TEC/Peltier CPU Chilled Water Cooling

[tech-wreck]

great thread ryan

...and then i remembered this: http://www.tomshardware.co.uk/answers/id-2630545/closed-loop-cooling-system-modification.html which is still sat in a box in the shed...
so now i'm inspired to add another project to the ever growing list of things to finish...

there is a radiator attached to that system, and the peltier has a heatsink on both sides. dell must have had a reason for designing it that way.

when i get round to it, i plan to replace the tubing and test it as designed. then i will rip it apart and start again, but following your design, and compare results.
one question though, will increasing the size of the reservoir have any effect, good or bad?

 

[4Ryan6]

one question though, will increasing the size of the reservoir have any effect, good or bad?

The larger the reservoirs coolant volume, the longer it will take for the peltier to drop the coolant temperature to a point of equilibrium or usable below ambient cooling level.

You'll have to experiment to see how much coolant volume vs the cold output of the peltier or peltiers you run.

 

[tech-wreck]

okay, i have much to read, and then i can start breaking stuff
i found the thing and grabbed a picture.

to the left out of shot is a 120mm rad, which i don't think i need.
the cpu plate is still wrapped in cardboard, the pump is a delphi?
the TEC bit is strange. heatsink-peltier-waterblock (the blue section)-another heatsink?
all i need now, is time.

 

[4Ryan6]

It looks like a 2 peltier water block sandwich?

Looks like: air heat sink > peltier > water block > peltier > air heat sink?

Is that correct?

Interesting design, how well did it work?

 

[4Ryan6]

Would you post a picture of the whole assembly including the radiator you mentioned?

Did it have a reservoir?

 

[4Ryan6]

After doing some online research on that assembly it seems it was used in some Dell large server towers.

Including the radiator would ensure the peltiers could not drop cold levels down to condensation areas, as the radiator has an ambient canceling effect on the peltiers chilling capabilities, so it probably would run as close to actual ambient as possible and maybe a hair below.

IMO, since this is the first time I've ever seen one of those, so I am just guessing at this point?

 

[tech-wreck]

just had a closer look, and the wiring goes to both sides of the waterblock, so yes, it is a double peltier sandwich, yum.
there is no reservoir in the loop, for phase 2 i'll be making my own or possibly an insulated cylinder like the one attached to the solar panels on my roof...i don't know yet.
full pic:

 

[4Ryan6]

Thanks for the picture!

That is a very interesting design, I wonder how well it actually performed?

 

[Deepli]

Hi, could you please explain why transfering heat from peltiers hot side with watercooling is not possible. It is a bit confusing because it is stated that temperature difference between peltiers hot and cold side is around 40-60c. If cold side is down to 0degrees celsius the boiling issue shouldnt be a problem or is it?

 

[4Ryan6]

Hi, could you please explain why transfering heat from peltiers hot side with watercooling is not possible. It is a bit confusing because it is stated that temperature difference between peltiers hot and cold side is around 40-60c. If cold side is down to 0degrees celsius the boiling issue shouldnt be a problem or is it?

Welcome to Toms hardware Deepli, So you joined Toms just to ask me this question, that's kinda suspicious wouldn't you say, considering what happened a few days ago in this same thread, never the less I'll humor you.

I never said transferring heat from peltiers hot side with water cooling was not possible.

I chose not to go that route because to do it properly cost way more than I wanted to invest, this cooling is expensive enough as it is.

You have to have enough radiator cooling field to cool the heat load from a combined 600w generated from the peltiers hot side with water. (Running the same spec'd peltiers I am running that is, there are other alternatives this is only one solution.)

But why invest that kind of money in an entire independent water cooling loop, requiring special water cooling blocks that can handle the peltier generated heat load, when it can be done with $120.00 of heat pipe air coolers.

http://www.performance-pcs.com/thermalright-venomous-sb-e-universal-cpu-cooler.html

Just one high quality 120 x 480 radiator cost more than that, excluding the rest of the loop, pump, reservoir, special water cooling blocks, etc.

http://www.performance-pcs.com/alphacool-nexxxos-monsta-480-radiator.html

If you want to cool the hot side with water cooling that's your prerogative, Go for It!

 

[4Ryan6]

@ Deepli

Cooling the hot side with air cooling and changing to the 110cfm 120mmm cooling fans is:

3 heat pipe air coolers at 36.95 each and 3 110cfm cooling fans at $9.99 each totals $140.82 plus S&H.

Cooling the hot side with water cooling:

Got to have a pump, I would use no less than a D5 variable speed Swiftech MCP655 = $76.95

Water blocks that would fit a 50mm x 50mm peltier which is 2" x 2", using copper would keep the loop all copper thus avoiding mixed metal problems. 3 copper water blocks 3" x 3" PT # WBA-3.0-0.85-CU-01 rated 1200w and supplies full coverage of the 2" x 2" peltier at $164.50 each totals $493.50.

A cheaper alternative is the Aluminum blocks WBA-3.0-0.85-AL-01 at $83.25 each, totaling $249.75, if you use a coolant additive for mixed metals in the loop, I prefer 100% pure steam distilled water and BioCide.

You need some type of reservoir? (That really depends on what type of coolant you use)

Even if you go the Aluminum water block route that's more money invested right there, than the entire cost of air cooling the hot side!

Monetary water cooling investment vs air cooling cost my point is already made, but now we come to the radiators to handle the total heat load?

I run 3 peltiers that pull 20a each and use 200w each actual, 2 are powered when the machine is booted, the 3rd is switch controlled when needed, so the radiator cooling field has got to be able to handle all 3 powered up.

Any suggestions?

 

[Deepli]

Firstly thanks for these informative answers, really helpfull.
Heatpipes are too firm and it will take time to reshape them and redesigning has its limits.(reason im trying to avoid them)
Therefore I've heard a little about super expensive flexible heatpipes, those look much more attractive.

One more, hopefully not too stupid question. Theoretically if peltiers cold side is directly on cpu and I cool down peltier module to 0degrees celsius, then the hot side should be around ~40degrees, does not seem like a impossible amout of heat to transfer away. In that case, how much will cpu temp differ from 0c.

 

[Deepli]

As im really a beginner in the subject, i do not have any suggestions. Just found myself to be interested in the world of peltier cooling

 

[4Ryan6]

Firstly thanks for these informative answers, really helpfull.
Heatpipes are too firm and it will take time to reshape them and redesigning has its limits.(reason im trying to avoid them)
Therefore I've heard a little about super expensive flexible heatpipes, those look much more attractive.

One more, hopefully not too stupid question. Theoretically if peltiers cold side is directly on cpu and I cool down peltier module to 0degrees celsius, then the hot side should be around ~40degrees, does not seem like a impossible amout of heat to transfer away. In that case, how much will cpu temp differ from 0c.

Sounds like you may be more interested in direct mount applications and many at Overclock.net do that type of peltier cooling.

http://www.overclock.net/f/62/peltiers-tec

 

[Ruwed]

Hey Ryan,

Sorry about the previous debate on your thread. I didn't realize it would bother you, I'll leave the proof of concept and explaining facts in your thread to you from now on

I had a couple of questions about your setup if you have a bit.

Roughly how long does it take to cool your reservoir? Do you have a startup procedure? As in, turn on cooling system for X amount of time prior to using the machine.

Have you timed the cool down time for just the loop with no CPU powered on vs with the CPU turned on?

How much heat do you feel coming off of your heat sinks when you are running full blast? (Or do you have a temp gun that you have taken some measurements with?)

My build is almost complete, a couple of minor setbacks, one completely preventable with a bit more research... the other I think just Murphy coming to visit, or I did something that I'm not aware of.

I should be up and running tonight and getting my own results, I was just curious as to what I'll see vs what you see. I am running one 230w Qmax and two 130w Qmax peltiers, so I'm a bit under sized compared to your setup and may have to step up. Our res size is pretty similiar, I think my res holds just over a half gallon, the full loop takes just under 2/3 of a gallon.

 

[4Ryan6]

Hey Ryan,

Sorry about the previous debate on your thread. I didn't realize it would bother you, I'll leave the proof of concept and explaining facts in your thread to you from now on

I had a couple of questions about your setup if you have a bit.

Roughly how long does it take to cool your reservoir? Do you have a startup procedure? As in, turn on cooling system for X amount of time prior to using the machine.

Have you timed the cool down time for just the loop with no CPU powered on vs with the CPU turned on?

How much heat do you feel coming off of your heat sinks when you are running full blast? (Or do you have a temp gun that you have taken some measurements with?)

My build is almost complete, a couple of minor setbacks, one completely preventable with a bit more research... the other I think just Murphy coming to visit, or I did something that I'm not aware of.

I should be up and running tonight and getting my own results, I was just curious as to what I'll see vs what you see. I am running one 230w Qmax and two 130w Qmax peltiers, so I'm a bit under sized compared to your setup and may have to step up. Our res size is pretty similiar, I think my res holds just over a half gallon, the full loop takes just under 2/3 of a gallon.

Your debate has not bothered me at all, it was the jerk that opened an account at Toms and his first post was to attack this thread, the moderators got him, that kind of attitude and nasty language is not tolerated here, he was banned and everything he said deleted.

Roughly how long does it take to cool your reservoir?

Normal start up after being off all night the reservoir has warmed up to ambient room temperature after powering up it takes about 30 minutes to reach 10c coolant temperature.

If I'm in a hurry to drop the temperature I can energize the 3rd TEC assembly and drop it in about 18 ~ 20 minutes, with the machine just idling.

Do you have a startup procedure? As in, turn on cooling system for X amount of time prior to using the machine.

I do out of habit, not operating requirement!

I get up every morning at 4am to help my wife get up, and get ready for work, I usually go in my office and power up the machine, then go feed and water the cats, fix coffee for she and I, and do anything else she needs me to do to help her.

Then once all that is done I get to use my computer and it is at gaming temperatures as far as the coolant temperature is concerned, it is a perfect time situation to let it chill, because those things have to be done.

I don't always do it that way sometimes I power it up and do what I need to do, because once it is powered it drops below ambient very fast.

Have you timed the cool down time for just the loop with no CPU powered on vs with the CPU turned on?

No? When I power up everything starts.

How much heat do you feel coming off of your heat sinks when you are running full blast? (Or do you have a temp gun that you have taken some measurements with?)

Right now my ambient is 23.4c the peltier exhaust heat is 36c that's taken with a temp gun.

My build is almost complete, a couple of minor setbacks, one completely preventable with a bit more research... the other I think just Murphy coming to visit, or I did something that I'm not aware of.

I should be up and running tonight and getting my own results, I was just curious as to what I'll see vs what you see. I am running one 230w Qmax and two 130w Qmax peltiers, so I'm a bit under sized compared to your setup and may have to step up. Our res size is pretty similiar, I think my res holds just over a half gallon, the full loop takes just under 2/3 of a gallon.

It will be interesting to see how well it does, did you ever get your voltage controller worked out?

 

[Ruwed]

I did get that voltage controller working! So far it looks pretty good, I'm sure there are a few bugs that I've missed and will find eventually, but as of right now it puts out a nice smooth max of 11.8 out with 12v in, which is pretty good efficiency for a home built buck regulator! I can turn it down as low as 2v that I've verified, possibly lower, I haven't tried.

I've gone through some other headaches figuring out how to take all of the readings I want but we are in the home stretch now.

I was only asking about running the cooler without the computer on as it would cool much faster that way. I wired up a 5v relay for the power enable line on the cooler and then ran a molex plug up to the computer, that way one button turns everything on. I was just considering adding a bypass power switch on the cooler to be able to turn it on independently if I want without the computer running. I'm sort of toying with the idea of building a real time clock for the Arduino and having it turn on the cooler before I get home from work, that way it's all ready to go when I get home. We'll have to see how that idea goes though. It might not be worth the headache.

I appreciate the info on the setup! It will give me a good comparison to see how my results are stacking up to an established cooler.

 

[Ruwed]

Quick update:

Get bigger TEC's. One 210w Qmax and two 130w Qmax TEC's just don't cut it!

I can drop the Res temp in about 30 minutes with no computer on. Once the computer is on the res temp starts slowly creeping up. I haven't found it's balance point yet, but I'm suspecting its around 18c° or so.

I have two TR Venomous coolers on the way along with two more 210W TEC's, we'll see how much of an improvement they make.

The voltage regulator worked, right up until I realized how hot the switching was making it. It was hot enough that once I took the lid off I could feel the heat waves rising 12".... whoops!
I bypassed the PWM and just tied them high for now, they don't generate any heat that way. I'll go through and remove them and add in a couple of control relays instead. I'm going to lose efficiency but I'm not sure I can fit large enough heat sinks inside my case to run those, especially once I add in two more 20 amp TEC's..... Besides, the heat generation alone from them will probably negate any benefit I'd see!

Just thought I'd share my learning experiences as I go! No regrets though, it's been a great learning experience.

 

[tech-wreck]

hi, i have an update: work has begun on a crate to hold it all in

 

[4Ryan6]

I can drop the Res temp in about 30 minutes with no computer on. Once the computer is on the res temp starts slowly creeping up. I haven't found it's balance point yet, but I'm suspecting its around 18c° or so.

I have two TR Venomous coolers on the way along with two more 210W TEC's, we'll see how much of an improvement they make.

What is your normal ambient room temperature?

Because 18c actual coolant temperature just from the machine running is going even higher under a constant load like gaming, and won't support much of an overclock.

What you have coming will make a big difference!

 

[4Ryan6]

hi, i have an update: work has begun on a crate to hold it all in

Am I seeing 2 cases mated together? Outstanding!

Details Please?

 

[tech-wreck]

i'm not releasing any more details yet, mostly because there aren't any. lol, i'm not even doing a build log, i'm just going to throw it together and unveil it when it's done.
but yes it is two cases in a state of permanent coitus

 

[rubix_1011]

I like that kind of DIY ingenuity.

 

[Ruwed]

I can drop the Res temp in about 30 minutes with no computer on. Once the computer is on the res temp starts slowly creeping up. I haven't found it's balance point yet, but I'm suspecting its around 18c° or so.

I have two TR Venomous coolers on the way along with two more 210W TEC's, we'll see how much of an improvement they make.

What is your normal ambient room temperature?

Because 18c actual coolant temperature just from the machine running is going even higher under a constant load like gaming, and won't support much of an overclock.

What you have coming will make a big difference!

My overclock is currently 4.4Ghz from 3.6 (5820K), no throttling, it's full 4.4 all the time. Ambient room temp was about 24c last night, the heat sink temperature was 36c, I think 38c at it's highest point.

Yeah, I could run my computer as it sits and the current build would keep it just a bit cooler than I had with watercooling, but they wouldn't make enough of a difference to justify running that big ole box.

So bigger TEC's and coolers it is

Thanks for posting up that link to the venomous coolers. The price really helped a ton. I wouldn't have been able to upgrade both TEC's if would have had to buy another Noctua.

Here's a quick pick of my setup in it's current phase, the box over the PSU will be going away and I'll be mounting relays for two of the TEC's instead.

 

[4Ryan6]

Thanks for posting up that link to the venomous coolers. The price really helped a ton. I wouldn't have been able to upgrade both TEC's if would have had to buy another Noctua.

U R Welcome!

That is an excellent price, and they won't be available forever either, I think once Performance-PCs sells off their on hand stock, they will disappear from the sales inventory, because I don't think Thermalright is making those coolers anymore?