Using H2O as thermal compound.

[CheeseBall]

I have been thinking about this for some time. I haven't been meditating on it as much as I should be here is what I'm thinking:

We know direct die watercooling doesn't work long term. But what if we did the following:

1) Cut a 1/32" deep pocket into the bottom center of your waterblock. The pocket being slightly smaller than the size of the die (and centered over it).

2) Then cut a small channel around your new pocket for an O-ring. For an AMD XP, you would have the O-ring a milimeter or two outside the die. So water would be able to touch the CPU's PCB, but only right next to the die. I need to find out if there is any thing electrical there. But w/ A64s and pentiums this would be alot easier.

3) Now the trouble is getting water into our pocket. The unpocketed area around the pocket will be in contact w/ the outside edge of the die after we mount the block. So to get our water in there, we drill a 1/32" hole at an angle from the highest spot on the side of our waterblock base, to the pocket. We want to start the hole on the side of the block that will be the highest side when the block, mobo, and case are in their normal position. Would would like to enter the pocket at the edge or corner, so the air in the pocket floats UP, out of our hole when we put water in it.

4) Now after we mount the block w/ the O-ring. Position the case so that our hole is on top. Then take a syringe (or something you know will work), and fill it w/ water. Then stick the syringe into your hole, and inject water slowly. Keep the perimeter of the hole surrounded w/ a towl, so when/if you fill it up to much, it doesn't make tooo big of a mess. Once the hole is filled completely, tilt the case around a little to allow any air bubbles to get out that haven't. Then make sure our hole is most of the way full still. Then plug the hole w/ whatever will work well. Finally, wipe up any spilled water and let the sytem dry for a day (if you made that big of a mess), and put some paper towels under the waterblock so you can find any obvious leaks.

5) Boot up the system and check temps right away.


So what do you guys think about this? The hardest thing will be drilling that damn small hole, and angling the hole as to not go through the O-ring channel. Any other ideas?

Since the water is not moving, it should remained sealed for as long as we want w/ out leaking. Yes? And the solid bay of water between the CPU and waterblock should beat the heck out of AS5. Yes?

And As you have already noticed, if we don't get enough water into our pocket, we are screwed. And if it leaks, we are pretty much screwd. But when did that ever stop anyone? Do you think a refined version of this would be better than both AS5 and that solder method.

Can someone post a link to the soldering method of mounting a waterblock or heatsink?

[jaydee]

Seems to me this is a very bad idea. What happens when that water boils? It will create steam and a LOT of presure blowing the CPU core up and the water block right off the mobo.

Or maybe I am not understanding this correctly...

[CheeseBall]

Hmm, I didn't think the die of the CPU would ever get warm enough to boil. The heat would go right from the die, to the water, to the copper, and then finally to the water in your cooling loop. I wouldn't think the die, or the water between the die and the block would ever get above 75C.

[jaydee]

Originally Posted by CheeseBall

Hmm, I didn't think the die of the CPU would ever get warm enough to boil. The heat would go right from the die, to the water, to the copper, and then finally to the water in your cooling loop. I wouldn't think the die, or the water between the die and the block would ever get above 75C.

Well true enough. I thought boiling point was lower than that but it isn't.

[BalefireX]

Classic - fill your head with so much esoteric watercooling knowledge that you forget water's boiling point!

[jaydee]

Originally Posted by BalefireX

Classic - fill your head with so much esoteric watercooling knowledge that you forget water's boiling point!

LOL, thats a quote for the sig!

[CheeseBall]

Any thoughts about practicallity and performance? I know it won't be extremely practical, but better then trying to do the soldering method.

Should I just give it a try?

Also, how well does ice conduct heat as opposed to liquid H2O? I have been thinking about how to apply it to below 0C phase-change cooling. You want the ice to be frozen solidly to the die and block, but the ice will expand... so if it''s frozen from liquid in a closed chamber, it will probably break the die.

[topdragon147]

correction, water expands when it freezes...so the ice would contract when it melts

[CheeseBall]

Umm... yes, that is what I said. I am talking about -30C phase change cooling here. So once the water is frozen, it's going to stay frozen unless your compressor stops or something. The thing is getting a good joint between the two. If I did the same method from above, once we powered on the phasechange system, the liquid water would freeze and expand w/ no place to go. So we need a different method of holding the water there. I dunno if you put a couple drops of water on the CPU die, then mounted the block, then turned on the phase change system if that would work. I thinking that method would not be reliable enough. If it's at all possible, we would want the ice totaly frozen to the die and block as good as possible. Like when there is a puddle on some concrete, you throw a peice of plywood on it, and then the puddle freezes and you have to pry on the plywood like crazy to get it off. Again, ice might conduct heat like crap, I dunno.

[HAL-9000]

Originally Posted by topdragon147

correction, water expands when it freezes...so the ice would contract when it melts

I believe the only other material that contracts when heated and expands when cooled is plutonium. I have always found that strange.

The delta phase of isotope 240 is the worst, must be very careful machining a piece close to alpha because if it heats up and contracts, density goes up, and alpha = >1, which is....bad. Just use to work in that business and I've always thought that property being unique to Pu and H20 was weird. Pardon my off topic rant all.

[jaydee]

This ice theory is probably the better way to attempt this. Cut out the block like you mentioned in your first post, don't install it, fill the cavity with water, throw it in the freezer, take it out when it is frozen and install it on the CPU. No worries about leaking or filling.

[topdragon147]

i think that ice would have the same thermal properties as water, some one correct me if i am wrong

[CheeseBall]

This ice theory is probably the better way to attempt this. Cut out the block like you mentioned in your first post, don't install it, fill the cavity with water, throw it in the freezer, take it out when it is frozen and install it on the CPU. No worries about leaking or filling.

When doing this method w/ a phase change system, the problem is that the ice will not freeze flat. It will most likely be convexed. MAYBE it will be flat/smooth enough, but if it isn't, and it's not making good contact, and the CPU get's over freezing, then you have some melting/leaking going on.

As for using it in non phase change cooling, and still using the O-ring... I think that is a good idea . Just have to account for how much volume you are going to loose when the water contracts, then decide how you want to make sure there is still good contact between the water and CPU/die after the loss. Which wouldn't be too hard. I will try it out... sometime

[topdragon147]

Originally Posted by HAL-9000

I believe the only other material that contracts when heated and expands when cooled is plutonium. I have always found that strange.

hmm, never knew that

[CheeseBall]

Looks like the Thermal conductivity (W/mK) gets better as it gets colder, while the Specific heat (kJ/kgK) gets worse. That is w/ below 0C temps.

I noticed that the low point for water's specific heat CAPACITY is about 35C. Then as temps increase or decrease, the specific heat CAPACITY increases. I think I said all that right.

http://www.engineeringtoolbox.com/24_576.html

[topdragon147]

i think you just contradicted yourself, or i read it wrong

[Butcher]

What exactly is the advantage of water? It has a thermal conductivity significantly below that of AS (and copper).

For reference, some thermal conductivies:
Water 0.56 W/m-K
AS5 8.89W/m-K
Copper 390 W/m-K

The specific heat capacity of water is not very interesting if it's not being pumped so, if I understand your proposal correctly, water is about as bad a thing as I can think of (well except say, air).

[CheeseBall]

Crikey, well that settles that I guess. I didn't think AS5 was that high nor water that low. Thanks for the info. THe only advantage now would be that the "joint" between the die and block would be almost 100% solid, no air. But yeah... AS5 it is then.

The thermal conductivity of ice at -35C is 2.5W/m-k... but that is still too low. And I guess the thermal conductivity of mercury is only 8.34W/m-k. But I won't even go there.

[bigben2k]

I'm just glad to see an end to this nonsense, before someone lost a CPU!

[jaydee]

Originally Posted by bigben2k

I'm just glad to see an end to this nonsense, before someone lost a CPU!

No spirit of adventure. What happened to water cooling the back side of the CPU?

[bigben2k]

You wait and see, buddy!

[Butcher]

There shouldn't really be any air in the interface anyway, if properly assembled. the main reason people moan is because AS5, even at 8.89 W/m-K is still over 40 times worse than copper.

[Groth]

There will always be air in the TIM - tiny bubbles trapped in microscopic ridges and grooves of solids, and little bubbles introduded into the goop by the very acts of dispensing and spreading.

Unless, of course, you mount your blocks in a vacuum chamber.... I half suspect that's what Intel does when putting the lids on its CPUs, explaining their better than average TIMs.

[Butcher]

But using water isn't going to improve that anyway surely?

[Groth]

Nope, water won't help with the air bubbles and has a sucky thermal conductivity.