Full Immersion Cooling

scrlk

scrlk

scrlk

Soldato
Joined
25 Sep 2009
Posts
3,630
Found this:
http://www.corvalent.com/martinscorner/archives.shtml#cpuboil


Traditionally, full immersion cooling has only been done with mineral oil (example) which poses problems with corrosion or very rarely due to its expense (plus the high environmental damage it causes), Fluorinert which was originally designed for the Cray 2.

This method uses another liquid called Novec 7000 (another 3M product) which boils at 34C and they use a TEC to change it back to liquid. At 100% load on a i3, the core temp was less than 55C which is quite amazing. Shame that it is one of those $250 p/gallon liquids though like Fluorinert.

http://www.xtremesystems.org/forums/showthread.php?266069-fluid-immersion-(boil-the-whole-computer)
 
Imagine running on air without a heatsink though ;). If they used a properly designed heatsink in this fluid the results would be way better.
 
Considering its only a proof of concept it is awesome. I agree with above too, if there was some sort of heat sink on it would achieve better temps but then it doesnt look as cool lol.
 
Would a heatsink really make much of a difference? More surface area yes, but then do you compromise the gains by having an inferior TIM for the heat transfer?
 
Sin_Chase said:
Would a heatsink really make much of a difference? More surface area yes, but then do you compromise the gains by having an inferior TIM for the heat transfer?
Click to expand...

I think if it makes a difference in air, it will in liquid, but it depends on how good your tim is...

hopefully it doesn't dissolve in the liquid!
 
Last edited:
I would have thought the risk with using TIM in this situation would be contamination of the liquid.
 
Good find, cheers. I was all ready to say "sure, it sounds a great plan, but then the blasted liquid destroys capacitors and pulls all the labels off". Much more interesting OP than I was hoping for :)

Sin_Chase said:
Would a heatsink really make much of a difference? More surface area yes, but then do you compromise the gains by having an inferior TIM for the heat transfer?
Click to expand...

Difficult to say reliably. The CPU - heatsink interface costs two to five degrees, then further loss in the temperature gradient built up in the heatsink. Boiling at 34, with a core temp of 55, suggests a roughly 15 degree improvement should be achievable with a heatsink, 10 degrees easily achievable. I think this would be fairly invariant with wattage however, as the liquid boiling removes the heat at a rate that should be grossly in excess of a processor wattage.

So, I believe a core temperature of around 45 should be sustainable with a heatsink, pretty much independent of cpu wattage. For an i3 this is poor, for a 250W i7 chip it would be rather good.

ScratchFive said:
I think if it makes a difference in air, it will in liquid, but it depends on how good your tim is...
Click to expand...

Different game. Processor hotter than boiling point means it's the change in phase from liquid to gas which is removing heat. Think of a kettle, it gets hotter until the water starts boiling, but it'll then stay at the same temperature until the water runs out. The steam is taking the best part of 3kW away with it.

edit: As an afterthought, if these liquids don't attack components with too much enthusiasm, then an airtight container with a condensor mounted on top would be entirely silent and cool the whole system. Anywhere that gets hot, the liquid boils, and moves into the condensor (I'm taking a guess that the gas is less dense than air, if not one can always fill the second half of the box with a denser fluid). Here it turns back into liquid, and runs back down the tubes. That's very elegant indeed.
 
Last edited:
Only concern I would have with a heatsink would be if the tim could come of due to the boiling liquid.

Very cool idea though
 
You must log in or register to reply here.
Back
Top Bottom