What would happen ( waterchiller with a twist )
 

Had a few thoughts, related to watercooling, but using some pelts in there too, let me know if any of them make sense.

Thought 1 :

Before the CPU waterblock, water goes through a block mounted on the cold side of a low powered ( 40-80W ) peltier, hotside aircooled.

Would this drop the temp of the water by say 2-3 degrees, just before it enters the block? If so, wouldn't this give better temps, and be reasonably simple to implement?

I know that a peltier in this range, if mounted directly onto the CPU, would be a disaster, its not powerfull enough.

As I see it, the main advantage of this system would be that the pelt would be very easy to power, probably enough to use the PC's own PSU, if it was a good one.

Thought 2 :

A slight modification of the above idea, but with a more powerful peltier ( 176W or so )

The water would go over the cold side just before entering the CPU block, and the hotside of the pelt would be cooled by the water coming out of the CPU block, before it went back to the radiator.

Basically, a waterchiller, but using only one loop. Condensation wouldn't be an issue, the water wouldn't be cold enough, and again, it would be very simple to implement, 2 standard maze blocks would be prefect, and it would be a very simple matter to get enough clamping pressure to run the peltier in a very efficient way.


Would either idea actually work?

[EDIT] Rayman2K has already posted the first thought, so ignore first Thought above. [/EDIT]

Idea #2 has been posted before too!

The problem is that the chiller needs to have a large surface area, to significantly cool the water, and that means a very flow restrictive block, with a zillion channels going back and forth.

Overall, you're also adding heat to the total loop, so you have to improve your cooling performance, by either using a bigger heatercore, and/or a faster/bigger fan. Otherwise, the heat will raise your overall coolant temp, and you'll loose all the benefit you aimed to get.

Someone should do the math on whether it is more cost effective to set the house AC to 65 and put the radiator on the vent compared to running all these big peltiers to chill water.

Now look, I'm barely getting enough sleep as it is, you can forget about making my alarm clock ring at 3h00am!

So assuming that the heatercore is nice and big, and the airflow is good enough, would you see lower temps at the CPU?

And will the lower water temp more than compensate for the loss in pressure/flowrate created by adding the chiller section?

Why do any calcs? Look at your electric bill in July and August and just how bad it would be with the A/C set to 65. I'll use power supplies and pelts, thank you. And what do you do in winter?

In fact one month's additional electric cost would pay for the added cooling gear and a few months of use.

It's going to depend a lot on the design of the block that you're thinking about using.

If I was going to give it a shot, I'd make one of XJinn's block, and solder a series of return pipes (i.e. "U" shape), and clamp a couple of those bad boys together.

You will definitely need a very powerful pump to push water through it though.

Have you run a few 200+W peltiers before 24/7? I am not so sure that the cost is SO different considering you are running the AC anyway in the warmer months (just not quite so cold). And in the winter you put the rad outside :)

Erm neither will work too effectively GTA, you need 800watts of tecs in cooling power not input power to make an effective solid state chiller.

As for electricity bills, pfft, 99% of you leave a 400watt or multiple 400w computers on 24/7 whats a few hundred watts of tecs...

GTA read the article on the front page here about stirling engines :)

]JR[

i may be mistaken but i'm pretty sure a 400w power supply generally doesn't pull that much power (since no one really uses the 400w).

Most PC PSU's are around 70% efficient. (Power Out / Power In)

So, if a PC is drawing 280 Watts out of the PSU, around 400 Watts will be drawn out of the wall outlet.

Hey GTA,

I've tried both... :)

1. Placing a TEC cooled waterblock right before your CPU waterblock will act as a chiller and drop the water temp a few degrees. If you are air-cooling the hotside of the TEC it will take a big heatsink with a high CFM (very noisy) fan. The amount you will be able to cool the water will depend on your amb air temps and how cool you can keep the hotside of the TEC. I was dissipointed in how little (about 3 deg C with 172 watt TEC) it atually dropped my CPU temps.

2. I used two 226 watt TECs sandwiched between two Maze2 waterblocks and a Maze3 on the CPU. One E-1250 pump, two heatercore rads in parallel and 1/2" ID silicone tubing.

pump > Rads > TEC cold side block > CPU block > TEC hot side block > Res > pump

It actually made the water hotter... :confused:

This really drilled home the point of how much heat the TECs themselves dump into the system. Flat didn't work. You have to use two loops: hot side and cold side.

IMHO TECs are not very efficient and are very expensive (with dedicated PSUs) when used as water chillers.

Good luck,

Cheers all.

Very interested as to why you got higher temps with the second idea Robotech, seems very odd.

What sort of rad were you using? As BB said, and I agree, a very large rad would be required for this to work.

Kudos for trying both methods though, and cheers for sharing the results.

Looks like its back to the drawing board :)

Correction: After checking my notes the two TECs I used in the single loop test were both 172 watt Pelts (not 226 watt).

I found running a TEC chiller in the single loop configuration didn't work because the TECs themselves dump so much thermal energy into the water that it overloads the system (radiators).

I used TWO Black Ice Extreme radiators in parallel (each with a Panaflo 120 mm H1A fan @ 12 V) and that wasn't enough radiator and air flow to dissipate the heat of the TECs. I never mounted the waterblock to a CPU, just running on the bench with no-load, the water temp went up.

(2) 172 watt TECs: 24 VDC x 9.85 Amps X 2 = 473 watts
Amb. air = 20 deg C
Water temp (after equalizng) = 26 deg C

I tried a couple runs at lower voltages (22, 20, 18 VDC) on the TECs and that didn't help either. Theoretically it should work but the flow rates and radiator surface area will have to be very large.

I hope this helps... :)

Here is a pic after I split the system into two loops - cold loop on the left, and hot side loop on the right... :)

http://www.leesspace.com/images/Pelt...iller_test.jpg

With this setup I could get the cold side water temp down to about 0 deg C (note frost on un-insulated cold side) no-load. The Maze3 waterblock is bolted onto a home-made thermal simulator. The orange bottles are home-made reservoirs (cell culture flasks).

Excellent work Robotech!

Are those rads BIX?

I wonder if using 2 * 172W TECs is actually worth the electrical usage... The simplest most efficient (IMO) system I've seen so far, is #rotors dehumidifier mod.

If I had the time, I'd use a couple of Xjinn blocks (modded as above), instead of Maze2s, as the heat exchanger. Heck, there's got to be a number of block designs that are more efficient than the Maze2.

But a water temp of 26 degC is not bad! I will certainly be building something similar, but with #rotor's chiller as the hot side cooling solution (sometime by the end of this year...).

Hey Ben,

Yes, two BIEs.
No, IMHO TEC water-chillers are generally not worth the expense.

26 deg C is not that great - no load and still 6 deg C OVER ambient! (I was hoping for sub-zero temps when I first put this puppy together... :) ) Some of us just have to experiment and learn the hard way...:rolleyes:

Whats the effciency of a pelt heat pump? Somehow I doubt it approaches a compressor.

Anyway, do a little math here!

4.186 j/degree C is the specific heat of water. Plug in the output of your pelt and your flowrate and see for yourself what the maxium performance possible is.

Without getting a calculator, I'd think you'll need about 500w of cooling per degree below ambient in a reasonably high flow, 1/2 inch system.

Robotech - where did you measure the water-temp in the single-loop chilled setup? I had posted about this close to a year ago in the main forum, the idea being that the coolant SHOULD be hotter in that setup (thereby increasing delta-T of rad to ambient, make rad more efficient), except for the few inches of coolant-path between the hot-block and cold-block (which would include the CPU-block) which should have very cold coolant in it.

In the CPU waterblock (where it should have been the coldest)... :)