wwenze wrote:But what I've added to speed up this process is the heat-exchanging block which incorporates a peltier. The peltier causes one pipe to be cooler than the other. The much greater temperature difference speeds up the convection process. And the CPU gets cooler water.
Intriguing concept indeed. There are few caveats, tho: the water flow must be so slow that peltier elements can pump heat faster than flowing water circulates it. Otherwise it will end up just heating the water. What I mean is that water can contain *a lot* of heat, so even if it flows like 1 liter per minute, it's still very hard for any peltier to pump the heat out before the heated water comes back to cold side and vice versa. OTOH, radiator becomes more effective as temperature raises, so is might be possible to reach an equilibrium at some point.
In addition to that you need to make the loop very unrestrictive.. something like Reserator tower might work. More like thin vertical two-channel reservoir than traditional cpu block and tubing and use some kind of valves to make sure water has only one way to flow.
Also the radiator and peltiers need to be seriously overpowered. If you plan to pump for example 300 watts worth heat continuously (to make peltiers pump faster than water circulates the heat back) you need to throw in 1500-3000 watts worth Qmax and undervolt the peltiers. Then you can pump some 3-10 times the heat than the TECs generate. This is not as insane as it sounds, as in this kind of configuration TECs would consume just some 30-100W of electricity and you can get 320W Qmax elements at $25-$30 apiece. I am in process of getting some 1500W worth Qmax for my own experiment, and I count myself being at least somewhat sane
.
While being intriguing, I have to say it needs a lot of money to test it out and then there's no certainty that it will work. But if you can afford to take the risk, it will definitely be educating.
What I would test first is a bit altered version of your system.. put the peltiers between CPU and the water loop and have water flow only through the hot side. This way the peltiers have to pump only the heat generated by the CPU, not something they have already pumped before. As radiator effectiveness increases linearly in relation to temperature gradient, this might even keep CPU at reasonable low temps.
You just have to prepare the hot side and the whole water loop to stand high temperatures.. The CPU can very well survive some 60-70 degrees of Celcius, but at that point water in the loop would be almost boiling. Definitely not for the faint of heart.. And almost calls for full metal loop, since I bet most plastics will not stand 80+ degrees continuous temps reliably..
But, if you want to try those out without spending hundreds of euros/dollars to it, you can do a simulation with like 1/5th of 1/10th of the final power. If you divide all variables with the same number, you should get pretty close simulation of what would happen in real system. All variables here means heat produced by heat source, radiator size, water block/heatsink dissipation area, tubing cross section area and total peltier Qmax.
So, buy a single powerful TEC module, like 170-320 watts Qmax and then scale all the other parts similarly.. like simulate CPU die with resistor that produces 8W of heat etc.
And post us images and results
.
The idea is interesting though, and if you can a practical test would definetly be interesting.