[Beowulf] immersion cooling for ASIC Bitcoin miners

Bogdan Costescu bcostescu at gmail.com
Wed Nov 27 03:22:11 PST 2013

On Tue, Nov 26, 2013 at 4:28 PM, Eugen Leitl <eugen at leitl.org> wrote:
> This is one of 3M Fluorinerts (not at all cheap), so there is phase change
> enthalpy as the liquid boils -- I presume there is a heat exchanger at the
> top where it condenses, and drips down again.

Indeed, this is how it should work at a macroscopic scale. I was
thinking more at the microscopic scale - when the bubble forms, the
heat transfer becomes suddenly much worse locally and only a fast
convection (and therefore moving of the bubble) avoids a hotspot. I
think that this was a problem previously with viscous (oil-like)
liquids, as the bubbles were not moving fast enough.

> The ASIC-bearing boards are vertical, so there should be considerable
> bubble-assisted convection.

If the boards are too high and placed too close to one another, there
is a risk of getting a too high ratio of bubbles to liquid, in which
case the overall heat transport doesn't work as planned. In the pic it
seems like the whole liquid is bubbling...

> The reason why I posted it is that these guys obviously expect to
> recover the expense of building and operating the data center by way
> of mined Bitcoins. Not sure it will turn out as planned.

Well, this report (as many others like it) only works to increase the
hype around Bitcoins - which leads to a higher value :)


More information about the Beowulf mailing list