The most efficient way of watercooling?
 

There is an ongoing discussion in a swedish oc forum (sweclockers.com) about how efficient direct die cooling is and that it should be alot better than using something inbetween to spread the heat over a larger surface (ie. a waterblock).
There is a claim that cooling a peltier with the same technique should be better too...
I just wanted you all to vote for the most efficient way of watercooling something hot. :)

Please post pros and cons about the diffrent cooling methods that might come in mind. :cool:

and the MCW50-T is a commercial product whose TEC is cooled directly

Yeah, but is it better than blocks that spread the heat over a larger surface area?
The water must be easier to heat up if you have a design that consists of a large amonts of pins instead of a flat surface (when cooling a TEC).

I did'nt know that.................off to swiftnet

so you mean that it does not use a coldplate, it is block direct to pelt, right?

The waterblock's base is the TEC, water actually hits the hot side of the peltier. It still has a cold plate under it.

What about the lack of clamping pressure on the tec? Do you not need it if it is directly in the liquid? Last I heard TECs need clamping pressure alot more than say a cpu.

Don't remeber where I read it but i beleive there is an issue with water slowly saturating the die shorting it out... any truth to that?

Yeah, the water eventually eats away a the core. Direct die would be more effective with tons of water velocity, but so far nobody has found a very good and reliable system. As of now, having a lump of copper on top seems to work the best.

I'd think a water block would be better because you're spreading the heat over a larger area.

Also copper is a much better heat conductor then water (I think like 400 W/m.k vs .58 W/m.k).

Also water is a much better conductor then air.

On a large surface, I beleve direct die would be the way to go if there is no other destruction, in a situation like on a peltier is what I would be talking about. A cpu on the other hand has such a high heat density that the heat needs to be spread out a little bit to cool efficiently. The cpu is a peice that can be destroyed because of water soaking into the core or substrate, also, unless you completly bond the block to the cpu, it makes quite the mess to remove.

Jon

Pretty much summed up my experience with it. Excpet I used RTV silcone to seal and didn't have any problems taking the block off the CPU.

Thread: http://forums.procooling.com/vbb/showthread.php?t=7698

If we could figure a way to "tin" the core somehow...
maybe with silver... then that problem would go away...
I don't think that there would be any benifit to spreading the heat with a block prior to using water.
a hot spot on the core could be minimized quicker with direct die cooling...
but untill the issue with the core being eaten away by the water is addressed,
i will have to vote for using a water block.
my 2¢

There were a few that were actually trying to plate the core with some copper or silver, I dont know how well it went. I know neomoses was giving it a shot and another guy.

Jon

I guess the best thing would be to start with a CPU that has an IHS, most are fairly thin and usually made of copper. You could probably even put an o-ring right up against it. Get a nice powerful jet over the core and youre set :D

Direct water is problematic, and better.

Noodle, if that was the case, make a block with a empty thin bp.

Jon

since the die is so small wouldn't it amke more sense to distribute the heat?

Given the same heat transfer area and wattage copper would conduct heat better. So to make up for water's lower thermal conductiviity by using copper to spread the heat right?

You'd need a very high flow rate to compentsate or the smaller area wouldn't you?

Maybe the thermal paste needed inbetween the block and the processor makes this less apparent?

Now adding an IHS would be interesting.

I'm looking at Fourier’s Law:

q = k A dT / s
where

q = heat transferred per unit time (W)

A = heat transfer area (m²)

k = thermal conductivity of the material (W/m.K or W/m.°C)

dT = Temperature difference across the material (K or °C)

s = material thickness (m)

I could be wrong i don't know much about heat transfer.

unfortuatly I'm not good at math... or in this case arithmetic...
what i do think is that having a piece of copper sitting on the core...
won't the copper be a consistent temperature?
even after the water removes the heat from it the copper,
won't the copper's temperature will still be warm?
with direct die approach the water is cooler and is remove immediatly...
heat transfer of copper may be better but the copper isn't phisicaly moving...
the water is... does that make any cents? :shrug:
perhaps someone may be able to explain what i'm thinking better...

Cant put enough mounting pressure on it, and why do this when it would be more effective just to have a bare core.

I said that because of the detrimental effects of cooling the core directly.

If you had a powerful enough jet on it nothing could beat a direct die logically; at one point the benefits of copper would be outweighed by the thermal delta between the point of contact with the core and the dissipation area.

Am I wrong?

EDIT: I guess I'm ignoring the question, this method is not the most effecient, but probably can achieve the lowest possible temperature.

Problem here is the IHS is designed to have a center force on it and distribute the weight evenly across the IHS so it will make contact with the CPU die below it. The IHS is not directly attached to the die and needs pressure from the HS to be effective. What I thought about doing was making a base simialr to the Cascades but drill the holes all the way through and use the IHS as the base. The water would hit the IHS and also get a secondary cooling fro mthe sides of the holes in the copper. But that still isn't direct die cooling.

Direct Die cooling shouldn't even be worth discussing anyway untill AMD or Intel design a CPU that is direct die cooling compatible. Not to mention there is no proof (that I have seen) that direct die cooling is any better than a decent water block. I had a pretty good Direct die cooler with jets and it was no better than a Maze 4. Just not enough room to work with especially on the outlet part of it.

What if a design like the storm was utilized for a direct die cooler. Isnt the basic principle to use the jets to remove the boundary layer on the bottom part of the cups? Perhaps you can put an open honeycomb shape directly against the core with tubes through it to guide the water out of the core area. I would have to think about it longer.

Yes, BUT, part of the reason for the walls to be there is for heat that doesnt get absorbed to transfer up and be absorbed where there is more surface area.

Jon

The honeycomb would sit right on top of the core, yes.

The problem is that the water flow is going to eat away at the core. Plus, you still have to seal the honeycomb's edges, which is easier said than done.

True, but this is simply theory :D

[sinister Kim Jong Il voice]
"Nobody can!"
[/sinister Kim Jong Il voice]