Use Peltier to cool water... not CPU (theory)

[BadPenguin]

What if you made a resivor, cut a small square on the bottom, sealed the cold side of the Peltier so that water would run over it, and attack a (shoot me) heatsink/fan combo on the back to keep it cool. would that work? I mean.. it seems like a good idea. any thoughs? Or if you cant actually get Peltiers wet, make the bottom of the resivor coppor, and attack the pelter to it with some AS5 and just chill the whole bottom. Would it work?

[iroc409]

to make it somewhat efficient, you'd need some sort of heatsink on the inside of the res to better transfer heat (just as a heat sink transfers heat away from the pelt).

there are units that run on a/c you can get made for fish tanks, already assembled for around $100. those claim they work fairly well, wether they do or not in a pc setup, who knows.

but, from what i've heard (you might want to check out the water cooling part of the forum ), this method isn't particularly efficient. much better to run on the cpu itself, although there are the obvious drawbacks of such.

but hey, try it out. pelts aren't too expensive, if you've got the power to drive them.

[pdf27]

It will work to cool the water, but will be less efficient than cooling the chip directly. The problem is that watercooling the hot side of the peltier will get it much colder than aircooling it. Hence, the cold side will also be much colder. This means you can get a lot colder at the CPU if you use CPU -> Peltier -> Water than CPU -> Water -> Peltier.
However, such an arrangement will allow you to get sub-ambient water temps if you attach the Peltier after the radiator in the loop, and it is powerful enough. The other advantage is that if the Peltier dies for whatever reason, the computer doesn't.

[feathers]

"It will work to cool the water, but will be less efficient than cooling the chip directly." - Yup, A peltier's surface area is very small and so the cooling is tightly focused. Hardly ideal for cooling a large volume of water flowing at high-speed!

The most effective place for the peltier is on the CPU itself as PDF27 stated, chilling the water would serve to increase the performance of the CPU peltier though!

I really don't like the idea of room-temp. water flowing through my system. A watercooling system with a peltier sucks bigtime! If you combine water+thermoelectric then you get pretty good performance. It would be great to go that step further and chill the water.

The first chiller I made was based on the Aquarius 2 system with it's pathetic 6 or 7mm ID tubing. I used the Aquarius 2 waterblock as the chiller itself. It was connected to an 80W peltier and the waterblock was submerged in a thermally insulated watertight box.

The water would flow into the box through tubes and directly to the waterblock inlet/outlet nozzles (rather than using the box as a reservoir). However, a crucial step in improving the chill performance was made by filling the small thermal box with water. The peltier/waterblock would then freeze the water solid (overnight) and the solid block of ice would further chill the water as it flowed through the tubing (looped at least twice) before reaching the waterblock. The results were very good and the water would get so cold that anti-freeze would have to be used!

Unfortunately this design was never tested on a computer and I scrapped it to build a chiller based on 12mm tubing and Eheim 1048 pump.

For the Mk2 chiller I made a simple custom waterblock which consisted of an aluminium box with a vaned heatsink fixed inside. Water would flow into the box through nozzles and pass through the vanes of the heatsink.

This mk2 unit used a 130W peltier.

It was a disaster.. The flow-rate with 12mm tubing and Eheim pump is so great that the water doesn't have sufficient time to chill.

The key to successful water-chiller design is to force the water through as long a route as possible (like a radiator) and to ensure that the water is forced through as flat a channel as possible so the water is closer to the surface of the peltier.

Conventional radiator design is far from ideal because although the water flows through flat tubes.. Those tubes are generally quite smooth with nothing along their path to create turbulence.

If you want to get your water-chiller project off to a good start then take a look at the Swiftech MCX4000 heatsink. It consists of a large copper CPU core-plate (where the cold-side of the peltier will sit) and hundreds of densely spaced pins. If you rotate the heatsink to 45 degrees (so it becomes a diamond shape) and pump water through the top point of the diamond and expel through the bottom.. You have a water-chiller that's close to perfection! The water has to follow a very torturous route from the top of the heatsink to the bottom (why rotate the heatsink 45 degrees? - Because at right-angles the pins give a clear path from one side of the heatsink pin matrix to the other!).

Even so, you would need several of these heatsinks in order to get effective chilling!

A cheaper alternative would be to take two cheap vaned heatsinks (with long vanes spaced quite far apart) and interlock the two (have the vanes of one extend between the vanes of the other). Seal the sides and drill a inlet/outlet holes on the two outermost vanes of the heatsink. The water is then forced through a zig-zag route through the vanes and this would give very effective cooling since the heatsinks would be chilled.

The other trick is to use metal which you wouldn't normally use.. Metal that is slow to heat-up becomes preferable in this application. I know of at least one Japanese company that makes these peltier water-chillers. Their custom built waterblocks enhance the cooling performance of the peltier.

I can give you a link to their website if you wish? and also give you pictures of the prototype units I made.

:-)


I have built two prototype water-chillers using peltiers.

[feathers]

"I have built two prototype water-chillers using peltiers." - That line at the end shouldn't have been there!

:-)