[Beowulf] Innovative liquid cooling

Prentice Bisbal prentice.bisbal at rutgers.edu
Fri Mar 1 07:45:00 PST 2013

On 02/28/2013 09:12 PM, Mark Hahn wrote:
>> Actually, since the thermal conductivity of the liquid is so much better
>> than air, you probably don't need finned heat sinks. Heat transfer rates is
> well, you don't need AS MUCH surface area and/or can tolerate lower flow.
> even given better conductivity, and acknowledging greater heat capacity
> as well, can you really just cool one wall of a tank and expect everything
> immersed in it to be well-cooled?
> actually, does anyone have numbers on this 3M fluid?
> if the following table is to be believed, 1000x would be quite impressive:
> http://upload.wikimedia.org/wikipedia/commons/1/1e/Thermal_conductivity.svg
> (if I'm reading it right, it shows water about 30x higher W/(m*K) vs air.

Yeah... I was just re-reading my e-mails this morning, and was wondering 
where I got that 1000x number from. I think someone else might have 
thrown it out on here at one point. It could have been 100x, and I added 
a zero. Either way, I should have googled the thermal conductivity of 
Air, Water, and 3M Novec before posting yesterday. Qualitatively, my 
argument is still valid, the numbers are just waaaaay off. I think I 
have a future writing for infomercials.

For future reference, I just looked up these thermal conductivities:

Air : 0.024 W/m-k (at 298K)
Water:  0.58 W/m-k (at 300K)
Mineral oil: 0.162 W/m-k
3M Novec 7000:  0.075 W/m-k


I went with the 3M Novec because it looked like "7000" was written on 
the bottle in this video and the PDF lists "supercomputing" under 
applications. Looking at a comparison of all 3M Thermal management 
fluids, Novec 7000 has the highest thermal conductivity. (see 

Using this numbers,

Water is 24x better than air
Mineral oil is 6.75x better than air
3M Novec 7000 is only 3.125x better

What is clearly much better than the other alternatives for heat 
transfer. If only it wasn't conductive, and corrosive (rust, anyone?)

I hope this makes up for my earlier exaggerated numbers.


>> Boiling at these temperatures would require a relatively volatile liquid,
>> which would probably flammable (explosive), or bad for you in other ways
> pages around here:
> http://solutions.3m.com/wps/portal/3M/en_US/3MNovec/Home/Applications/HeatTransfer/Semiconductor/
> seem not to be too alarmed about boiling.  I'm guessing we might be talking
> about the novec 7300, which claims to have "low toxicity" as well as
> being nonflammable.
> I was thinking of boiling just because such a cycle moves a lot of heat
> and has some natural engineering advantages (bubbles, as you mention).
> even there, you'd obviously want to design your system to keep components
> out of the CPU's heat plume...
> I guess my main point is that you can't take a current air design,
> remove the fans and have an efficiently-operating fluid-cooled design.
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

More information about the Beowulf mailing list