SuperMicro CEO predicts liquid cooling will rack up 2,900 percent growth in two years

SuperMicro CEO Charles Liang expects liquid cooling will be installed in 30 percent of racks the company ships next year – vast growth, given the market for such kit has been moribund for decades. In his keynote speech at Taiwan's Computex event, Liang revealed SuperMicro has tooled up to build 1,000 racks a month equipped …

[DJO]

I canne break the laws of physics Jim

DLC's ability to cut energy costs by removing the need for some air conditioning in datacenters

So how and where are they proposing to get rid of the heat? It's the same amount of heat irrespective of the cooling technique.

Just dumping it out of the back of the rack will soon turn a computer room into a sauna and pumping outside is exactly what AC systems do and if you have to do it separately for each rack it is going to end up more expensive than a single room AC unit.

DLC is helpful in reducing the space and the noise but other than that there's little incentive.

[Caver_Dave]

Cool and quiet for a home PC

I've gone the water cooling route a few times, making tiny, silent, overclocked PCs. A couple of times I've had a single external radiator and one large, slow, quiet fan, another time I had a convection cooled radiator. If I had a rack full of kit to cool, I'd probably put the radiator outside, as you would with an A/C unit.

Remember that in a normal datacentre there are (often multiple) fans in each 'pizza box', plus the A/C unit. But with water cooling you can avoid all the noise and electricity consumption of the fans in each 'pizza box' and directly feed the warm water into the radiator or A/C unit replacement.

[DJO]

Re: Cool and quiet for a home PC

If you are cooling a whole rack then either you need a separate cooling system on each box or the coolant will need to be split into a separate pipe to each box - you can't daisy-chain them as the end of the line wouldn't get much cooling so either a stonking big pump or one smaller pump per box.

I can't see the pumps using significantly less power than fans, actually as the load on the motors is much higher moving fluids than for a fan moving air I suspect the pumps will consume more power than fans.

Of course in fluid cooled systems the waste heat is far more manageable being concentrated at a heat exchanger somewhere rather than blasted into the room and that is a considerable advantage.

For rack cooling if they use the big pump method they could have a redundant spare in parallel with automatic switchover if the first fails which would give a level of resilience impossible with fan based systems. (But in the real world, how often do fans fail in computers installed in air conditioned server rooms? And are pumps more, less or roughly equally reliable as fans?).