Is a peltier water chiller a dead idea?

[Exoce†]

Suggestion?

I've been thinking about trying a setup of this sort, I figure the best way would probably be to use more, and more powerful peltiers than actually required and then relay them through a thermostat of some decription. This would prevent icing in the reservoir and if desired keep the coolant warm enough to prevent condensation on the hoses. I speculate (I'm kinda sketchy here) that you could use a smaller reservoir as well. While it might work like a hot **** it'll probably be very expensive too.

I'm new here so all this might be old news, otherwise; If I ever trade in the Wife for my computer habit and build it, I let you guys know how it works out.

Exoce†

 

[Paxmax]

Well, Exocet,(<- I'm not gonna plow through my keybrd to find that "T" ), If you got the TEC power to cool the CPU 24/7 you'd be better off putting it directly on CPU with a coldplate. That way you're not adding pumpheat to TEC coldside, you don't have to have it on while you don't 'pute, less losses in insulation, more compact.

 

[doc]

I've been gone for a couple of days, but I see Paxmax needs a little help on the calculations. The specific heat of water is

1 cal. / gram C not 1 J / gram C and a calorie is 4.18 Joules

A gallon of water is about 3800 CCs and we will assume that at the temps we working at the density is 1 gram / CC. That means
we have 3800 grams of water, which would take 3800 calories or
15884 Joules to raise the temp 1 degree C. One watt is 1 Joule/sec therefore 100 watts is 100 Joules/sec. So for 1 gallon of water ( 3800 CCs) we divide 15884 Joules by 100 Joules/sec and get 158.84 sec or 2.64 min to raise the temp of 1 gallon of water by 1 degree C. Now if you are running a TEC during this time, lets say maybe a 40 watter. We would subtract the 40 watts of heat removed by the TEC from the 100 watts from the CPU. So then it would be 15884 J / 60 J/sec, which would be 264.73 sec or 4.41 min per degree C. If you have 5 gallons or the temp can rise 5 degrees the time would be 5 times as long.

 

[Caffinehog]

You don't need fancy equations to figure this out.

The closer to room temperature, the less heat the radiator gets rid of.

At close to room temperature, (and we know that water can hold the wattage of a processor without rising more than a couple degrees a pass,) the peltier is moving essentially all the heat of the processor. If the pelt can't handle the full heat load of the processor, it can't cool it and the temperature will rise or at least stay constant.

So if your pelt is less than the wattage of the cpu, it can't get below ambient.


Ever.

So you're going to need a massive pelt. And that's not cheap. Nor will be the power supply for it.

 

[Paxmax]

doc! Thanks I'm lost without you !
Argh.. I was so close to nailing it... I "just" messed up calories and Joules. I think I got misinformed on another thread(excuse excuse)
I was thinking.. was it really that bad ?!?!


Caffinehog: Some of us are talking about PRECHILLING the water and then use it, for a limited time... some posts are about 24/7 driving of CPU, what scenario are your post adressing?

 

[krag]

Alright...alright...settle down fellas I got something you guys might find interesting. one of the reviewers on [H] or 3dgpu or NVnews (can't remember)uses this little baby to cool his h20 setup. And he said it works killer. If I wasn't such a Vapochill slave this is what I would be using.

 

[doc]

Looks like a TEC to me, so whats the diff.

 

[Caffinehog]

I wasn't considering pre-chilling. If you are interested in that, though, a bunch of ice cubes in a reservoir will be much more effective.

 

[doc]

Hey, I've used that before. It works OK, ice is 334.4 Joules to melt a gram of ice. Works great in the winter, lots of free ice.

 

[Caffinehog]

Watts=joules/seconds

or

seconds=joules/watts.

so, for one gram of ice and a 100W processor....

seconds=334.4 joules/100W

A gram of ice would last 3.3 seconds.

If you were to use a couple ice cube trays full, lets say 1000g of ice, well, that would last almost 55 minutes.

Of course this assumes you water starts at 0C and your ice also starts at 0C. I know this is not the case, but I also know the water will be warmer and the ice will be colder.

 

[doc]

A poor mans phase change cooling.

 

[wymjym]

Hi,
FWIW………this unit has been chugging along for around a year now. (I had planned on upgrading it but never found the time)
It hasn’t been apart since I changed out one of the 71 watt pelts for a 156 unit, that was last October.
I currently have a xp2600 running at 2600mhz. My ambient room temps are around 26C, the chilled water is always around 20~21 (because of the VPC controllers) and the cpu temps under load never rise above 33.5….idle is generally around 28. The computer temps are coming from MBM…the water temps from external thermometers.
I am totally satisfied with this setup, but it is not extreme.
I have been accumulating parts for my next unit. It will have three copper blocks that measure 65mm x 225mm. I can then mount up to six (6) 50mm, 226watt pelts into this sandwich (up to a total of 1,360watts of cooling). I will use several VPC controllers to control the water temps as in my current setup but I am also going to use a pelt or two on the cpu and chipsets. These will also be controller controlled. This way I can still have a ‘pretty’ mobo setup (I hate the looks of insulation) and when I want colder temps…I just adjust the dial and it automatically keeps things at that set temp. (oh one other very important thing…this new setup will have the mobo in a ‘cooled’ container so that I can achieve my condensation free operation at much lower temps than ‘room’ ambient.


I have a friend that is using an ac chiller to cool the water that is going to the hot side of his pelt water blocks. This way the pelts (on the cpu, gpu, etc) can still be controlled for temp but the potential for really cold operation exists. (if the cooling water temp is lowered by 10 degrees then the cooled side is lowered by almost the same amount).
So…..it is all about what you want, I am thrilled with 2600mhz (will do 2650 but the memory is flakey at that speed). I could probably get 2700~2800 out of it with really extreme cooling but it isn’t worth it to me, to some it is.
wj

 

[four4875]

if you put the pelt after a rad that cools to ambient, you CAN reach sub ambient temps on a cycled pelt chiller system. but it's not worth it IMO. i'd just get a little mini fridge and put your rad in there.

 

[wymjym]

if you put the pelt after a rad that cools to ambient, you CAN reach sub ambient temps on a cycled pelt chiller system. but it's not worth it IMO. i'd just get a little mini fridge and put your rad in there.

I have NEVER seen that approach work. I have never heard of that approach working, I have heard people postulate that it would work...but no proof. The pelts only cool a certain amount every time the fluid passes by them. The radiator 'equalizes' temps.
If you have a big enough compressor in the minifridge that will work well. You could also use the cooled fluid from the minifridge to cool the hot side of the pelts which would in turn make the cooled side (of the pelts) even colder.
wj

 

[Bender]

I think some of you are unwilling to accept or try new ideas. For crying out loud my Thermotek water Chiller is AIRCOOLED. It has 10 (yes 10) 54 watt TECs cooled by 2 decently sized heat sinks and 2 120 mm fans. The chiller set at max cooling spits out 12v for a total of 440 watts cooling power. It is capable of taking a 225 watt heat load down to ambient or 95 watts -20c below ambient.

Don’t fall for the misconception that you cannot make anything near the performance of this $2000 (retail) chiller. The chillers specs are excellent but there is major room for improvement. The hot side of the TECs uses standard white thermal goop BUT the cold side uses a thermal pad much like retail Intel heat sinks. Using even cheap thermal goop on both sides will increase cooling capacity and increase efficiency greatly. Using AS3 or Ceramique would give even further performance gains. The water block, which removes the cold water for circulation is aluminum and has poorly designed internal fins. Simply switching to a decent copper design is good for a few more degrees dropped. The 2 aluminum heat sinks clamp the assembly together with 8 Philips head screws. Hex head screws offer much needed clamping force to the TEC assembly. All added up there are huge performance gains to be had without increasing power consumption.

If a company can turn out a product like this, there is no reason you can’t whip up something better in your basement with crude tools. You might not have the fancy electronics to control the unit but they really aren’t needed or wanted if you are going for absolute coldest temps. I believe TEC water chilling is capable of being more efficient than a water cooled TEC on top of the CPU. The TEC on top of the CPU makes for one big problem; you have almost no surface area to focus your water cooling on. With a TEC chiller you can use low power TECs spread out over a surface area much larger than that of a CPU water block.

If you’re not looking for extreme low temperatures you can still do quite nicely with a single 80 watt TEC. Here is what I would use with an existing water cooled setup.

Low power TEC
2 heat sinks
1 decent 80mm fan
A reservoir tapped into your water if you don’t already use one
RTV sealant
4 hex head screws

The low power TEC uses the heat sink / fan to keep the hot side cool. The second heat sink will reside inside the reservoir, directly fastened to the cold side of the TEC using 4 hex screws. Use RTV sealant for everything that needs protection from water. This is not something I thought up entirely on my own. I forget who it was but someone here at OC did something very similar with respectable results. If I remember correctly it dropped water temps 5-7c! That’s a nice improvement for not much work. You don’t even have to worry about condensation until you get a lot more TEC power.

Don’t be afraid to try something new

 

[four4875]

thanks bender. im glad someone agrees with me that it'll work. i think i might look into havin a nice copper block for TEC chiller use milled up. put HSFs on 2 (or if i could get complex enough, 4 sides) make it like tunnel and spiral around the inner outside of many blocks put end to end like. then mountup all of the TECs and HSFs. and have a cooling tube... just my idea.

 

[Nico3k]

Alright...alright...settle down fellas I got something you guys might find interesting. one of the reviewers on [H] or 3dgpu or NVnews (can't remember)uses this little baby to cool his h20 setup. And he said it works killer. If I wasn't such a Vapochill slave this is what I would be using.

Item Name SubTotal
1 Ice Probe with Power Supply 104.95
Subtotal: 104.95
Shipping: 10.00
Tax: 7.61
Total: 122.56

Thank you for your order. Please print this page for your records. You will receive an email from UPS, FedEx, DHL, or Airborne once your order ships out.



Lets see what this thing can do
thanks for the advice krag

 

[raskren]

Yes, let us know how it works out. I didn't think that little popsicle would cost more then 10 bills though.

 

[Nico3k]

will do!

i found what krag was talking about:

I have been using a home made water reservoir with a built in thermoelectric water cooling device called the IceProbe. This gives me a reservoir that drops the temperature of my coolant to 15c without the use of a radiator, for a more compact system.

from http://www.hardocp.com/newsarchives.html?news=TWFyY2gsMjAwMywzLDIw

i cant wait to try it out... i will still be using both my radiator and the iceprobe

 

[camel]

I hope its helpful, but I cant believe the claim of 15c from [H] that cant be right. I am very interested to see what it does do If it made a difference of 5c, thats not too bad for the money.

Another seller
http://www.drsfostersmith.com/product/prod_Display.cfm?siteid=6&pCatId=4902