Charlie Stross
HOW TO MISLEAD WITH STATISTICS 1010:
26 million litres is 26,000 tonnes. Divide by 365 and you get 71 tonnes/day/MW, or 3 tonnes/hour.
1MW converted to heat with 31% efficiency will boil 3 tonnes in almost exactly an hour.
So our 1MW data centre is converting about 69% of its energy into computation.
The real efficiency problem is not the water consumption (although we need better heat sinks than potable fresh water!) it's the MIPS/MW ratio.
https://mastodon.social/@Kjaerulv/113690631295042803
Charlie Stross
@Kjaerulv FOOTNOTE: a typical UK household averages about 3KW of power draw (excludes those with PV panels). So 1MW is roughly 300 households.
UK per capita water use is 140 litres/day, assuming 2 person households that's 84 tonnes of water/day per MW of consumer use compared to the 71 tonnes/day for data centres.
Charlie Stross
FOOTNOTE: a typical UK household averages about 3KW of power draw (excludes those with PV panels). So 1MW is roughly 300 households.
UK per capita water use is 140 litres/day, assuming 2 person households that's 84 tonnes of water/day per MW of consumer use compared to the 71 tonnes/day for data centres.
Graham Sutherland / Polynomial
@cstross I also strongly disapprove of this whole "it uses water" fad in reporting. I need that goose meme but with "but where does the water go?" written on it.
like, yes, they use it as coolant, and there are environmental effects from using natural water sources as heatsinks, but the flow rate through the facility is not directly indicative of the problem. it's shoddy reporting that leaves holes for the real questions to be dodged.
Graham Sutherland / Polynomial
@cstross if we're going to evaluate the merits and deficiencies of this cooling approach, a better question is whether a comparable closed loop system, such as phase change cooling, results in reduced long-term environmental harm when all factors are considered. that's a much harder thing to quantify, but it's what actually matters at the end of the day if we're trying to build things that cause less harm.
Josh Simmons
@gsuberland @cstross yeah it's particularly annoying too because people seem to struggle with the idea that different places have different issues with water supply. you can even use the rejected heat for good, potentially, so while i'm extremely not excited in general about the whole thing, just directly taking the big numbers and write a headline is really harmful.
Charlie Stross
@dotstdy @gsuberland Yep. It's a totally different situation if your data centre is stuck in the Arizona desert as opposed to being colocated with a dense residentual district in a cold climate (eg. cities in Scotland, Scandinavia) where it is set up to provide a municipal hot water supply!
Ysegrim
@cstross "69% into computation" is physically impossible. I mean ... how could it? It's not like you go in with a hard drive, and when you take it out there is 10kWh of data on it. Energy just doesn't work that way. Except for the miniscule amounts of energy that leaves the data center as light (through windows or fiber), all electrical energy that goes in leaves the building as heat. Through coolant water, through coolant air, conductive/radiation through the walls and roof ... but as heat. In that sense, a data center has an energy efficiency of almost 0%.
(The exception being, the heat is used in some way. To heat buildings, to heat water for residential use, greenhouses, ... heat will then, again, leave these buildings again in the same way that it now leaves the data center building. But people can make use of the warmth in these buildings before that happens.)
Charlie Stross
@ysegrim It's not 69% into computation. A chunk of energy gets lost running the cooling system: more gets used up doing peripheral tasks (shunting bits between devices uses energy even if it's non-productive).
The question is what to do with the waste heat (and how to minimize stupidly wasteful (in human terms, not physics) uses of data centres, eg. cryptocurrency mining).
Charlie Stross
@zeri @Kjaerulv This is Scotland I'm using by way of an example: most of our renewable electricity is from wind power (we've got 20% of Europe's available wind just offshore: there's a 50-60mph gale blowing right now outside my window).
LightlySearedOnTheRealityGrill
@cstross @Kjaerulv 3kWh ... 3kW on average would be 72kWh and solar cells average 6h of runtime a day so 6 MWh for a 1MW plant ... 2000 households
but that electricity I think it's 10x energy consumption if you include heating
GNU social JP is a social network, courtesy of GNU social JP管理人. It runs on GNU social, version 2.0.2-dev, available under the GNU Affero General Public License.
All GNU social JP content and data are available under the Creative Commons Attribution 3.0 license.