Extreme Extreme Cooling Setup

I had an idea...what if you had TWO watercooling loops. On the first loop, there would be *>pump>res>radiator>waterblock>* and the waterblock would be connected to another waterblock with a pelt in the middle, like a 226 watt, then another waterblock on the other side (cool side) going through another pump, res, but to another waterblock with something like a 175watt peltier cooling the CPU directly. So...it would look like this...

*>pump>res>radiator>waterblock*>pelt>+waterblock>pump>res>other waterblock>+

* = loop point for first cooling loop
+ = loop point for second cooling loop

If you cool a cool peltier, doesnt it become even cooler? It just makes sence.

There is practically no use to have a secondary loop for a TEC setup. All you are doing is adding pump heat to the cold-side of TEC and that will very negatively affect overall cooling capacity. You are allo adding several more heat transfer points and everywhere you are effectivly adding losses.

However, using a small pump on TEC cold side and go for direct on cpu-die cooling is something I have pondering on.

You have a small gain by going direct on die cooling, since there isn't any TIM between WB and CPU. That has to be balanced against extra pumpheat induced on coldside. Every 1W of extra heat on coldside will render like 2 W heat on hot-side due to the inefficiency of a TEC.
So far I don't know since I haven't calculated anything about that scenario.

Thats why theres a 226 watter on the first loop and only a 175 on the second. Compensation for pump heat.

Oh my.... that's even worse I'm afraid...

First: you don't get 226W cooling, since you prolly don't feed TEC 15V/30A.
Second: Cooling at hand is around 180-190W at 12V.
Third: Add-> pumpheat 10-20W + cpuheat 80W? + 150W TEC heat.... = ??? W !!!

I'm afraid that what we get from that is one burnt 226W TEC.
The rule of thumb I use when TEC stacking, use three times more for every stage. To Cool one 172W TEC use three 172W TEC. Total no. of TECs = 4
TEC stacking is extremely expensive and inefficient.
Running TECs in parallell is another matter.

Cooling the water(/metanol or whatever you will have to use) with a pelt has one more advantage that "tickles me in a way that if Loretta tickled me that way I would say "Yeah, that's nice.. that's the spot""
And that advantage is that you could use the cool liquid to cool your RAM using WBs.

How about that? ..sooner or later I'll have it on my RAM!

alright then...TWO parallel 226watters...on the first loop. I guess theres only one way to find out....do it

or you could do what i ma doing and use the TEC to cool the res and put a hsf on the TEC and keep the res well insulated. with the setup up i am putting together i expect to see coolant temps of around 5-10C. i got real lucking and found a small frig that's for like a 6-pack of cans. i've left it running and measured the temp inside the cooler and it's about 0C. with little modding effort and a custom res to accomadate a submerged pump it should be fairly effective. and then you take into account that i only paid like $30 for the frig it's pretty cheap.

Yup KLowD9x, that's the way...

I'm going for two 226W on coldplate on cpu, for starters...
Then I'm trying a two 226W waterchiller and circulate icy water direct on cpu die. I have no idea which method is the best yet.

Seems like a good method in my brain to skip that tiny tiny thermal interface between a ice-cold copper bar and 90W+ hot cpu. There has to be significant losses in that particulat tim, and having cold water rushing by seems to me be a very effective way to remove unwanted heat.

Ofcoz... should the second system pump ever stop I'd have a dead cpu in 3 seconds...

by the way, I'm doing this to a P4.

quit while you are ahea................behind and losing there is no reason to have a second loop on the TEC, you will just add more heat!