diamond waterblock concepts

[Natedog]

I am contemplating making a waterblock with a CVD or Bort diamond base insert for my p4. A 20mm radius circle or 13x13mm square would do. I am planning on using the diamond directly on the core(I took the P4 heatspreader off). I need info on where would be the best place to get them and what tickness sould I get the diamond? I was thinking of making the block with a 1mm base and have the water input shooting directly onto the ruff side on the diamond. Like cathars microchannel waterblock but without the channels.

[Joe]

One thing to beware of wiht CVD, or Graphite, or Carbon heat sink material is the heat moves very well in one one dimmension. There is very little if any heat spreading ability with either CVD or Graphite since the heat moves fast in only one direction and the other directions seem to act as thermal insulators. Now that Carbon Foam thing could be different, but I know from my email conversations wiht a MFG of CVD diamond plates for XRAY machines was that these used for a Heatsink/WB would be rather pointless and VERY fragile.

[Natedog]

If the water input is directly over the CVD then how could the awesome heat transfer (especially sub-zero) be pointless?

[Joe]

Umm why not just do it on the core then?

no heat spreading ability would make the actual heat density the same on the CVD as it is on the Core ( but you add another 2 layers of thermal resistance between teh coolant and the water.)

see pointless. Oh and CVD plates are killer expensive to get.

[airspirit]

Dude, if you want to blow money, just use silver. It'll give your thermal transfer a boost, and you won't be supporting the diamond cartels.

[Joe]

most CVD diamond is NOT natural diamond, its synthetic lab made diamonds. so it is not going in the pocket of dabeers or anything

[Natedog]

will the fact that this block will be used for a chiller system(-20c water) make any difference? What different considerations should I take when building a block for chilled water

[Volenti]

If you've already taken the heat spreader off you my as well do something like this.

with chilled water it would be awsome (don't go below 10 degrees (c) or you'll risk SNDS (sudden northwood death syndrome))

[Natedog]

Yeah I have been thinging of doing direct die also. I just dont know about the leaking problems that everyone who goes direct die tends to have.

[Albigger]

volenti - is urs o-ringed?

I would think, as long as you have proper and even pressure an 0-ring would be perfectly fine.

[redleader]

One thing to beware of wiht CVD, or Graphite, or Carbon heat sink material is the heat moves very well in one one dimmension.

Are you sure about that Joe? I'm well aware of this with carbon fiber and other sythetic carbon materials, but I've never heard of it with respect to CVD. Seems like that wouldn't be the case, after all CVD is like a normal diamond, just with a less perfect latice structure.

Am I confused?

[Joe]

The geeks I talked to said the lattice structure was essentially the same as real diamonds, and was meant to have heat/xrays/light transmitted straight through with no aberrations. (sp?) I asked about heat spreading ability and they didn’t seem to know more than that it was made to have everything pass straight through and not disperse.

[bigben2k]

Joe is right.

I'm looking at an add right now from graftech.com, and they're showing off their eGraf HS-400 material.

Thermal conductivity (in-plane): 370 W/mK
Thermal conductivity (thickness): 6.5 W/mK

Density: 1.94 g/cm^3 (lighter than aluminium)
Heat capacity: 846 J/kgK (Alu is 900, copper is 380)

[Volenti]

Originally posted by Natedog
Yeah I have been thinging of doing direct die also. I just dont know about the leaking problems that everyone who goes direct die tends to have.

You don't get leaking problems when the block is epoxied to the cpu

but yea that is a valid point, I just don't trust o rings in a suitation like this, a little lateral pressure from a hose can break the seal the o ring has.

Using a silicon sealant may be more preferable option for a removable direct die setup, some of the grades of silicon are strong and flexable enough to mantain a seal (with mechical assistance) while still being "weak" enough to remove without damaging the cpu.

[MadDogMe]

silicone strong?, yep!, don't see many fish~tanks falling apart! ...

[Volenti]

Originally posted by MadDogMe
silicone strong?, yep!, don't see many fish~tanks falling apart! ...

yea, so long as their assembled properly I've seen the results of a big marine tank letting go, it ain't pretty

Actually there is a whole range of automotive silicone sealants, many used for gaskets that have to seal yet be able to be removed periodically. one of them surely would be suitable for sealing a direct die block. hmm...

[nicozeg]

Silicone is used to hold glass in curtain wall facade systems (mirror like office buildings). In a hurricane the glass could collapse, but the edges are still firmly attached to the supporting frame.
To get this adhesive strength you only need very clean and degreased surfaces. In that case the only way of separating them is easy: cut the silicone from the side with a sharp knife.

[dream caster]

some links:
http://www.chm.bris.ac.uk/pt/diamond/othergps.htm Links to other groups studying CVD diamond
http://www.micrelinc.com/lines.html Micrel Associates, Inc. representants from some diamomd makers
http://www.p1diamond.com/terms.html P1 Diamond Inc.: Products and Services: Terms and Conditions - Affordable Industrial Diamond (my best bet to try buying from)
http://www.rhobest.com/eng/indexa.htm Rhobest
http://www.sp3inc.com./ sp3: advanced diamond tooling -also heatspreaders-
http://www.diamonex.com/ Welcome to Diamonex,

my current idea: to use 0.5 mm thick CVD diamond plate and make a version of Cathar's block with more channels (as fins will be half the thickness) with all heat transfer active parts made from diamond; using metallized surfaces in the inner side of base plate and edges to tin solder diamonds

[Cathar]

Hmmm, interesting.

My block really does focus most of the thermal spread in 2 dimensions, which would somewhat suit a substance like what's being mentioned above. However, we can't underestimate the need for side-ways thermal spread since even if the heat does spread up and out along a plane (ie, up the fins), it still needs to move sideways to make contact with the fin->water interface (wall).

Very interesting substance to be working with.

[Joe]

Or you could just buy an Alpha 8045 and a Vantec Tornado fan and be done with it hehehehe

Just had to take a HSF jab in here

[nicozeg]

Originally posted by Cathar
However, we can't underestimate the need for side-ways thermal spread since even if the heat does spread up and out along a plane (ie, up the fins), it still needs to move sideways to make contact with the fin->water interface (wall).

An easy solution to that is tapering the fin surfaces, allowing every conductive plane to touch the water.

[bigben2k]

Do you have any idea how hard it would be to taper that stuff? Never mind the size, it's basically a carbon fiber process, which can be as hard as steel.

Anyways, the heat transfer property would be proportional to the angle of the cut. If the cut was 45 deg, then (for example) the eGraf-400 stuff's thermal conductivity would be right in the middle between 370 and 6.5 W/mK, which would be 188.

I think that this stuff would be best used to transmit heat to a copper plate that sits higher than the normal fin area. It would basically allow a designer to have a much larger surface area, without impeding flow.

[dream caster]

Originally posted by bigben2k
Joe is right.

I'm looking at an add right now from graftech.com, and they're showing off their eGraf HS-400 material.

Thermal conductivity (in-plane): 370 W/mK
Thermal conductivity (thickness): 6.5 W/mK

Density: 1.94 g/cm^3 (lighter than aluminium)
Heat capacity: 846 J/kgK (Alu is 900, copper is 380)

Graphite has a layered cristalline structure; if you look at graphite heat conduction values all of them come with values for conduction parallel or normall to the layers; diamond has no layered structure and heat conduction in diamond is isotropic; this is it is equal in all directions. Some anisotropic properties though may be introduced by the orientation and shape of grain interfaces and different materials on them (graphite is formed there); this asymmetry (when it exists) depends on the method used to make CVD diamond.
P1 Diamond Inc. in its engineering notes mentions CVD diamond heat conduction as isotropic and lists its heat spreaders as having 10W/cm/ºK in their conditions and terms; diamonex list heat conduction as 13W/cm/ºK; natural diamond has a somewhat higher heat conductivity.

[jaydee]

Originally posted by Volenti
(don't go below 10 degrees (c) or you'll risk SNDS (sudden northwood death syndrome))

If this is going to be a rule then don't bother with Diamonds. Copper will work just peachy and save you a hell of a lot of money. Diamonds seem way out of line for this situation but this is just my opinion. Why have a cooler worth more than computer it is cooling and for what? A few more mhz? Wish I had that kinda money to piss away.

Non the less good luck!

[Joe]

Originally posted by dream caster
Graphite has a layered cristalline structure; if you look at graphite heat conduction values all of them come with values for conduction parallel or normall to the layers; diamond has no layered structure and heat conduction in diamond is isotropic; this is it is equal in all directions. Some anisotropic properties though may be introduced by the orientation and shape of grain interfaces and different materials on them (graphite is formed there); this asymmetry (when it exists) depends on the method used to make CVD diamond.
P1 Diamond Inc. in its engineering notes mentions CVD diamond heat conduction as isotropic and lists its heat spreaders as having 10W/cm/ºK in their conditions and terms; diamonex list heat conduction as 13W/cm/ºK; natural diamond has a somewhat higher heat conductivity.

Hmm Good info to know. Maybe my udnerstanding of what the company I talked to was off.