Direct Die cooling...

[MeltMan]

I am sure someone has been crazy enough to attempt this. I was wondering if the actual die surface is electrically insulated from the rest of the electronics in the computer. This might be a dumb question since silicon is a semiconductor and wont conduct under normal circumstances- uless man creates circuits within the wafer. Basically, if someone were to touch the green die surface with a ground, would the chip blow up? (ignoring the fact that you couldnt, since it would burn up in 2 seconds anyway)

[iceheart]

Well the copper or aluminium water block you would have on it in normal cases doesn't break anything... Direct die cooling isn't particularly efficient anyhow btw.

[Triton]

Tom Leufken just released his component fluid injector. URL=http://www.leufkentechnologies.com/index.shtml]web page[/url] This is something similar to what you're describing. I haven't seen a single review on the waterblock yet. Would love to see it compared to the maze 2 and spiral.

[Triton]

Doh, to answer your question he recommends painting over the exposed traces with clear nail polish to prevent shorting them with water.

[BrianW]

Well I think he was talking about the core, which i am sure is non conducting.

[MeltMan]

Exactly ^ Thats what i was looking for. Core conducting.

Thanks guys.

[Mr Evil]

I know of several experiments of direct-die water cooling:

It seems that it actually keeps the core cooler than a conventional waterblock, which is contrary to what I would expect, but multiple experiments seem to confirm this.

The core does not conduct, so contact with water is not harmful.

Mineral deposits build up on the die after some time, even if purified water is used and the die is coated in nail varnish or whatever. This is eventually a terminal problem.

[Triton]

Deposits build up on the core even if water wetter is used? Just wondering. Thanks.

[MeltMan]

well the way i see it, if you directly cool the die, and have no copper in between, you eliminate several thermal barriers. First, you eliminate the copper. Even though its a great conductor, water is better. Second you drop the need for thermal goop. Arctic silver is great, but no where near the conducting abilities of water or copper. So basically, you let the water take the heat directly off of the cpu. seems like a good idea to me, if you dont mind the risks involved. I wonder about that water wetter. Possibly, noone has experimented enough to actually know... Just think, you could be the first.

[Brad]

not too many people do it....

imagine if you got a bubble going over the core, the cpu heats up astronomically in a second or so. the bubble moves on. the next lot of water comes on and gets boiled.

bye bye one athlon


imagine your pump fails. water heats up very quickly in the waterblock, eventually boils

bye bye one athlon

[Butcher]

Time to death before pump fail is pretty easy to estimate... If you have say a 10cc block (2/3 cu.in), then you probably need to heat about half that water (top half) up by 40C to have your cpu fail (bear in mind internal temps will be higher). CPU puts out 80W, 5 cc of water needs 21J/C to heat it, you'll gain 3.8C/s. That gives a time to fail of 10 seconds... OUCH.
If you have a copper block you have more water and the block will lose heat to the air as it heats, you'll have a much longer time before fail in which to catch it.

Even though its a great conductor, water is better.

Actually water is a lousy conductor, but as it's convection that's the issue that's beside the point. If you eliminate the copper you convect water directly off the cpu, this is better in terms of interfaces, but worse for surface area. It's better for heat differential though as the heat is cooling from a smaller area. That is till you get mineral deposites on the core, then you have a much worse interface than the copper.

Also the long term stability of the cpu in a wet thermal cycling environment is an unknown factor, you might find out the hard way when you cpu cracked and leaked water all over the motherboard.

[LiquidCool]

It's just to risky, especially since it uses the socket clips. You've got the block, tubing and water pressing in whatever direction, and now you have to make the clip tight enough to get a watertight seal?

People are already breaking the socket lugs off with heatsinks, imagine when they crank it that last little bit to make sure there are no leaks!

Not to mention when it does break, your going to empty your water out in about 2 seconds all over the inside of your case...

[Mr. Thompson]

One of the folks at Overclockers already had a problem with Leufken's system. The nail polish did not protect the exposed circuitry.