LN2-level CPU/GPU cooling

[DrMrLordX]

We've all seen suicide benchmark runs done under LN2 with impressive results. This is usually how world records are set. LN2 pots on top of the CPU are obviously not practical for everyday overclocking since the LN2 boils off so quickly. Unless you can afford to hire a butler to pour more in on a regular basis, it just won't work (and that would get expensive anyway).

What other ways are there to achieve such low temperatures (-190C)?

I was looking at TEC stacking, but that is entirely impractical. In fact, any solution involving TECs will probably fail since the ones I've seen operate at temperatures between -60C and ~110-130C.

The lowest temps I've heard of from triple-stage cascade phase-change cooling are -120C (or lower?), but my knowledge here is considerably lacking. Also, I'm not sure if a phase-change solution would be capable of dealing with the massive heat loads that LN2 can successfully dissipate.

Any thoughts? Obviously attempting to chill any PC component to those temps is a little crazy, and I have no serious intention of doing so . . . just idle speculation.

 

[jmmtn4aj]

Well, you don't need a butler, just a pipe connected to a tank that refills as fast as it boils off, so the pot always has the same amount. The tank can be large, so you'll only D) have to refill it with LN2 like once every few days..

^^

 

[Billb2]

There are commercial 24/7 phase units available, though -180*c is out of the question.
For Example:
http://www.frozencpu.com/cat/l1/g41/Phase_Change.html

A custom built two stager would be the most practical way to obtain the lowest temps possible on a 24/7basis. But $$$$ for initial cost and daily operation.

Anyway, now WR are all set with LN2. And every 3-6 months or so the newest CPU/MOBO/memory combinations on air will beat past LN2 records. It just costs a lot of money to stay at the bleading edge whether on air, water, TEC, phase, dice or LN2.

 

[aigomorla]

Originally posted by: Billb2
There are commercial 24/7 phase units available, though -180*c is out of the question.
For Example:
http://www.frozencpu.com/cat/l1/g41/Phase_Change.html

A custom built two stager would be the most practical way to obtain the lowest temps possible on a 24/7basis. But $$$$ for initial cost and daily operation.

Anyway, now WR are all set with LN2. And every 3-6 months or so the newest CPU/MOBO/memory combinations on air will beat past LN2 records. It just costs a lot of money to stay at the bleading edge whether on air, water, TEC, phase, dice or LN2.

yoxxy had a vapo and he burnt it out on an overclocked quad.

Your going to need a full custom phase unit built by a legend like nol (n00bofl33t) or Jinu over at XS. <thats if you live in the US>

Triple cascades arent practical for everyday use. The size and volume of a triple is big. I believe fuggar was wondering if it would even fit in a trunk off a mustang when he was taking it down to that intel presentation along with a QX9650.

 

[DrMrLordX]

Originally posted by: jmmtn4aj
Well, you don't need a butler, just a pipe connected to a tank that refills as fast as it boils off, so the pot always has the same amount. The tank can be large, so you'll only D) have to refill it with LN2 like once every few days..

^^

Well, see, I was thinking of that myself, sort of. Actually what I was thinking of was a LN2 condenser that would refill the pot dynamically. Granted, a LN2 condenser is no laughing matter. I don't believe small-scale refrigeration plants for LN2 production are manufactured or sold by anyone, and even large commercial plants only operate at 50% efficiency. This is the smallest unit I've found available here:

http://www.rigaku.com/cryo/nitrogen.html

Minimum output is 10 litres per day, maximum is 60. According to my rough calculations/estimates, you might need 270 litres per day . . . yuck.

Still, I think it could work. You'd only have to fill the system with a set amount of LN2 and then start it up, and the condenser would ultimately be responsible for removing all the heat introduced into LN2 by the CPU plus whatever heat seeped into the system (which is a problem in LN2 storage . . . I noticed that dewar flasks designed for storage of LN2 are rated by the liters of LN2 they lose per day as heat seeps in, causing LN2 to gassify and escape through valves to prevent the flasks from exploding). The above unit is designed to pull N2 from the air (unnecessary) so a device for CPU cooling could be simpler.

Assuming 50% efficiency which you would probably not get from a small-scale unit, you should be able to condense all gaseous N2 coming out of the pot for refilling purposes with a 800W-1KW condenser easily (assuming you're only cooling the CPU that way, and assuming a hot quad putting out 300-400W of heat). Of course, the entire device would have to be designed carefully to prevent explosions or N2 leaks which can be very dangerous if the gas accumulates in an enclosed area with poor ventilation.