Benchmark Heating
So if you're in one of those buildings and it's a bit cold, just log in and get it to run a few benchmarks and generate a bit more heat
IBM has delivered a water-cooled supercomputer to the Zurich ETH* university, saying it uses up to 40 per cent less energy than an air-cooled machine, and has a reduced carbon footprint because its waste heat is used to warm adjacent buildings. The 6Tflop experimental system is called Aquasar and has been designed and built by …
This post has been deleted by its author
The Computer Science department at Manchester was supposed to be heated on a similar basis, albeit using air heating. The idea was to circulate air heated by valve based machines around the building, eliminating the need for an actual heating system.
Of course, it didn't work so well when computers became entirely solid state. Otten it was actually colder inside than out.
used the heat from their water cooled IBM 360/65 and 370/168 to heat Claremont tower when I was at Durham oh so many years ago.
Interestingly enough, on the current big IBM Power6 575 water cooled IH nodes, the exit temperature of the water from the frames is less that 30C, with a temperature rise of around 10C input-to-return, which makes the heat difficult to use, even if there is a lot of it.
That would require a monstrous flow rate?? I can't even see at the innefficient cold end only a five degree drop across the standard building's radiator assuming a maximum room ambient or a very comfortable 20 c?? A very well insulated transport system between buildings might lose that much and I'd appreciate clarification/checking on this. It may always be some screwy volume figures involved as far as what the size of the building being heated is but after all the emphasis on CPU cooler efficiency pumping 60 degree water in seems a bit contrary.
I wrote a blog in the Reg on this story and checked out the numbers. I agree that the 5 degree delta seems too low, but that was the best info I could get at the time. I was expecting to see a higher output temperature given the amount of gear being cooled and the operating temps. So, as you surmise, we must be talking about a very high flow rate. I also believe that perhaps the numbers are a bit off in the materials that I was able to find (primarily a press release with a few supporting, but non-technical docs.)
for a friend who spent too much money on gaming machines and had SLI graphics cards worth more than his car. We used some Zalman Reserators in the sump filter of his marine tank to dump the heat from the PC into the fish tank which saved electricity on the tank heaters.
On the server side though this fiddly water block on the chip approach is not really scalable. It only takes part of the heat from the system, much of the heat is from the chipset and memory in these systems and the manufacturing & installation cost of the water blocks is prohibitive for volume technology. Check out proper immersion cooling of commodity servers such as Iceotope ( http://www.youtube.com/watch?v=K5e9_cqFiNE ) to see how water cooled servers can be economically installed in real data centres.
...makes enough heat to warm a room up quite satisfactorily (or less than satisfactorily, in Summer). It also generates enough heat to make the glycol evaporate and, thanks to a flaw in the design of the £1000 cooling system, blow a leak all over the place.
That friend of mine now has a single graphics card that's almost as powerful as the previous three. Hopefully next time he'll just get a refrigerated box for less.
I used to work to a nuclear power plant. It must cool the reactor, after 2 closed cycles, with an open cycle that uses sea water. Due environmental reasons, the power plant can only heat the 'ocean' by just a single degree. Disruption of natural life is a no-no, as usual.
The catch is, it must waste some 2650MW (yes, you read it right: two thousand, six hundred and fifty megawatts) of heat to this salty water. Yes, large-bore pumps exist for a reason, thank you.
OK, if you didn´t do the math, here it is: roughly, 76m3/s (cubic meters per second). That's more than a few rivers. Anonymous because memory fails, and regulations change over time.
Your assumption of huge flows are correct. Except that you are not trying to keep below 30ºC, you are trying not to reach past 65ºC. And you are not using that amount of power.
My sums are that 9kW with 5 degree C drop corresponds to a water flow of about 0.4 kg/s, or 2.5 m/s in 15 mm pipe. This seems about right by comparison with domestic central heating where 25 kW or so is transferred with twice the temperature drop and usually a bit higher average temperature via 22 mm pipework from the boiler at less than 3 m/s.
Some specs for radiators are here: http://www.stelrad.com/UK/docs/elite.pdf
Yes, there are some means of gain in harnessing that amount of waste heat.
Some German power plants keep the rivers where they are installed in navigation-friendly condition during winter, otherwise they would be frozen. Don´t ask me how environmentally sound is that.
Others use multiple facilities that take advantage on that heat for steam and such, aptly named "combined cycle" or "cogeneration". The trick is, rankine cycle prevents further use of all that steam for proper turbine generation, since moisture in the steam acts as an abrasive on the aforementioned (imagine a sandstorm at 800km/h and 250ºC, give or take), and the thermodynamic results are not good (read cost-effective). However, for proper heating, this excess heat is perfect.
Anyways, back on the point. That amount of waste heat has its uses. A hot bath after a hard day of server-farming feels nice.
Paris, 'cause, y'know, a hot bath...