[JoT]

I'm totally new to TECs, so bear with me.

Would it be possible to make a hollow copper cube, probably with HS type fins extending into the interior, hold water as a reservoir, and be TEC chilled?

My idea is that the copper cube acts as a really big coldplate and reservoir. Four TECs sit on top, running at only half power (high wattege TECs, to be most efficient). The TECs are cooled by a (1) custom waterblock. Water flows out of the pump, to the cube, to the processor, to the VGA, to the NB, back to the pump; this eliminates the need for ALL fans, with the exception of PSU.

I'm not looking to get sub-zero temps--heck, I'm not even looking to get freezing, I just want to get the water cooler than room temperature by at least 20F degrees (typical room temp for me is 80F or higher).

I'm not sure if I've got the pump inlet/outlets right in my diagram, but that doesn't really matter, just as long as you get the idea.


This is the only way I can think to have sub-ambient temperatures and still be a mobile computer (I take this rig to LANs).

Edit: I just realized that I have the diagram completely wrong because my brain farted when working with the WhiteWater. I'll make another diagram real quick.

[JoT]

Ph33r the paint skillz!!

Edit: Wow, the first one looks so much cleaner Perhaps if I actually do this, I should rethink the tubing layout... o_O

Edit #2: Err, I just realized, I reversed the inlet/outlets is what happened, not the WW messing with me. In the first one, the arrows go...the other way

[Shadowcat]

Hm, the diagrams seem sound but water chillers are extremely inefficient and troublesome.

You will have to protect all your tubing etc against condensation and the amount of power the TEC's will draw to chill the water may not be economically worth it. The TEC's will also be heating up your loop if they are cooled in the same loop. Also in order for them to cool anything, the water needs to be VERY slow in the reservoir / chiller part, depending on how you set up, it could hurt your flow rate or not.

Water chillers can work, but don't forget TEC's are highly inefficient, you would probably be better off just sticking one on the CPU surface.

[SwooshingRob]

Good thought mate, but unfortunately it will never ever work.
No matter what you do, the peltier will always generate more heat on the hot side than it can remove on the cold side.
That's the bad thing with equilibrium. (Hot side = heat removed from cold side + power used to remove it)

[JoT]

Even if the TECs are only cooling down water, they're not trying to run at full power, nor are they trying to remove 60-80W of heat, they are merely trying to cool down near room-temperature water.

Shadowcat, you do have a good point in that the water needs to be in the res longer; with the planned Iwaki, I doubt it would be there long at all.

I'm trying to chill all parts that are watercooled, not just the processor. This is the only way I could think of to cram a chiller of any sort inside the case.

Might there be any way to make this work, such as submerging a radiator in water, then freezing/chilling the water with the TECs?

[czand1]

I am using something called a cold plate. It is a solid flat surface with copper tubing running through it. I mounted this outside the case and used two 72 watt pelities to cool the plate with a large heatsink and fan. I did not use a radiator. With no flow through the cold plate ice would form. Must use a small amount of antifreeze. CPU temps at idle 20 load 28-32 MSI 865 Board

[Meow Cat]

Using the pelts to chill the water, not good idea.

[ogboot]

czand1 that sounds pretty cool, no pun intended, nice temps, are those on die or socket temps?

[JoT]

I'm not getting too many helpful responses here

One of the other things that could be done with the pelts is air-cool them, rather than water, to prevent dumping that heat back into the loop.

[doc]

If I understand your diagram (no labeling on diagram), you are using the same coolant to cool the TECs and the CPU ect. If so, did you ever hear of "China Syndrome". The heat must be dissipated somewhere, you have all the heat from the CPU ect. and TECs going into the coolant. You need to cool the TECs separately either with water or air. Air in tight spaces will be hard, because of the large heat sinks that will be needed.
I have a water chiller that is external to the computer, 5 gallon res. with TECs and heat sinks in the res and in the air. I can run the TECs at very low power when I'm not using my system and store some extra cooling. Then when using my system I can kick up the power to the TECs and keep it cold for longer. My circulating pump is only on when the CPU is on, I do not run my system 24/7.

[JoT]

Yeah, sorry about the labeling, I'm in school during the early hours of the day, so I was in a hurry.

When I first thought up the idea, I knew about the excess heat being dumped into the water, but I thought that since water-cooling isn't 100% efficient, some of the heat would go out into the air, possibly through the inefficient waterblock.

Since I plan on putting this whole thing into a Chieftec full tower, space isn't really an issue with this setup, so there would be room for a couple of SLK900s or something. I could stick my two blowers on the heatsinks at 12v and the system would still be nice and quiet.

What would be the best heatsink for mounting on TECs?

[doc]

If you had lets say a 40mm X 40mm TEC, I would use a heat sink that was close to that size 40mm X 40mm up to maybe 50mm X 50mm. I have modeled many different types of heat sinks, if you are using extruded Al type the fins larger than about 50mm sticking up from the base, are overkill. Also going much larger than the surface you are cooling, like hanging 10mm off on each side, is of little value. And beyond that is even less use. So from a cost point of view, as close to the size of the TECs as you can.

[JoT]

Perhaps I could get a couple of MCX462+T's and air-cool them? Would seem to be the easiest way to go about air-cooling a TEC. The only problem I see is figuring out how to mount the block to a sheet of copper.

[czand1]

The reading is from the MSI motherboard. I did the cold plate set up because i did not want to deal with condensation. The tecs do a great job and it is small and compact

[JoT]

*kicks thread to the top*

[Metalwave]

You know just reading this thread, it would seem to me JoT like you're going through a whole lot of trouble with these TEC's to chill water. You'll still have condensation problems if the water is cooled below the dew point (changes daily, but usually between 50 and 60F indoors). And SwoochingRob is right too, you'll have all the extra power used by the TECs to get rid of, which would be far in excess of what the CPU uses even if you're heavily OC'd with volt mods and other bells and whistles.

Your best, easiest and simplest bet is to mount the TEC right on your CPU and insulate the socket/coldplate from condensation (which is really not that hard). This would also be the most efficient use of the TECs and the power they comsume. It'll also make the whole wc system more efficient (as you're removing more heat with the water).

Anyways, that's my 2 cents.

[JoT]

The point is not to cool just one thing to a lower temperature, it's to cool all three. If the processor is reliant upon a pelt, then if/when the pelt dies, the proc is gone; this system would be less reliant upon things that could fail rather easily.

I don't want to get my system under sub-zero, heck, I don't even care about getting it to freezing. I just want it to be lower than room temp, which is usually 80-85F.

[JoT]

*bump*

[Unoid]

think of a way to get a giant heatsinks atatched to the pelts and col them with 1-2 120mm fans blowing over. it could be done to work efficiently but takes a lot of deisgn.

[JoT]

That's why I asked about the MCX462+T's, they're built with peltiers in them, and are designed perfectly to take away the heat of a TEC.

[Exoce†]

Originally posted by czand1
I am using something called a cold plate. It is a solid flat surface with copper tubing running through it. I mounted this outside the case and used two 72 watt pelities to cool the plate with a large heatsink and fan.

Very cool, I built something very similar using laminated acrylic to build a body over a piece of 3/8 Aluminum billet. I haven't used it yet but I am happy to hear that it might actually work.

[JoT]

*bump*

I'm still not getting much help on this idea. Since starting this thread, I have learned that you cannot cool a TEC with the same watercooling system it is attemping to cool, but that's pretty much the end of that.

Could someone please tell me how well the MCX heatsinks would work?

[Lotec25]

JoT,

Me and my nephew were talking about this. We were thinking of using like a cooper reservoir to run the water threw. And attaching some pelts to it. I think it would work and probley work preety good. We were going to use 2 220w pelts. and hook them up so they ran on there own that way we could leave them on let it cool and then start the system later. I think it is a good idea i just never followed threw on it for that fact that i am just going to go all out and get an A/C unit and freeze the heck out of the water.

Hope this helps some. Lotec25

[SniperXX]

JoT I think using the MCXs should work. I dont know much about them but they should cool the TECs well since Swiftec sells good quality products. I would read up on some reviews of those heatsinks. One thing I think I remember hearing is the more air you are cooling the MXC with the cooler the thing runs.. Basically I would just read some reviews like I stated earlier in this post, then just go for it and try it if you are willing.

[JoT]

Eh, guess you're right, Sniper, I just have to try and see if it works.

[kanuuker]

I think I am attempting something similar. I just did it this morning so no real results yet, but I'll post them later on.

Anyway, what I did is dismantle one of those mini-fridges that hold a couple of cans. The cooling apparatus has essentially 3 parts: peltier, heatsink, and an alluminum block. I'm not sure what the block was for, it had something that I"m guessing is a thermomitor attached to it - I removed it.

I opened the top off of my reservoir (an electrical junction box) and simply submerged part of the block with the peltier and heat sink still attached, makiing sure the peltier doesn't come in contact with the water. Note that my cooling setup is external so I have more room to work with.

I only did this 10 minutes ago but my temp has dropped 1c to 20c.

[felinusz]

If the processor is reliant upon a pelt, then if/when the pelt dies, the proc is gone; this system would be less reliant upon things that could fail rather easily.

I'm pretty sure peltiers last forever, and won;t stop working suddenly, altough I may be wrong.

Why not do the same thing czand1 did, but use the Swiftech MCX heatsinks in place of the heatsink and fan czand1 used? It would be somewhat compact, probably very effective, but pricey to build. Condensation would probaly still be an issue as well though.

I think that if you can fabricate a copper tubing coldplate maze type thing as czand1 has done, and setup solid mounting points for the MCX heatsinks (one on each side would be effective), you'd have a Air cooled Peltierwich in place of a radiator.

The only major problem would be the restrictivity of the copper coldplate maze the water travels through. You would have to have several 'passes' to make the most effectie use of the two MCX peltiers.

[Jawadali]

Originally posted by JoT
Might there be any way to make this work, such as submerging a radiator in water, then freezing/chilling the water with the TECs?

I was thinking the exact same thing.

I want to get into water cooling, but only if my temps drop by ~20C (from mid-40s/low-50s)

kanuuker seems to have a good idea.

I was thinking to have the radiator submerged in a sealed plastic container. At the top of the container, there would be a small cutout were the pelt makes contact with the water via a cold plate. The other side of the pelt would be on the outside of the plastic container, and would be cooled by a heatsink + fan.

This setup may or may not fit inside a case, though. It depends how big the radiator is, and how tall the peltier heatsink is.

The radiator would then cool the water flowing through it.

The only things I don't know are:

1) How much wattage should the pelt have

2) How to power the pelt

3) What type of heatsink I would need

4) If I would be allowed to get a water cooling system in the first place

[kanuuker]

Well, my experiment didn't do squat. My temps only dropped by a degree. I would imagine that if I used a higher wattage pelt (I don't know what the one I used is) I would get more of a difference. Oh well, I'm still sitting pretty at about 21-23C.

[Jawadali]

kanuuker, if you don't mind, can you elaborate on your experiment a little more?