TEC waterchiller, second attempt

[drifter]

in my first thread here i almost completed my project but i was not satisfied with the result : the hot side cooling was insufficient, the temps were not that great and the case was way to heavy to move around. so i decided to go back to the drawing board and came up with what seems to me a better plan. most of the parts i kept from my initial buid, but have improved the cooling, looks and mobility. (for parts list refer to my older thread if you're interrested)

So here it goes: i bought 2 mid tower identical cases (cooler master centurion 590), i picked this model because of its size, (u'll see later on), simplistic looks, and because 2 cases one next to the other would look like a single case.

The i dea is to have a tec chiller in one case and the pc in the other.

The chiller case got stripped of all the unnecessary drive bays, etc, and some additional holes were drilled.

The new cooling harware: 3 360x120 aluminum koolance radiator with 9 schyte ultra kaze 1000 fans. so the total cooling is 3 360x120 rads with fa,s pushing, 1 360x120 rad with fans pulling and 1 240x120 rad in pull. pump is still d5, and the tec heat exchanger is from the previous build.

top

front

top

back


The chiller is almost assembled, as soon as i am done leak testing, i will install the psu, front panel and that's basically it. This build was harder then my first one because i tried to squeze in evrything tightly so plumbing and wiring was insanely hard. I will update with results and any input is very welcome.

This unit is independent of the pc case (apart from a relay to have psus working in sync) and has quick disconnect valves for the cold side loop. so it is much easyer to move aroundthen my firs build, looks better, and working in the pc case became easyer (because i upgrade often it was one of my main concerns)

[systemviper]

WOW, can't wait to see some results, nice looking stuff!

[drifter]

Well the results will most likely disappoint (maybe a few C gain over my first build, but like i said, mobility and convenience were my main priorities, and i do not want to go under 15C in any case which i could do even with inferior cooling), but hey, it was fun nonetheless.

[luf1c3r]

It seems a good job, I'm also eager to see results of the invention

Greetings

[eligray]

this is epic in the fullest sense of the word...

[Uncle Jimbo]

The new cooling harware: 3 360x120 aluminum koolance radiator with 9 schyte ultra kaze 1000 fans. so the total cooling is 3 360x120 rads with fa,s pushing, 1 360x120 rad with fans pulling and 1 240x120 rad in pull. pump is still d5, and the tec heat exchanger is from the previous build.

Drifter -
I'd like to make a few comments on your cooler - which is really quite a good one. In your previous thread, some posters made assumptions which aren't right for this piece of gear.

Your exchanger has 8 DRIFT-0.6 TECs (appropriate for 'drifter' I guess ). Those are rated Qmax of 229W at 300K, which is 27C. In this design, Kryotherm is using them in series parallel, operating at about 50% Imax, as you assumed.

With your hot side loop at around 30C, and a desired dT of 20C (around 10C cold side) each of those TECs will be at an efficient operating point - CoP of between 1 and 1.3, depending on differential and heat load. Each TEC can move about 80W across a 20C dT. So 8 can move 640W.

Even with an OC'd i7 and a couple of big GPUs, you should have plenty of cooling, although perhaps not at your desired differential. For example, at 10C dT, they will move 120W each, for a total of 960W cooling.

I have measured an i7 965 puts out 180W at 3.6G, 240W at 4G. I have more experience with ATI GPUs than nVidia, but I would think 200W would be an upper bound per GPU. So even the 640W load should give you 20C dT.

If you are not getting that, perhaps when you took the unit apart, did you re-torque the assembly? The TECs need to have between 150 psi and 300 psi, evenly applied. You would tighten the assembly as you would a cylinder head - take them all just tight, then increase torque crosswise by a slight amount until you hit the right number.
It looks like you have 10 bolts. If they are 6/32 or 3mm, you need about 13 in/lb per bolt to get 250 psi when done. If they are 8/32 or 4mm, you need 16 in/lb.

That is a fair amount of torque and you may not have applied it evenly enough, or in the right sequence. If torque is low, you will see a big performance drop as the hot side will not transfer to the cooling loop efficiently. You can also damage the TEC.

If you are using a suitable controller (I assume temp control off the cold loop), you can also allow higher voltages to manage peak heat better. Even if you limit to 70% Imax, you can go to 32V. That will take cooling at 20C dT to 140W per module for total cooling of 1120W. You are at a less efficient operating point, but presumably you won't need that cooling all the time. You can go to 48V and still be in spec, but I would stay at or under 32V - you won't get much more performance going higher, and will burn a lot more juice.

If your heat load is really around 650W peak, that same 32V will deliver dT of 40C. So you would have the ability to control your load for any power level, at any temp you like down to -10C at the IHS. At Idle, your controller should ramp power down to 5-6V and 8A or so, not much power at all.

*EDIT* BTW you should be using at least 10 GA wire to go between your power supply and the cooling assembly. It's less important with a controller, but you don't want the wires to heat up.

[drifter]

So i pulled everything together and ran some tests. the results are rather weired. to remind you i'm running 8 226w 24v tecs in pairs at 24v (50% cause when you run in pairs V doubles). I tried running th tecs at 24-30V (12-15 per tec) and for some reason the cold side temp would stay the same but the hot side would vary by 4C (hot side 36-40, cold side 15) with the pc at idle. in my first attempt with inferior cooling i got better resuls on the cold side when i increased the voltage. And the hot side temp would not vary regardless of fan speeds.
I have a few theories what is wrong:
1. Maybe the aluminum heat exchanger due to it's size affects the performance by absorbing heat from the air (cold side)
2. maybe the thec is broken or a wire is cut inside the tec assembly, so the inactive tecs act as a heat exchanger between the hot and cold sides.
and 3. maybe the NB cooler which is attatched to the mb heat spreader acts absobs acts as a radiator (the mb head spreader is quiete cold so not only would it absorb heat from the NB, sb, vregs, but also exchange heat with air which theoretically could reduce the cooling efficiency as the surface is rather big, made of copper, has fins and airflow.)

My last theory looks more credible to me because i really struggle at 15C to lover the temps by 1C. so the heatspreader acts as a radiator and is counterproductive in an exponencial way, meaning that every C below ambient is harder to gain. (hope you get what i mean)

However the hot side cooling seems superior as the radiators are colder to toutch then they were in my firs attempt.

I REALLY do not want do disassemble everything to see what's wrong and might aswell keep it as it is, 15C water temps were my goal from the very beginning, and the unit seems to handle the load with a minor hit to the temps (~3C). fans are really quiete and the cooling capasity seem capable of handling the extra gpus.
I am sure something is not right, ant it will haunt me for the rest of my life be4 i find out for sure what, but i am too tired of playing with this particular unit so will leave it as is.

Uncle Jimbo, thank you for a very informative post (your post on this forum are a great contribution to tec community).
I have 2 similar units and the one i have disassembled is still opened up. The wire i'm using is 6ga i believe.

Some more info:
The hot side of the heat exchanger is around 38C, cold side 15C, and ambient 22, the tecs are running at 50% at 24v.
when i up the V to say 30 i get hot side temps up by 2-3C but the cold side temp does not improve.

Like i stated earlier in my post i think the problem is due to some heat exchange (other then cpu and chipset cooling) somewhere in the loop which acts counterproductively. Also i believe kryotherms heat exchanger has a few flaws, one is the material used, which is aluminum (should've been copper, but that would've been more expensive and 1.5 times heavier), 2 is the thermal compound which is ceramic based i guess from the looks of it but could've been better, and looked as if it was applied unevenly, also the surfase is not lapped properly. so i guess i could have improved the efficiency overall and gain a few C by correcting the said flaws.



As for some results, well nothing spectacular my 920 with 1.375V at load running prime 95 goes all the way up to 60C, but the waterblock surfaace temp is under 20C which leads me to believe that either my waterblock is not making good contact with the cpu or the temp sensor is wrong. I think so because i have another i7 rig with a different mobo (ga-ud5) and when i compared the 2 with stock cooling and volts and the other cpu was running much colder no matter ho many times i resitted the cooler.


A little update:

Here are a few pics:




I did not take any from behind as it is still a bit messy.

Also, just when i thought i was a "happy camper" i decided to double check everything, all the wires and tubing (the tubing on the cold side was bent really bad apparently), and on next startup i started seeing temps of 15C at load and 10C at idle (after 10 minutes, i did not test longer as the temps were still going down), also hot side got colder a bit. i am not sure what i did but now the temps are not as easily manageble as before so i might need a controller after all...

As for the cpu, well the temps went dow a few C at load, and i am sure a different cpu would provide different temps but the temps are not disturbingly high so i will leave it as is.

[zipdogso]

2. maybe the thec is broken or a wire is cut inside the tec assembly, so the inactive tecs act as a heat exchanger between the hot and cold sides.

A working TEC is a heat pump and non-working TEC is an insulator.

Also the way the unit is wired I think you would have more problems if one/some were not working.

I think it is probably down to construction and TIM.
-----------------------------------------------------

Glad to see your getting some likeable results after all the trouble you been through building this.

[Uncle Jimbo]

Drifter - where did you get the Kryotherm unit? How much did it cost you? I see it in the Kryotherm catalog but didn't see a price.

[drifter]

I got it from kryotherm russia for 600$, which is reasonable if you ask me because the asking price for their 226W 40x40 tec is over 60$ and this unit has 8ght of those. (it says on their site the unit has lower speced 80w modules but i popped it open and am 100 percent sure those are 226w modules, also they confirmed it on the phone).

a lil update:
I got my gtx280 in the loop. With the cpu running prime and gpu running furmark stability and xtreme burning mode i get water temps under 20 (ambient close to 30C). I softvoltmodded the gtx and upped the V to 1.27 and am running 783 core, 1566 shader.

[repsol_23]

Drifter - where did you get the Kryotherm unit? How much did it cost you? I see it in the Kryotherm catalog but didn't see a price.

It's actually closer to $75. If I remember correctly its $50 for the tec and another $24 to ship. I was looking into getting one last year but decided against it since I didn't think it would be enough to cool my old HD 3870.

[Uncle Jimbo]

I got it from kryotherm russia for 600$, which is reasonable if you ask me because the asking price for their 226W 40x40 tec is over 60$ and this unit has 8ght of those. (it says on their site the unit has lower speced 80w modules but i popped it open and am 100 percent sure those are 226w modules, also they confirmed it on the phone).

a lil update:
I got my gtx280 in the loop. With the cpu running prime and gpu running furmark stability and xtreme burning mode i get water temps under 20 (ambient close to 30C). I softvoltmodded the gtx and upped the V to 1.27 and am running 783 core, 1566 shader.

The cooling power and voltage / current also says those are the big DRIFT modules... that is not a bad price considering what's in it.

[repsol_23]

The cooling power and voltage / current also says those are the big DRIFT modules... that is not a bad price considering what's in it.

Not meaning to hijack this thread but I had a question for ya Uncle Jimbo.

Do you think that the Drift 06 would be sufficient to cool a stock HD 4870? I am not looking for crazy low temps but I would like to put a tec on it. The card is currently running ~35C (idle) and ~44C (gaming 2+ hours) and furmark/ATI tool testing ~52C. It would be in a loop (2 x 3.120 swiftech rads) with the 345W tec from arctic spider with the (water) loop running ~35C. The Drift would be running off of a Meanwell 12V supply.

Given my current temps do you think it would even be worth it?
Thanks in advance!!!

BY the way, the price I mentioned a few posts earlier was for Kryotherm USA, hence the reason for thinking they were too costly. I felt that to ship it over one state (in my case) that $25 was too much.

[Uncle Jimbo]

Not meaning to hijack this thread but I had a question for ya Uncle Jimbo.

Do you think that the Drift 06 would be sufficient to cool a stock HD 4870? I am not looking for crazy low temps but I would like to put a tec on it. The card is currently running ~35C (idle) and ~44C (gaming 2+ hours) and furmark/ATI tool testing ~52C. It would be in a loop (2 x 3.120 swiftech rads) with the 345W tec from arctic spider with the (water) loop running ~35C. The Drift would be running off of a Meanwell 12V supply.

Given my current temps do you think it would even be worth it?
Thanks in advance!!!

BY the way, the price I mentioned a few posts earlier was for Kryotherm USA, hence the reason for thinking they were too costly. I felt that to ship it over one state (in my case) that $25 was too much.

At 12V, the Drift TEC will move 130W across 10C and 105W across 20C, which should be plenty for a stock 4870. My guess is you will be around 15C at idle and maybe 25C at load with 35C water. You will draw less than 8A from the 12V line, so you could run it on a Molex from your power supply at that level. If you need more cooling, you can run the voltage up but would need a better connection for power (like spade lugs). The Meanwell will go to 13.8V - at that level, you move 150W across 10C, and 120W across 20C, drawing about 9A. That will probably take you to 10C Idle, 20C load.

[drifter]

another litle update.
I was checking my loop and found out that all this time i was running my horside loop pump at 60%, and after making it run at 100% (12V) i got a nice improvement in hot side temps. then after investigating further i realised that the radiator closest to the heat exchanger was much hotter then the furtherst down the loop, so i decided to install a second D5 pump to runi n parallel. and guess what, the temps improved some more. (i am not giving any numbers as for them to be comparable i would have to take measurements in identical conditions, so for the moment i will just say there was a definite imrovement but no numbers yet).

so more flow = better performance.

i would have to replace one of my pumps though cause it seems to make more noise then the other.

[vodka3dg]

seems you solved some kind of bottleneck-like issue with that second pump.. congrats! cant wait for some more results
have you figured out how much power the entire thing actually consumes? i'm actually curious about that :p

[drifter]

2 d5s improved my temps but i still feel there is more room for improvement.

The problem is i want a quiete system and the d5s run really quiete (and the noise from the fans at 1000 rpm is barely audible).
so my options are:
run more d5s in parallel (not so good cause as far as i understand the efficiency goes down a bit with every extra pump added)
or get a new pump (noise is the main concern here)

So i went a little pump crazy and bought 3 aquarium pumps (guys at the aquarium shop looked at me as if i was naked):
eheim 1262 - 3400 l/h 3.5m http://www.eheim.de/eheim/inhalte/in...ail_27591_ehen
iwaki md40 - about 3700 l/h 3.5m http://www.iwakiwalchem.com/Literatu.../WMDmanual.pdf
deltec hlp 4040 - 4500 l/h 3.5m http://www.deltecaquariumsolutions.com/delpumps_1.php
i did not get more powerful pumps for the same reason, to keep the noise down.

I tested the eheim and iwaki and contrary to what i thought iwaki was crazy loud, eheim also a bit loud but bareble (well loud to the barely inaudible d5s). Deltec still not tested but i got high hopes for thiss baby (i especially like the 3 speed selectable controller).
A pity i do not have any equipment to du a real review of the pumps and post it in the LC section, maybe i'll try to come up with something later on...

as for power draw, i do not have a kill-a-wat meter so can only mension the amp draw readings during my firs build attempt. i saw 20A draw at 24Von the psu (20*24=480w and to calculate the inefficiency 480/8*10=600w power draw). but with tecs the power draw varies depending on load and hot/cold side temp, so it is a rough number.