[Beowulf] immersion cooling for ASIC Bitcoin miners

Prentice Bisbal prentice.bisbal at rutgers.edu
Wed Nov 27 10:12:15 PST 2013


On 11/26/2013 07:22 AM, Bogdan Costescu wrote:
> On Tue, Nov 26, 2013 at 12:05 PM, Eugen Leitl <eugen at leitl.org> wrote:
>
>> In fact, this cooling technique is of the planet most advanced.
> Thanks for limiting to a snippet ! I found the article a little too
> light on tech details... inexperienced reporter :)
>
> The heat exchangers on the building top are in a VVVV shape. The
> middle panels seem to me to blow hot air against each other - is this
> efficient, given that they seem to have enough space to put them a bit
> farther apart ?
>
> One of the pics shows a bubbling liquid surface, we've seen recently
> similar pics. From what I remember from past discussions on this very
> list, the bubbles are bad for heat transfer - it's OK to have them at
> or very near to the surface, but not down below. Has the physics
> changed behind my back ?
Actually bubbles are very good for heat transfer - changes of state are 
great for storing/releasing energy. This is how refrigerators and steam 
radiators work. It works like this:

As the computers give of heat, it's absorbed by the cooling liquid. The 
liquid increases in temperature until the boiling point is reached. At 
that point, the liquid stays at that temperature, but absorbs A LOT of 
heat, until it has enough energy to become a gas. This difference in 
energy is known as the latent heat of vaporization, and this can be a 
very large amount of energy compared to the energy to raise the 
temperature of that liquid a few degrees. The gas bubbles then rise, 
taking all that energy with them. Eventually, the gas will cool and 
release it's energy. Once it cools to it's boiling point, if it releases 
energy equivalent to its latent heat of vaporization, it will condense 
back to a liquid.

For example, let's look at water, which is probably the best known 
coolant and the most studied. The specific heat of water is 4.817 
KJ/Kg-C. That means it will take 4.817 KJ of energy to raise 1 Kg (1 L) 
of water 1 deg C. By comparison, the latent heat of evaporation is 2,270 
KJ/Kg, so that same Kg of water will absorb 2,270 KJ of energy as it 
turns to steam. That's an increase in energy capacity of 471 times vs. 
raising the temp by 1 deg C. (source for numbers: 
http://www.engineeringtoolbox.com/water-thermal-properties-d_162.html)

The same principle applies when a solid melts and vice versa. In the old 
days before refrigerators, this is why blocks of ice were sold for 
cooling purposes. As a block of ice melts, it can absorb a lot more heat 
than the same quantity of water at the freezing point.

Now gases don't have good thermal conductivities or capacities, so once 
the gas is formed, you want to get it away from the heat source and get 
it to a heat sink where it can give off it's energy and condense back to 
a liquid. This is probably what you're thinking of. Bubbles that form 
but don't go anywhere will definitely hinder conductive heat transfer, 
but when they boil off and rise, it's a tremendous help to heat transfer.

I hope this clarifies things.

--
Prentice






More information about the Beowulf mailing list