Peltier waterchiller?

[aasmaukr]

I'm thinking of trying to build a peltier waterchiller with eight 27w peltiers. Do you think this wil be okay?

Here's the specs of the peltiers:
Dimensions: 25x25x3.7mm
Imax/Vmax: 8.5A/5.9V max.
Qmax: 27.4W
49 couples

[Holst]

27x8=218

This should work ok for upto a 100watt heat load. (graphics card or CPU, but not both)

How are you going to cool them / power them?

These sorts of small pelts running in paralell will work as a chiller.
Most people never build these systems for cost reasons.

6v is quite a nice voltage as you can run 2 together on a 12v supply or 4 together on 24v. Should make choosing a PSU easy. 24v will probably be cheaper as you wont need so many amps.

[aasmaukr]

27x8=218

This should work ok for upto a 100watt heat load. (graphics card or CPU, but not both)

How are you going to cool them / power them?

But, the water wont give a 100w heat load? I was thinking of cooling GPU, CPU, and chipset.

I think i will have two water cooling systems, one for the peltiers and one for cpu, gpu etc. I will make eight waterblocks where the water is in directly contact with the peltier or i will make a big block for all of them.
I will use the 5v line or run two and two on the 12v line.

[Holst]

Remember that the TEC is rated for 27watts at Qmax!

This means that if you put 27watts onto the "cold" side you will get a delta T of 0 .. so you wont be any cooler than whatever temperature is at the hot side.

To get delta T of 20-30*c then you must load the TEC below Qmax.

To acheive this you must have TEC that has a Qmax capacity of around double your heat load. Any less than this and you will overwhelm the TEC and your temps will suck!

Read the TEC stickies if you want to know more.

As for the waterblocks, With such small TEC I think that one large block might be easyer and cheaper to make. But it depends on what your capable of making.

I would not recomend having the water in direct contact with the TEC hot sides.
Its hard to stop leaks this way and having a more conventional copper waterblock will work better anyway.

Good idea using a PC power supply and splitting the load across the 5v and 12v lines You may even be able to twek the 5v lien upto 6v to keep them all running at max power.

[aasmaukr]

Remember that the TEC is rated for 27watts at Qmax!

This means that if you put 27watts onto the "cold" side you will get a delta T of 0 .. so you wont be any cooler than whatever temperature is at the hot side.

To get delta T of 20-30*c then you must load the TEC below Qmax.

To acheive this you must have TEC that has a Qmax capacity of around double your heat load. Any less than this and you will overwhelm the TEC and your temps will suck!

Read the TEC stickies if you want to know more.

As for the waterblocks, With such small TEC I think that one large block might be easyer and cheaper to make. But it depends on what your capable of making.

I would not recomend having the water in direct contact with the TEC hot sides.
Its hard to stop leaks this way and having a more conventional copper waterblock will work better anyway.

Good idea using a PC power supply and splitting the load across the 5v and 12v lines You may even be able to twek the 5v lien upto 6v to keep them all running at max power.

But do you know how much heat the water gives when it has gone through cpu, gpu and chipset? Will 8 27w peltiers be enough?

I have made a block where the water is in direct contact with the hot side of the peltier, and there was no leaks and it worked okay. I have a lot copper and 6mm plexi, but i think i will make a big waterblock for all of them since it's a bit more simple...

Thank you.

[n00b 0f l337]

8 wont be enough. Now if you get 8x 80watt peltiers, and run those at half voltage, then you could hold around 160watts of heat. Still not much. The plan I'll do is going to be 14-18 320 watt peltiers, running at 1/4 or 1/2 depending on benching or fun.

[Holst]

I dont think that 8 27 watt pelts will be enough to cool CPU. GPU and chipset when overclocked.

It will work but temps wont be as good as they might be.

Its difficult to predict exact temperatures, allot of these things need to be tested.

The nice thing about using those pelts is that you could allways just add a few more should you need them.

If you can make a block that cools the pelts directly with water that wont leak then go for it. It will work pretty good and will be cheaper than buying loads of waterblocks.
Can you take any pics of the ones you made allready?

[aasmaukr]

Okay, the pics are here: Link

I think i will give it a try with eight 27w peltiers. If it performs real bad, i can always put 4 of them on my 6800le's memory.

I think i have some pictures of the block before it was finished, but they are on my brothers camera, so i do not have access to them right now.

[wittekakker]

Don't look to much @ Watt's here because when chilling water the amount of energie to be cooled in a period of time is just not the same when using pelts ondie.

What I want to tell is that 8 pelts and big reseroir with not much flow could work very well to chill your comp.

Only hard thing to do is to not heat up your waterloop with ambient room temp :s

[gbomb944]

Keep us posted even if it doesn't work so that we may benefit from your experience. I've been wanting to build something like this for a while now but will use air cooling for the hot side. Most of the time when somebody says "peltier" and "waterchiller" in the same sentence they get deluged by "won't work" and "build an AC chiller" responses ad nauseum. Hopefully those who feel that way will practice restraint and not try to discourage you. I'm subscribed and look forward to any updates.

One tip that might help is to use a dc pump that you can control the speed in order to keep pump heat as low as possible on the cold side. Good luck.

[Bloody_Sorcerer]

pumping heat on the low side shouldn't be more than, say, 15 watts (D5 at seting 5), which could be lowered drastically by using a different pump, such as a 50z (something like 9 watts?) without sacrificing performance.
good luck... while i think you're nutters, i salute your brevity/stubbornness

[ls7corvete]

What I want to tell is that 8 pelts and big reseroir with not much flow could work very well to chill your comp.

Lets not start about how less flow will improve performance.

Keep us posted even if it doesn't work so that we may benefit from your experience. I've been wanting to build something like this for a while now but will use air cooling for the hot side. Most of the time when somebody says "peltier" and "waterchiller" in the same sentence they get deluged by "won't work" and "build an AC chiller" responses ad nauseum. Hopefully those who feel that way will practice restraint and not try to discourage you. I'm subscribed and look forward to any updates.

No one has suggested a phase unit yet. There is a difference between discouraging new ideas and simply informing someone that they do not have enough power to have a temperature drop from the TEC.

He either needs to reduce heat load(only cool cpu for example) or increase the TEC size. Sorry but 200w is not alot of power for a TEC. This project is a failure from the start unless changes are made like I said

[Holst]

Don't look to much @ Watt's here because when chilling water the amount of energie to be cooled in a period of time is just not the same when using pelts ondie.

This is totaly wrong.
A TEC chiller works in exactly the same way as an ondie watercooled TEC.
The wattage of the CPU is the same, and if you want the same temps you will need the same cooling capacity.
You need to pay as much attention to having appropriate wattage TEC as you do in a on-die watercooled setup .. probably more so.

What I want to tell is that 8 pelts and big reseroir with not much flow could work very well to chill your comp.

As I have allready demonstrated the above 8x27watt system is insufficent for both CPU and GPU.
Lower waterflow will give you WORSE performance in such a system!
This is just a fact, not open for discution.
Follow this progretion of logic .. if lower flow = better performance, no flow at all = ??????????????
its the CPU your cooling not the TEC cold side waterblocks!

Only hard thing to do is to not heat up your waterloop with ambient room temp :s

Valid point but this is the easy bit.
Insulating the waterblocks, pump and tubing is fairly easy. neoprene and armaflax are your friends.

[wittekakker]

Jup, true. Donno how I came up with those ideas

[Bloody_Sorcerer]

yeah; many people feel that lowering flowrates should increase time in cooling mechanism (be it radiator or peltier or whatever) but in reality it changes nothing; think of it like a racetrack. you go around it at 60 mph and you will be in a given spot for 1 second every pass around the track. double your speed and you will only be in that spot half a second every pass, but you'll pass around twice as frequently, so you net equal time in that given spot. While this isn't a perfect example, it is completely valid.
Higher flow improves performance. its a proven fact; europe is wrong.

[n00b 0f l337]

Europes not wrong persay, they just think different. They're not on ur idea, but low restriction slow-flow blocks, so in that 60 second loop of yours, the waters in the block for 2 seconds instead of 1 or .5 and in that time there idea is it will pick up more speed. Depending on how you design things, either way is good, but high flow is better by a good margin. Lets not start that flame war again shall we? Back to the tec chiller. Ive got some good ideas if I can find some money to make em. But flow isnt a problem at all.

[Bloody_Sorcerer]

yeah, and european and american are slowly convening... the UK though, wow, they really got it right (thermochill, anyone? )

[n00b 0f l337]

Here's a sneak peak of my design...

[aasmaukr]

I don't think that i will make a peltier waterchiller anyway. Now i'm looking at direct die phase-change. I'll buy a old freezer and use everything but the condenser and the evaporator in a new system. I'll fill it with r290.

But i have to read a lot more about phase-change cooling before i start the project.

[<Wouter>]

I don't think that i will make a peltier waterchiller anyway. Now i'm looking at direct die phase-change. I'll buy a old freezer and use everything but the condenser and the evaporator in a new system. I'll fill it with r290.

But i have to read a lot more about phase-change cooling before i start the project.

Good idea. You could also add those tecs after the rad in your wc loop. That should give you a couple degrees colder water.

[wittekakker]

Nope. U know, it's not only your cpu, gpu and nb who's heating up the TEC's then, but also your radiator. And even some small radiators can do a lot.

[n00b 0f l337]

Adding the tecs after radiator doesnt work. Thats not how it works. the area after tecs wont be any "colder".