Another idea for peltier cooling

[David|SP]

Hi people

I´ve been on these forums for like 3-4 months but haven´t posted yet, so this is my first time

So here goes an idea I had some time ago:

I currently have a basic watercooling system: 1048 pump, innovacool rev.3 block and innovatek 120x120 radiator, and i want to jump to peltier cooling

I thought putting the peltier directly on the CPU would be pretty risky, as i´m going to have the computer on 24/7. So i thought cooling the water would be a better idea. I´d need 2 water loops for that (CPU + peltier cooling) and that´s expensive

Enough talking: i´ve attached an mspaint file so you see my idea. Could you tell me if it´s possible?

Thanks

[g.l.amour]

dangerous!!!

u risk letting all the water bypass the rad...

[Since87]

It won't work because TEC's are too inefficient.

You would be putting a large amount of additional heat into the water because of the power dissipation in the TEC. This will raise your average water temperature by an amount larger than the "cold block" is able to cool the water as it passes through.

[koslov]

Well, I have been considering this for a long time now, except using heatsinks to cool the pelts. Get a slew of 80W pelts, hook em up in a little array, and slap a massive Al HSF on there. I was thinking of mounting this to my Lian Li PC-60, just right on top. You would need insulated tubing and all that, but you could even set up a PWM to control the pelt power consumption. Your CPU would be in no danger if they failed (well, it might get unstable if you have an extreme OC), but it wouldn't fry. Make a large WB to clamp to the pelts and heatsink, and you would have some chilly water with 3 or more 80W pelts.

The way I see it is this:

Advantages:

• Controlled power consumption using PWM
• No danger of CPU meltdown when pelts fail
• Can use as many pelts as desired, on-CPU limits to 1 (or rarely 2)
• Can dissapate heat through large heatsink (or secondary WC system)
• Much easier to build than a phase-change or similar system
• Most likely quieter than phase-change depending on method of pelt cooling
• No need for radiator

Disadvantages:

• More inefficient than phase-change
• Need insulated tubing
• ???

I would have already started work on this, but I have been unable to find an affordable power supply to fufill my 40 amp needs ! I am considering building my own, but even that path is costly.

If anyone has done something similar, I would like to see it. Otherwise, I would like to be the first to do it!

[murray13]

Hey, for once am I the first to try something out?


I have done something like what koslov has suggested.

I'm using a surplus Thermotek solid state (TEC) chiller. Modified to run constantly. Now I dont have the best WB (designing one) or the greatest setup. But I have a water temp of 12.1C and a MBM temp of 24C at load. Cooling a AMD 2100+ at stock freq and voltage. My attempt at unlocking was not successful and I didn't want to take it all apart to try again.

Koslov has got the basic concept down correctly. But you won't find a 12V40A power supply for any rational amount of money. You might however be able to find a 36V or 48V power supply that could supply 7 to 10 A that could work. Just put the TEC's in series instead of parallel.

Here is a pic of a similar model to the one I got as surplus.

[murray13]

oops

Oh by the way I'm running the coolers two fans in series (half voltage). It's much quieter that way. If I run the fans at full speed the water will stay below freezing and since I don't have everything insulated I run in quiet mode.

[jtroutma]

This sounds very much like an idea I had a while back.. about taking 3-4 pelts and running them between an old PII heatsink (used for a res. chiller) and a copper waterblock (separate loop to radiator).

It would be made using two independent loops and pumps. One side for the chilled water and the other for the hot water.

The res. would simply be a 2" copper tube that would have one section cut out to accomidate a PII like heatsink in which the cold sides of the pelts would be attached. The water would flow from the bottom up allowing longer contact time with the chilled surfaces. The hot side would be a loop of its own going to a massive radaitor setup.

[David|SP]

OK i think it´s too risky too, i sat down and thought about it, and the water would get hotter and hotter.

I´ll just put the peltier on the cpu, with proper insulation, and a thermal swith protection (i´ve burnt 3 processors in 2 months and i want to keep this lol)

any suggestions for a peltier? like how many W, etc... i want to build my own power supply so that won´t be a problem.

another question: how about an 80W peltier for cooling the GF4 AND its memory?

thanks again

[jtroutma]

An 80W for a GF4 could work. There was a guy that made a copper plate that when combined with a pelt would freeze both the core and the mem nicely.

Cant remember what thread I found that in though........

[David|SP]

hmm you mean some copper plate that would cover all the card, and a pelt on it? that would be enough, i think... but the problem would be applying enough pressure to the peltier.

[jtroutma]

Found the web site:

GeForce4 Copper Cold Plate

He used a dual plate design; an 80W on the front cooling the core and mem and another coldplate on the back cooling just mem.

[koslov]

Well, of course the water will get hotter the way you have it set up. You need two separate loops, or one large HSF to cool the pelts' hotsides.

murray: I've seen the Thermotek units and as far as I can tell, it's just a few pelts and a water cooling loop. I think getting a bunch of low-watt pelts and cooling them with a simple (large) HSF would be adequate for what I am thinking off.