Water Cooling Computers With A Swimming Pool

JimH

Looks like this guy has just completed a course of study
in galvanics... that cool looking blue anodised finish
provided practically no protection from the rapacious
chlorine ions!

http://www.thebuehls.com/pool_cooling/

And if you want to read what the nerds are saying about
this installation, you can look here:

http://hardware.slashdot.org/article.pl?sid=07/01/06/1410219

Maybe some of you steam guys can figure out how to
use a pentium to make steam!

-JimH

ChasMan

Interesting. I cool my PC with a device which is effectively a giant radiator with a small pump in it. I have seen others with lengths of fin tube etc. I may go the fin tube route one day if I find myself with not much to do.

Christian Egli_2

Let me work on steaming up the supercomputer

There is a simpler way: it’s no mystery IBM named its super computer Deep Blue... toss the whole thing to the bottom of the swimming pool, and ta daa, efficient CPU cooling and free pool heat. No pumps, no pipes, no exchangers, plus, it’s ultra hygienic due to the chlorine.

Add a chaise lounge and you’ll find me by the “computer room”.

As far as steam heat goes, I kind of doubt a CPU wants to work at 212F, but the picture of the one computer piped to a 5 gal plastic pail shows exactly how steam is effectively cooling this set up.

The article says the water in the system steadies at 96F and the PC is happy. The article also says that much make up water has to be added - that’s cool, because the heat is being taken away by water vapor at more than 1000 BTU per pound. The bucket itself does not transmit much heat to the surroundings by sheer convection - nope, the efficient cooling going on here is powered by steam. (Steam just being our special nickname for gaseous water)

Some fun computations

Say you consume 4.25 gallons of makeup water and let’s roughly pick the value of 1000 BTU/lb to boil away water and let’s peg the gallon of water at 8 pounds, and lastly considering there are 3414 BTU per kWh. Calling Deep Blue, we get:

4.25 gal * 8 lb/gal * 1000 BTU/lb * 1kWh/(3414 BTU) = 10 kWh

This is the approximate energy dissipated by one disappearing bucket.

If it takes two days to do so, about 50 hours, then it only takes 200 W of CPU power to get to 10 kWh. It’s all very believable. Isn’t that cool?

Personally, I would go with oil as the fluid flowing around the computer and in its simplest form, I would very simply droop a length of hose into the pool for circulating and cooling the oil while letting the steaming action occur at the pool surface.

Would Deep Blue have figured all this out? or is it just good at chess.

Mike T., Swampeast MO

He [seems] to have considered the chlorine problems carefully and made use of a nearly perfect heat sink.

While the circulator is extremely oversized, at least he's not using fresh water and dumping it down the drain.

Perry_3

So, Seymore Cray cooled his computers directly with

Freon. The circuit board racks were evaporators - with reall good contact between the edge of the circuit boards and the racks (and special cooling blocks for the CPU's).

His third major design immersed the CPU and other boards directly into liquid Freon - and you could watch through a window at the lighted circuit boards and the freon boiling away the heat...

Most of the immrsion computers are still used today (and freon for applications like this is exempt from the freon use treaty).

Of course, All of the really large supercompters use at least one of the cooling techniques developed by Seymore Cray.

His genious was had several facets. You couldn't build a supercomputer unless you could cool it, and you needed to limit the wire length to increase computer speed (speeds of the processors were faster than the signals could travel down the lengths of wire in the older computers.

Seymore is the first to turn a computer into a large freon evaporator coil (achieving more than 10 times the heat removal rate of anything else that was done); and then built a "circular" computer where the longest wire connecting any circuit board to another was 3 ft long.

People think nothing of a computer that can do 1 gigaflops per second now. But in the late 1970's early 1980's it was mind boggling. It was not just a modest increase over existing computing power - it was "Super" advanced - a leap of magnitude gain in capability. You had to buy 4 of the largest fastest and most expensive IBM computers available to support a Cray 1 computer. 2 to feed it the input data; and 2 to handle the output data (IBM liked Cray... for obvious reasons).

I long ago lost track what speeds Cray achieved on a single processor machine (and they are still in business selling whatever the latest "super" computers are).

Of course, multi-processor machines have a larger gigabite speed. And they are best for certain types of calculations (where you are comparing the how the next point over affects this point - like heat transmission or weather). For many forms of problems though - only a single vector machine will do (nodal analysis won't work - and you cannot break the problem into parts that can be solved simultainiously).

I hope Seymore Cray is floating somewhere in a boat on a lake of liquid Freon. No one else thought of the idea or tried to build a computer cooled by it.

Perry

mtfallsmikey

Been there, done that

Saw IBM mainframes cooled with chilled water, Saw some of Cray's more recent creations. Still the Cadillac of the industry.

Constantin

He's an interesting guy...

... his next potential project is using the swimming pool as a heat sink for the air conditioning. I dissuaded him from using his regular condenser coil (!!!) and to focus on a tested solution like the doucette desuperheaters made of cupro-nickel (i.e. for marine use).

For those that care, he is using a bronze-bodied Grundfos pump. I wonder to what extent the chlorides + oxygen in the water will attack the solder joints, however.