Cooling Water-cooling with the water cooled by pelts?

Hi all, right, this is my big idea !

Seeing more and more cpus/gfx cards cooled by pelts which are then in turn cooled by water, i thought how about using one (or more) pelts to cool the water before it reaches the cpu/gfx block? In effect the pelts would be doing the work of the radiator, but i was thinking about the radiator reducing the heat after the cpu block, then the cooler water going thru the pump, then being cooled down EVEN further by a copper block connected to say a 300W or 2x110W pelts or something like that...

Would that work?

Any ideas how cold the water would get (obviously there would be heatsinks cooling the opposite side of the pelt) ??? (i thought this might avoid condensation issues of standard setup)

Any possibility of water freezing in the lines ??? (in this way there is only a small distance for the V.COLD water to travel before it gets heated up by the CPU/whatever)

And how about overall efficiency of this kind of system, would it be more/less efficient than the normal CPU>PELT>WATERBLOCK system ?

I also thought it would be even better to use a slower pump than a fast one as this means that the water will have more time in contact with the cold pelt-cooled copper block which means the water entering the cpu block would be colder than with a fast&noisy pump! hooray for quiet pump!

Just an idea i had lol, could be a fun project!

Thanks guys !!!!

PS (sorry this is double-posted, did not see this section at first ! lol)
http://forums.bit-tech.net/showthread.php?t=82188

 

you still have to cool the hot side of the TECs...Some people have tried before to make chillers but not with alot of sucess...

 

Not sure if this is possible, if you think about it it's like putting a pelt onto a CPU block upside down, the heat has to go somewhere. Getting it to cool the water effectivley would be hard. Stick safe and with a high quality radiator

 

It would work, yes, but what are you going to stick the hot side of the pelt to? Heatsinks? Congratulations, you've just created an incredibly high power-using heater...INSIDE THE CASE.

Of course, if you externally mount the heatsinks, or add a bunch of fans to cool them, you'll just have a heavier version of a radiator that sucks up a bunch of wattage. But your water might almost be as cold as if you just stuck a radiator in there instead.

 

yea i was thinking that might be a bit of an overkill, i do know about cooling the hot side of the pelt and i do know it throws out about 3 times more heat than it suck up, so i was thinking of a mammoth heatsink on top of the case ? (i can manufacture my own using my schools milling machines/laser cutters etc)...

AND

the pelt block could potentially cool the water very close to 0C just before it enters the cpu block,
what temp could a decent 120mm rad redude the water to ? (i always thought nowhere near 0C)

any ideas ??

 

no cooler than the air that it is blowing accross the fins of the radiator...

You could do what your planning sure, it just wouldnt be the most efficent way imo...

 

Well, you're assuming a lot there...water temperature is much more a function of the equilibrium of the system...not everything near the pelt will reach that near 0c, in fact most won't. So plan on how much you can equalize the heat in the system as opposed to how much you can chill a couple molecules to.

It's true you can't get down to 0c or anything below ambient with a radiator setup, but remember that you're going to be creating a much smaller area with less contact time, so you may not even get that performance from the pelt when you figure you're really cooling the whole loop.

Unless, of course, you put in a *massive* TEC plate, and then your heatsink will function like a heatlamp when your computer is on, and probably draw enough power to black out 3 neighboring communities.

Well, not really, but you get the idea...I think it's just a bit impractical to build it this way. You're better off taking a TEC and sandwiching it between your waterblock and another flat piece of copper, then attaching the whole thing to the CPU (which is routinely done). Less elaborate, a lot more efficient.

 

Yeah :S ... i think an old 200w PSU could provide the 10A or whatever the pelt needs, so thats already making the case bigger ...

Having thought about your post - if potentially the water can be cooled to ambient temp in a standard rad with fan, say 22C,

and the pelt then further cools the water down to as close to freezing as possible (but NOT freezing lol) say, 3C, does that not give (potentially) 19C reduction in CPU temp over a standard water cooling system?

If this is true, it can even challenge the normal CPU>PELT>WATERBLOCK efficiency? Or not?
Also MINUS the risk of condensation around the cpu! At expense of condensation of the stretch of water-tube between the pelt-block and CPU block...

What do you think ???

 

Cather tried something along these lines a couple of years ago >>clicky<<

Not an ideal day to day system but good fun non-the-less

 

Good ideas, but still problems. Nice to see you're thinking this way, though...insanity is half a step towards invention, I always say!

The idea of cpu - tec - w/b is for a very specific purpose: It's the TEC that cools the cpu, the waterblock is then just exchanging the heat of the tec (which it can do better than a standard heatsink). So, you'd have a hard time being more efficient than that.

It would be fairly inefficient efficient to do the tec on the rad because the rad itself is a heat dissipating object. You couldn't get the cold temps INTO the tubing THROUGH a rad, it is only supposed to take the preexisting temperature differences and bring them to ambient by exposing it to larger surface area. What you COULD do is make a copper block, drill a hole in it to hold your tubing, and then attach the pelt, and attach a heatsink on top. But it would still be much less efficient than the first system....all you'd really be doing is making another waterblock with a tec, but not having it cool much of anything.

 

Dego that was my idea exactly, but for even improved contact area i thought about a standard CPU water block, but instead of using the waterblock to conduct heat it would be used to conduct the cold INTO the water before it enters the CPU waterblock. The the hot side of the TEC into a big heatsink ontop of my case!

Waddoya reckon!

Lol sure sounds like a bit of expensive-time consuming fun in hope to defy the undefyable efficiency of standard CPU>TEC>BLOCK cooling lol!

 

I just about managed to skimp thru the 24 pages (!!!) of Cather's thread, it seemd like good theory but failed in the end (????), i had something slighly different in mind, however i think his idea is like mine but elaborated (he got there first lol) with the two loops&that. I think i will still try it anyway, just purely to see, because i can! . So basically standard watercooling just with a TEC chiller right before the CPU block, im probly thinking a small copper block (waterblock would not have very large surface area for contact with the pelt), so ill try a few sketches and see if i can get a small&as efficient as possible block to attatch 2x80W pelts which should run off a dedicated 12V 250W psu.

Im starting to think of it as just slightly enhanced watercooling in terms of performance:S... I think the test here is to see whether it would reach same performance as the standard CPU/TEC/WATERBLOCK configuration, as the condensation issue around CPU tends to trow me lol (ive heard some stories... )

I keep thinking ..................................

 

Hehehehe....Mod on, dude, mod on....

The copper block would be interesting, but you're going to have to restrict flow if you want the TEC to take effect...hence the suggestion of a waterblock instead. What you almost need is a waterblock with inlet/outlet on the sides, all copper, and put a tec on each side.

After all, that cooling isn't going to be too useful if the water barely stays in it...so don't just make it a through-channel.

This could get really interesting...

 

I could see this working on a much smaller level. Like say using a 40 or 80 watt TEC and also a radiator. Couldn't these TECs be effeciently air cooled? Have the radiator in line first to cool the water to about room temp and then have the TECs cool just a little bit further. It wouldn't be a big difference, but it would work right.

Or instead of one big TEC, how about a few small ones? Like 3x40 watts all passively cooled. This would have to chill the water somewhat. Please correct me if I made any wrong assumptions as I am new at this stuff.

 

Right idea, but there's one problem, wedge. TEC plates can't really be air cooled. They function by pulling the heat using current from one side to another, almost like a pump. So if you don't have the ability to dissipate heat from the hot side, you lose most of your function to cool it on the cold side. Thus you fairly well run a bunch of electricity for no reason.

The hot side of the TEC needs SOMETHING to dissipate that heat, usually in the form of a heatsink and fan combo.

 

Another problem would be preventing the water from freezing, wouldn't do the pump or the loop alot of good

 

Swiftec make something like your discribing cooling the water with pelts then cooling the pelts with water it the "MCW-CHILL 452™ Thermoelectric liquid chiller" it cool the water in a second loop the the other loop cools the Pelts

 

Oooh the swiftech block looks nice, tho did you see the price at the bottom *ouch* !!!!

Well i had a small look today in an engineering part of our school and they have big round blocks of aliuminium about 15cm diameter and 5cm tall, (no copper unfortunately), and looking at it it would easily have the capacity to cool a couple of small pelts, will just need to machine it like a standard heatsink to increase the surface area so that may be worth it !

That would look right crazy sticking out of the top of your case ...

Im going to slowly accumulate these parts that i need and will keep posting up on this board, i decided ill go for it it will just take time, im going through a few PC upgrades at the mo...

BTW i think the right mix of antifreeze/water would allow me to run the 'coolant' sub-zero!

I appreciate all your comments guys !
Will keep posting pics, drawings etc watch this space !

 

It's been done. Overclex.net did an article way back where they stuck a pelt to a huge-ass aluminium heatsink, mounted on top of a lid of a reservoir with the fins of the heatsink sticking through the lid and submerged in the coolant inside reservoir. The hot side of the pelt had another (huge-ass) aluminium heatsink stuck to it cooled by a 120mm fan. No radiator required --it worked a treat.

Nevertheless taking this approach will lower your coolant temps below ambient and condensation is not far behind. Unless you use a regulated pelt power supply with a temperature probe connected to the CPU to keep it nicely within the ambient range.

 

Nexxo,

I think I saw that article you're talking about. I thought it was a rather brilliant idea to mount the heatsink INTO the res. The no-rad solution is great, except for that little problem you just mentioned...it will require an entire cpu-controlled power supply to keep temps near ambient and prevent condensation, and that is just a pain.

By the way, am working on a workaround to that...my buddies and I are developing a "smart" peltier system.