This is very interesting. The oil submersion of server heat generating components allows efficient heat transfer at 40C vs air which has to be much lower temperature to efficiently transfer heat. Keeping coolant down to 40C is much easier, requires much less energy, than say 24C for air.
I think that's a question of power density. The problem is often a limited power budget for the site. Filling as much of that budget with processing, storage and communications and not cooling is becoming paramount.
Remotely Operated Vehicles (ROVs), and other similar systems that operate at great depths in the ocean, have been using this technology for decades. In our case, the mineral oil is also used to equalize the pressure inside the electronics housing with the ambient seawater pressure so that very strong and expensive enclosures are not required. The oil lasts forever and is not aggressive, it does not exhibit chemical compatibility problems with most electronic components. As with all technologies, there are nuances to master but nothing exotic. Mineral oil is also environmentally friendly in case of a spill.
@agk: none of the above! FYI, the JEDEC 50.1 procedure to calibrate semi chip packages for thermal resistance does call for immersing in an Oil bath. Obviously the board assemblies are cleaned prior to being mounted on the server chassis, so there shouldn't anything leaching but there could be some outgassing.
So what Intel is doing is nothing out of the ordinary. It is a staright forward process to calculate the required oil volume for a given temperature rise.
@iniewski: we are almost there! There are microchannel coolers in CPU's that use propylene glycol and water but so far no liquid nitrogen... HP Z420 and Z820 Workstations already use liquid cooling.
thank you @docdivakar! you are very well informed, as always...would you be interested in giving a talk at emerging technologies symposium in Whistler next summer? details at www.cmoset.com, email@example.com