Slashdot Mirror
Supercomputer Cools Off Using Groundwater
gManZboy writes "The Department of Energy is no stranger to supercomputers, and its Pacific Northwest National Lab has proven that it can continue to be an innovator in the field by using what the lab calls a unique groundwater-fed cooling system in the lab's newest supercomputer, Olympus. The novel cooling system translates normal groundwater into big savings for the new 162 teraflop supercomputer, which is being used in energy, chemical, and fluid dynamics research. The setup translates into 70% less energy use than traditionally cooled systems."
It says it's a closed loop of groundwater?
That makes no sense at all. A closed loop won't get rid of heat, just transport it. There must be a system which exchanges the heat out of the water to the environment. Maybe a radiator system, maybe a chiller, maybe an evaporative cooling system.
Or maybe it's not really a closed loop?
http://lkml.org/lkml/2005/8/20/95
When I investigated a bunch of energy-saving options for a building I was managing, geothermal heat pumps were by far the most cost effective (not to be confused with geothermal energy). For our building, its payback time would have been 3-5 years. There are no fancy materials, no high-tech equipment involved. Just a bunch of buried/sunken plastic tubes with water flowing through them. The ancient Romans used a variant of it to air condition their homes.
Essentially they're the same thing as a window heat exchanger/air conditioner, except instead of using the ambient air as the heat dump, they use the ground or groundwater/pond. This provides a much steeper and more favorable temperature gradient in both winter and summer, allowing the heat exchanger to operate much more efficiently. Whereas air is about 90 F in summer, the ground is about 55 F making it much easier to pump heat into the ground. In winter the air is about 30 F, while the ground is still about 55 F, making it much easier to pump heat out of the ground. (Below about 40-50 F, most heat exchangers just shut off and run a heating coil, because it's so inefficient trying to extract heat from air that cold.)
They're an easy energy-saving measure which quickly pays for itself. I'm surprised more new building construction doesn't incorporate it. Makes sense for cooling computers, motors, etc. too if you've already got the infrastructure in place for your home or building.
The first "fluid dynamics research" project is to design its own cooling system.
The water is in a closed loop. They pump it down into the ground to cool the water, and then back up again to cool the computers. Then it just goes down into the ground again (and again).
no three-eyed fish here. From what I understand, the water in the system is pumped back directly into the water table, where any so far negligible temperature difference between it and the ambient water is swiftly negated by the sheer volume of water and rock it meets. Thereafter it becomes so diluted any contaminants would be quickly absorbed by the rocks several hundred, if not thousands of feet underground never to see the light of day again. Water effluent from power stations is generally sterile and has little in the way of dissolved minerals in it; it does have dissolved gases due to the nature of the cooling process, in an approximate mix of 78% nitrogen, 21% oxygen, 0.9% argon and 0.1% CO2 and trace gases.
Operation Guillotine is in effect.
It's a closed system, not an outlet per the article. Anyhow, the nearest river is a few blocks away - the Columbia. It's within the Hanford Site, but at the very last edge of it, adjacent to the town of Richland. A supercomputer's worth of heat sink there will be negligible in comparison to nuclear heat sinks just upstream. Also, depleting ground water reserves adjacent to a very large river seems unlikely.