

I guess they didn't bother because it's so thin already? Besides, there's direct-die cooling for those who really want the best performance.

[zombienerd]

I would think that you'd still just be limited by your thermal interface.  Having a better conductivity on your IHS would be minimal, due to diminishing returns, I'd think.

 

As Jurrunio said, if you're worried about the IHS losses, just do direct die cooling.

 

Also, using the diamond layers as a TIM would be worse than using liquid metal or (I'd bet even good paste) due to it's inability to fill micro holes and surface imperfections.  I'd wager most attempts at this would show worse performance than a stock TIM.

[Radium_Angel]

13 minutes ago, fasauceome said:

You can't call yourself a true gamer until your rig is cooled by literal diamonds.

You'll note they list no price tags on their site.

In other words, "if you have to ask....you can't afford it"

There are a number of things to try over copper, but the price/performance ratio of copper is hard to beat, hence the popularity...

[ace971]

it makes sense from a price / performance standpoint.

 

At that stage I'm thinking it'd actually be better to have that directly on the chip to conduct heat out of it.

 

Now I can only imagine an aftermarket cooler with a diamond base, but again price would probably make this hardly conceivable at current times.

 

I actually saw price tags, of the order of 250 dollars for a 10mm X 10 mm X 0.3 mm piece of it, highly prohibitive.

[For Science!]

34 minutes ago, ace971 said:

-

The limit will still be the material-to-air inteface. Just like how full copper finstacks on heatsinks went out of fashion due to the increased weight and limited performance gain, and nowadays even the most premium aircoolers still use aluminium finstacks, this is presumably because air is such a poor conductor of heat.

 

Therefore no matter how much you improve the IHS or any step thereafter (liquid metal loops, all that crazy nonesense) ultimately since the radiator/finstack is interacting with air, that is the rate limiting step.

 

Unless you move away from air cooling (yes, watercooling is still air cooling), there is minimal rationale for this.

[ace971]

I see your point, however, and stop me if I'm wrong, the mass of the cooler, should also increase its thermal dissipation.

 

For instance, and my maths are totally off here but on a high level I remember it'd take like 100W to heat 1KG of copper of 1°C or something like that, therefore I was under the impression that we could always get the heat away from the die as fast as possible, and get it to accumulate, sort of, waiting for it to be dissipated, in the cooler, hence the reason of getting rid of bottlenecks as much as possible.

 

Now it is true that to that purpose, a direct contact to the die would address that, but again, what if we'd use almost perfectly flat diamond coating on the base of a cooler, with something like Liquid metal to fill in the imperfections...

 

It's so frustrating to have materials that can dissipate heat so well, but to still sort of be able to only transfer heat in such an inefficient way.

 

Next thing I'm thinking of is 3M Novec cooling liquid, but that obviously requires a lot more modifications and prevents then a computer from being moved around, as long as a quite expensive price for the liquid itself. 

I'm living in a pretty warm country, and would like to improve as much as possible heat dissipation so I don't have to use air conditioning to keep my machines at acceptable temps.

 

Right now I have a 2700X non OC , with its prism cooler, that idles at 40°C+, which is far from ideal. I ordered a Noctua C14S but the case being quite small, I'm limited in that regard.

 

Overall , these temps aren't much of a problem, since the max doesn't go beyond 80°C, but I'd like to get things optimally cooled, and satisfy my curiosity in that domain as well

 

Thanks for your feedback guys!

 

[For Science!]

1 hour ago, ace971 said:

-

Increasing the thermal mass of the cooler is "fine", but if you do not improve the efficiency at which that heat is dissipated, all you have achieved is a massive heat reservoir that takes longer to reach equlibrium but not any better in keeping the temperatures. A parallel to this is a watercooling loop with a tiny radiator but with a massive reservoir, it will take a long time for the loop to warm up, but ultimately will heatsoak and reach the same temperature with just a small radiator and no reservoir (assuming the reservoir isn't radiating the heat).

 

3M novec idea is a bit better, obviously people like derbauer are using waterblocks or GPU die blocks to increase the surface area of the die. Since this is no longer air limited (although I do not know what the thermal conductivity of 3M is) and also incoorporates elements of phase change, it may benefit - although I also doubht it since the fluid is likely the limiting factor here, so a diamond waterblock would make it reach equilibrium faster, but the position of that equillibrium may not be different.

 

Don't want to stop you from doing your experiment, but these are my opinions. Im out.

[Neo-revo]

8 hours ago, Radium_Angel said:

There are a number of things to try over copper, but the price/performance ratio of copper is hard to beat, hence the popularity... 

price is more so affected by abundance than quality.

 

but i feel you on this note.

[ace971]

I see your point and it is indeed valid. 

 

So in theory, assuming the same cooler on the other end, a diamond base would allow for a faster heat transfer from the die to the cooler, but then the cooler becomes the limiting factor which in a way is easier to address imo. 

 

I'm going to start delidding (not the 2700X, some other intels I have and a 2200G) then use the conductonaut, lap the IHS on both sides to remove the nickel layer then use their kryonaut for the cooler to IHS contact (I don't think there's much gain to be made using LM here from what I found online), I think that's pretty much, when a IHS is still needed, the most affordable and efficient way to transfer the heat. 

 

Can only dream of a real "liquid diamon-like" product one of these days  

[elcamino85h]

On 10/17/2018 at 11:45 AM, ace971 said:

Hi everyone, 

 

I have been interested in the cooling side of things for quite some time now, and still feel that there's a lot of room for improvement in that regard, even though things have moved on a bit over the last 15 years.

 

I have been playing around with liquid metals, lapping and delidding CPU IHS etc.

 

I found a company, that is actually starting to sell synthetic diamond square blocks of which dimensions vary from 10mmX10mmX0.5mm to 10X10X0.3 , which is obviously too thin to replace an IHS as is.

 

However the interesting part is its thermal conductivity, the lowest is rated for >1000 Wm^-1 K^-1 @293 K

The best >2000 Wm^-1 K^-1 @293 K

 

Link is there: https://e6cvd.com/us/application/thermal.html

 

Question is: 

 

Does anyone know if there are - I have searched in vain so far - already manufacturers using these type of synthetic diamond based materials for IHS, and if so, or if someone was willing to try to work on these, what would be the thermal improvements observed over copper based IHS (as a reminder pure copper is rated at approx 385 Wm^-1 K^-1 

 

^Thanks 

 

^Mederic

I would think a custom cut of marble stone 2in.x 2in.x 6in tall best natural heatsink. Stone is always cold