Slashdot Mirror
IBM Deploys Hot-Water Cooled Supercomputer
MrSeb writes "With the ISC (International Supercomputer Conference) kicking off this week, there's been a flurry of announcements around new supercomputer buildouts. One of the more interesting systems debuting this week is SuperMUC — IBM's new supercomputer at the Leibniz Supercomputing Center in Germany IBM is billing SuperMUC as the first 'hot-water cooled supercomputer,' an advance it claims cut power consumption by 40%. Dubbed Aquasar, the new system looks like any standard water cooler: water is pumped in one side of the blade, circulates throughout the system, and is pumped out. The difference, according to IBM, are the microchannels etched into the copper heatblock above the CPU cores. Rather than simply being dumped, SuperMUC's waste heat is designed to be converted into building heat during winter. Presumably it is mostly radiated away in summer, rather than being dumped into the offices of angry German scientists."
What again? Someone call Chet Heath!
Hot and cold isn't that hard. Maybe I am missing a point somewhere.
You take the heat energy biproduct from a processor and dump it somewhere else. In "normal" this case, an air conditioned room. Heat dissipated is being countered by air conditioning going 24/7. More energy.
Instead of watercooling, which can refrigerate a fluid (more energy in put and unusable for anything else), this removes waste heat and reuses it elsewhere.
This isn't going to work but... Instead of sitting in the tub and pissing in it continuously, your waste is being used somewhere else. (This is Germany after all)
Has anyone ever tested if we actually need air conditioning for a server room? I mean transform one into a "wind tunnel" where the waste heat is either ejected outside or used internally? Instead of a giant cube... what about a rectangle?
Will this lead to.. yo dawg I heard you like blades so we made your rack of blades into a blade?
In related news, the UN Security Council has scheduled the start of WW3 to coincide with the 5 year anniversary of SuperMUC's online-date, pending the lack of adequate air-conditioning in the angry German scientists' offices.
greed@All_Evils:~#
Those scientists wouldn't be angry to have heated offices in the summer. Germany can be downright chilly in the summer. I remember some beautiful July days in Berlin with highs in the 50s.
On top of that, heated offices will make the German scientists think they're in Mallorca or Costa del Sol and they'll be partying all day and night to the hot techno beats.
Slashdot began tracking this one two years ago.
"Rather than simply being dumped, SuperMUC's waste heat is designed to be converted into building heat during winter. Presumably it is mostly radiated away in summer"
They might be storing the heat rather than dumping it in the summer.
We are building a meat processing facility. Meat processing facilities use a lot of energy for heating water, cooling carcasses, freezing and general storage & air conditioning. To reduce our energy needs we're storing winter in thermal mass so that we can use it during the warm seasons. We're also using the 'waste heat' from our refrigeration compressors to heat water in addition to solar hot water and the backup of propane heating for the water. All of this will save us enormous amounts of money since we won't have to buy as much energy. Good for our carbon foot print and even better for our bottom line as more money will stay in our pockets rather than being dumped into the environment. IBM could do the same.
See http://sugarmtnfarm.com/butchershop
My university building is 80m from SuperMUC; there is a large campus at the site with several thousand students and employees. In winter it most definitely makes sense to use the heat from SuperMUC, as the average temperature is about 0 degrees Celsius. In summer it might be a bit more difficult to dissipate heat on hot days, though the average temperature is still only 19 decrees Celsius for July.
This is very simple water cooling. The principle is identical to what is found in high end overclocked systems.
Your coolant only needs to be cooler than the core itself to remove heat. It's been known for a long time that dumping the heat of an overclocked system into a room through a water loop will heat said room.
Even in the dead of winter when it's 0C outside, my *one* overclocked computer can keep my 300SQ ft room heated to above 70 degrees with no additional heat sources.
News? I guess. Definitely a stale idea though.
BFD. I've been heating my house with computers for a couple of decades.
Nothing worthwhile ever happens before noon
People need to understand and respect just how awesome water is as a coolant. The specific heat of the stuff (basically, how much heat you can 'sink' into a gram of it) and its benign, well-understood nature, and the fact that its density only changes a little bit between freezing and boiling points make it quite awesome.
I live in a city with a river through it. I really don't know why they aren't doing cooling via air-to-water heat pumps. It's really absurd to blow fans all day when the river could carry away 100X the heat without too many ill effects.
"Sometimes, I think Trent just needs a cup of hot chocolate and a blankie." -Tori Amos on Nine Inch Nails
It would be trivial to upgrade 45C water to e.g. 60C with a heat pump. This could be done with high efficiency, certainly COP > 3. Of course an office building might not need that much hot water in summer (maybe for showers for those who bike to work?), but other buildings nearby might. Or use it for district heating, if they have that in the area, but with existing systems that would probably require more like 70C.
Or just get the CPU's running at 300C, of course. Then you could run a steam turbine on the coolant...
Finally! A year of moderation! Ready for 2019?
Non-overclocked air cooled systems also heat the room. This article is about heating rooms that the computer isn't in.
Yeah, I think the last time we dealt with the products of angry German scientists, it didn't go well.
"Computers are useless. They can only give you answers."
-- Pablo Picasso
Water can be a b***h to use in a closed-loop cooling system. If it has any appreciable electrical conductivity, you get electrolytic corrosion of different metals in the cooling loop. If you use 18 megohm DI water, you get corrosion for other reasons. Depending on whether you have exposure to air (like in an evaporative cooling tower), you get bacterial and algal blooms, dirt, dust, pigeon poop - it's not as simple as "pump the water around in a circle and move the heat with it". Many closed loop water cooling systems run about 50% glycol plus other additives to mitigate the nastiness.
Decades ago Cray heated their building in Mendota Heights Minnesota entirely using waste heat from the supercomputers. When they built their new campus a few miles away and sold the old building they had to go through some amount of trouble to retrofit it with heating from conventional fuels.
What's old is new again.
Cyrano de Maniac
The very rough idea is obvious. Going through the specifics of instrumenting a facility, of determining what the acceptable temperature and flow rate are to keep cpu die temperature at an acceptable level (note, if your cpu is still kicking, that may not be enough, voltage leakage increases with temperature, meaning power draw goes up, and you are being inefficient by letting the die get *too* hot. Also, this is the fastest x86 based system in the world. In part because the cooling is adequate to let the cpu frequency bump up more consistently. Go too far in the other direction, and you are spending too much money on cooling for diminishing returns in die temperatures and the benefits lower die temperatures bring.
We are also talking about a plumbing setup designed to keep ~10,000 servers serviceable without getting the conductive water in the wrong places. This means some significant consideration in how the plumbing connects and how to make it quickly disconnect without leaking to replace parts. The risk for a single water cooled system is small, you paid maybe too much for it and a limited volume of water is in the closed loop. This scale is a different thing.
This isn't just some overblown overclocking setup hooked into an aquarium. This is a non-trivial amount of engineering that goes into it.
XML is like violence. If it doesn't solve the problem, use more.
Laundry shops need a lot of very hot water to "cook" the dirty linen that they receive everyday
Almost all laundry shops are using water heaters - whether they be electric powered water heaters or gas-burner powered water heaters - to heat up the water
Here's my proposal:
On the side of all super-computer center or any large scale data-center the authority should draw up a special "zone" for laundry shops
That way, all the hot water generated from the computers will be put to good use - without any additional wastage of precious energy resource
It's kinda fits into the "Go Green" concept that is so popular these days
Muchas Gracias, Señor Edward Snowden !
No, the coolant does not just need to be cooler than the core itself. It has to also be able to absorb heat faster than the core is generating it. This means that all the thermal resistances in the path must be accounted for. Usually this means that the coolant temperature must be MUCH lower than the core, which requires chilling the coolant. The difference with this system is that the incoming water temperature does not need to be particularly cool - it can be 45C/113F, which means it takes less cooling to get it down to that temperature.
... is no more
When Slashdot was brand new - trust me, I was there - there was very few racist post, and there was no "mcpc" spams either
Sigh !!
Muchas Gracias, Señor Edward Snowden !
We do that at the NCAR Mesa Lab too. It used to be the various Crays that were the heat source, but now it is the IBM Power 6 cluster. They had to install 3,000 gallons of chilled water storage inside the computing center to ensure enough chilled water was present to cool the cluster until it could be shut down if the chillers were lost (i.e. power failure and the backup generators didn't kick work). When the new NWSC compute facility in Wyoming opens, the Mesa Lab will have to do with ordinary natural gas heat.
It was a lot of angry German scientists that left Germany which gave the allies the edge in WWII and gave the USA it's lead in postwar science.
"Tracking": is that our new euphemism for dupes? "Slashdot tracked eight stories this week." Sounds classy. I like it!
what about cooking oil?
What about it?
Muchas Gracias, Señor Edward Snowden !
Yeah, I think the last time we dealt with the products of angry German scientists, it didn't go well.
Well, their products got us to the moon and back. Maybe if we turn up the water temperature, they will produce some again?
Does anyone know how hot the Chinese keep their scientists? It must be pretty high, considering China's recent achievements in space.
"Yo! Turn on the supercomputer, and hose down the scientists! We're all going into outer space!"
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
Yeah, Einstein and the Manhattan Project...
thegodmovie.com - watch it
The problem with simply "dumping the waste heat" somewhere is that you need to find a place where to dump it. As the story indicated, this is not so much a problem in winter, but in summer, when no one wants your heat (no offices b/c sweaty people won't be smart, no industry b/c the water is still too cool). The innovative aspect of SuperMUC is that they achieve free air cooling, even in the German summer when delta t is very very low. Another cooling fluid is not an option as refrigerators are not as power efficient. Its all about power efficiency. And TCO: LRZ saves 500k €/a by this hot water cooling compared to classic AC with "refrigerators".
Computer simulation made easy -- LibGeoDecomp
I think the GP meant running cooking oil through the system instead of hot water, and then use the hot oil to fry stuff that the scientiests want to eat. That would be useful both summer & winter.
On a different note, I wonder - why not have some of these supercomputers built and installed in some of the world's coldest places - say Fairbanks, AK, or within Russia, the 2 North Poles of Cold - Verkhoyansk or Oymyakon. Another idea for a good location would be Svalbard Islands in Norway, just off the Arctic - they have a satellite station that has a 10Gb/s fiber connection to the internet. Avoid casing the computers in question, just have an open roof over them, and the ambient temperature of the area should be good enough to perpetually cool them. Toss in a shitload of Itaniums - and gobs of memory, and have it run. Due to the constant cooling, the MTBF of the CPUs should be much higher, and they can run non-stop.
Of course, if they desire to channel the heat in such cold places, that too can be arranged - and done.
Wow.
I never thought that somebody that has never heard of Einstein would be able to find their way to this site.
The problem with the ice slush pond is two fold:
1) they're in contact with the soil which limits the lower temperature due to ground heat. The economics of insulating the bottom of a pond this difficult.
2) they're not going down to as low a temperature as we can with the saline batteries. Salt solutions freeze at a much lower temperature. For the pond they want to shoot the ice water out over the slopes when this is used as ski areas. For environmental considerations salt water ponds are frowned on in most places since salt is highly corrosive and kills plants and animals. Big issue so it must be handled with care.
Our closed loop in a box solves these. We actually have boxes in boxes so that each box is at the right temperature for its function.
These are all ideas that could be used by a lot of industries and for home use too. Little ways that greening not only helps the environment but saves money. I suspect there will be a blooming in the interest in these sorts of things as the cost of energy goes up. Free energy could change that though. :)