Zippyc's Water Chiller Project:

[zippyc]

OK. It is beginning....

So I have done the ambient air/water cooling thing with excellent results, with a parallel setup cooling waterblock on CPU, GPU, Northbridge, and a custom water cooling loop for my video memory.

Of course this is not enough for me, so I am going down the road of chilling my water.

I have a techcool external all-in-one heat exchanger, which I will turn into my warm water loop. Its 300GPH output will be split into 2 loops going through the outer two blocks of the chiller sandwich, then recombined and retuned to the techcool for cooling.

I have three aluminum (anodized) blocks with side protruding 1/2 ID outlets that measure 4"x2" x 7/8", which will make the sandwich. I bought them on closeout almost a year ago...I believe they were originally designed to cool pentiums.

I will start with an 80W pelt on the mustard side of the cold block sandwich, which will run whenever the computer has power. I also have a larger 169W for the mayo side which will be temperature controlled by a device I picked up from a friend called a variable PC (variablepc.com), which will allow me to set the water temp. to a constant target temperature (around 10 degrees below ambient to avoid condensation)…and this way only the 80W will run at idle.

I plan to assemble this whole mess in an old mini-tower Dell, in which I installed an antec 510 Power supply (which is rated for 29 Amps).

And, to drive the water through my custom copper 4-way manifolds (*which i will insulate), I chose the Iwaki 30RLT (Mag. drive), rated for around 500-600 GPH with a 17 FOOT head! It sits nicely on a custom vibration-isolation mount that covers the floor of the PC case.

And I got one of those little $25.00 3.5 Drive bay fan controller/dual temp readouts to go in the floppy drive bay so I can monitor hot and cold block temps with alarms.

So, suggestions before I build the sucker? I already anticipate adding more pelts to the sandwich as needed (it will support four total), and I believe I will need more when I get my MCW-50 for my GPU. Of course I’ll need another power supply to add them, but I believe in principle this system should work great!

zippyc

[]JR[]

169Watt tec will never bring the water temp down to 10degrees below ambient with a cpu attached in the loop. To make an effective solid state water chiller your talking 3x220watt tecs (minimum) chilling the water and 1x220watt tec on the cpu. Then a small powerstation and cooling tower to cool the hot loop down from the 1.5kW of heat being dumped into it, as well as the powerstation to power all the tecs.

]JR[

[zippyc]

Actually, to simply chill my existing water to 10 degrees below ambient the two tec's do work, as it works sucessfully for this guy with a similar rig (who I know posts accurate, real-life results).

http://westech.home.mindspring.com/chiller/mychiller.htm

So I believe two will work as long as I don't use an additional TEC on the CPU or GPU.

It is when I pick up my GPU pelt that I anticipate I'll be adding at least one more (maybe two) pelts to the mix. Do you think (in theory) that a single 224W pelt will cool enough to keep the hot side of an 80 W block assby. cool?

The Techcool unit has a radiator rated to dissapate 2.2Kw, so I am wondering how well it will cope with the hot loop. Only time will tell.

I hope to have it up by the end of the weekend, and I will be happy to share my results.

[]JR[]

Ahh just chilly water is your goal! Okay you should be alright, just about.

I always assume 100watts for a cpu and 60 for a gpu, and maybe 25 for a nb = 185watts of heat, add to that your pump heat, and losses youll just be outside the max heat pumped so you might need a little more...

An interesting idea none the less :)

Edit: stacking a 220 and an 80, assuming the 80 is going to pump 50watts of heat your talking 80*1.6+50 = 178watts, so yes it will handle it.
I dont have the 220watt performance graph on me, but i should imagine with 178watt heat load the dT will be 15-18degrees. So... with no heat load on the 80watt tec and 20degree water the coldside of the 80watt theoretically should be -55degrees, but realistically it will be ~-30-35degrees.
Dont forget that the quoted pumped heat on tecs is the number of watts that the tec has a dT of zero, i.e. being useless.

]JR[

[hedge]

But this idea will works for just running chilled water orver the cpu, right? I may have to give this a shot.

hedge

[zippyc]

Actually it will run chilled water over my CPU, GPU, and Northbridge. There will be some performance gains to be had at the CPU and GPU(not much, but some) and an increase in system stability.

And the beauty of this is I can add more pelts and power supplies as needed.

My end goal is to run the system coolant at 10C less than ambient 24x7 (Zero condensation worries that way.)

Then, when I add a GPU or CPU peltblock (assuming I have the extra headroom after adding at least one more pelt to the sandwich), if I can keep it at 10C below ambient, I should see near or below freezing temps on the GPU. Same applies to the GPU if and when I decide to add it.

I will also insulate so I can crank the coolant temps down as far as possible, which will lower the pelt temps, which should allow even better overclocks.

And if I go really insane I could even use a superchilled vat of coolant for the hot loop (temporarily, of couse) to run by the CPU and GPU pelts....I think that is how Chong 345 gets his insane overclocks.

At least that's the theory!

[zippyc]

OK. So I have the chiller running already, so some progress has been made.

Biggest suprise so far was that I could not use AS III on the pelts. So I picked up some silicone-based heatsink compound and used it in the sandwich.

I had an 80w and 170 wat pelt initially. I hooked it up without the temp controller initially to see what results I had.

I ran Madonion 2001 and with the pelts cranking at full speed I reached a cold water temp high of 25.1C. My hot water loop got up to 49C. Cold water ranges from 18C to 25C during tests.

Not bad, considering all the heat I was adding to the system with 1.8V on the CPU, 1.74V on the GPU, and 3.14V on the memory. I was also running 1.94V on the VDD. Considering the heat from 250 watts of pelts, and one can understand why I'm not getting colder results.

So, if I wanted to I could leave it like this and enjoy water temps that never get above ambient, but of course this is not xtreme enough for me.

I want to add more pelts, but know I will need more cooling on the radiator side before I do. (Time to build a multiple radiator exchange.)

49C is acceptable , but If I could get the hot loop to run at 35C, I will enjoy almost a liner improvement with all block temps.

My CPU used to get up to 45C, now it hits a top temp of 41C.

As I type this my water temp is sitting at 21.2C with my CPU temp at 30.5C at 1.8v, 2343Mhz.

GPU is at 21.8C and memory is at 23C running at 360Mhz, 3.14V.

These are all "Idle " temps. I will post more results my next major revision (next week when I get another pelt and power supply.)

Zippyc

[]JR[]

21.2 degree water is hardly subambient, and its deffinitely not 10 degree sub ambient.

The cpu temp isnt that great either.

However considering a 49degree, hot loop temp, its what i would expect.

All considered my straight watercooling setup gets similar temps, although my rad can dissapate 800-1000watts fairly easily but its generally got 220watt cpu tec and a 172watt gpu tec in it.

Loose 15degrees off the hotwater temps, loose 10degrees off the cold water temps would be my guess...

]JR[

[zippyc]

21.3C is definately lower than my ambient temps, which hover around 25.5C. And it is a full 10 degree improvement from my previous liquid temps.

I, like you, am also convinced better heat disapation on the hot side will yield better cold water results right in line with your estimates.

Do you have a recommendation for a good heater core (vehicle make, year, and model plz.)? I'll use two or four....

[zippyc]

Update:

Ok,

An 80W and 170W pelt brought my hot loop temps up enough to necessitate more heat exchange if I wanted to hit my goal of a constant 10C under full stress. The Techcool cooler did an admirable job, as was designed to take up a small space and has only one core in it. But 200+ watts of pelts added more heat then my 1,2,3,4….no, SEVEN waterblocks that I am already cooling with it, well, 49C is too hot for MY hotside!

So I purchased four biggish heater cores from Van's Autoparts that measure 5.5 x10.5 x 2 inches. (GM9078 Heater Core)

I have successfully built one of two (planned) two-stage cooling “CoreDucts", which will run me about $85.00 ea.

(The heatercores were $25.00 each, and I spent $35.00 on the two fans for each box and additional building materials. Each CoreDuct contains two cores.)

Anyhow, imagine an almost foot-long rectangular core with two 120MM 112 CFM fans appx. 4 inches above it (pointing up, oriented in a rectangular position centered above the core), sandwiched by an identical core 4 inches above it.

Suspend all above the ground by four inches to allow air to draw from the bottom. Hot H20 comes in the top core (cools) transfers internally to the bottom core (cools more and with room temp air drawn four inches off the floor….Second stage =COLDER!)

Blows hot air up, simple.

Simply plug it into the wall (if you use 110V AC fans) and hook it in line with whatever radiator/pump combo you already have (if it is 300GPH or more), stick it in the other room, and voila!

I hooked it up inline with my Techcool on about 12 feet of ½ ID hose, and, did you catch that…I stuck it in the guestroom!

BETTER TEMPS! Cooler office!

Hot loop runs 34C maximum now in a room that is averaging 24-25c. It was hitting 49C. Cold loop gets to 8C, up to 21C under full load. I have another 224 Watt pelt on order and 10C constant should be no problem real soon....

Here is a pic. of the unfinished product. I am debating whether I want to stain or paint it…the wood came out better than I expected using just a jigsaw and drill (all the power tools I had on hand that day.)

I have already tested a 110v fan speed control switch with it which worked great, and will add that and slave switching soon. 224CFM's sounds like a Harrier (well, not THAT loud and it is in the other room, who cares really?)

And my wife hated the tubes on the floor under the door so suggested I run them through the wall to the other room. Now THAT’s support fellas! (And she shares my office from time to time and I suspect she felt the heat…)

I have pics of the guts too if y'all wan to see them.

[zippyc]

I made shelve "sliders" so that I could use the same dimensions for each radiator and fan frame, which are removable and interchangeable. The sliders are also exactly level all around so they make an airtight seal, forcing all air to flow through the radiators (making the box the shroud.)

The side panel that is off is designed to provide access if I want to change fans or configuration.

The switch shown is the fan speed control I will add when I finish it up. I will also add some 1/2 inch fiberglass board to the reflective interior surfaces to cut down on noise and occupy dead space.

[]JR[]

:) now were talking :banana:

]JR[

[zippyc]

So with two cans of spray paint, my ugly dirty old Dell case looks better now as my chiller/pump housing....:)

[zippyc]

And the guest room now has it's own computer-powered heater for those chilly nites!

[Dr. Possum]

Damn that's coooool ;) :D :toast:

So how many TEC's and powersupplies do you have running total? I'd hate to pay your electricity bill :-\




~DP

[zippyc]

Right now I have three pelts (226, 170, 80) and 2 power supplies.
29A powers the smaller two, and 20A to the bigger one.

"Bad" news is I have a fourth pelt (226W) on order and don't have the third supply (yet).

But, I think the groovy factor and good performance is worth some electricity, and these are switching power supplies so it could be worse....;)

I also have a couple of dangerden copper blocks or order, I'm going to get that aluminum block away from my cold loop as the anodizing was not good and I began to get grey water from the reaction. I put some water wetter in with new water and that fixed it for a while...

It never ends, where would the fun be at the end?

[zippyc]

I found an el-cheepo 400W computer power supply from Powermax that claims 25A on the 12V rail. Theorretically one would be sufficient for powering a single 226W pelt.

The good news is that I found them for $12.00 ea. here....

http://www.dealsonic.com/pow40sinfana.html


Even if I don't get 25A from them, I could run them in parallel (after a chassis isolation mod on one) to make sure I'm getting enough power.

I had purchased a 25A 12V power supply for $59.00 from radio shack (discontinued and on sale for $59.00 from $119.00), but that sucker is going back and I’ll get four of the “el-cheepo’s” for the same money, and I’ll make a power supply tower for them which should push plenty of air through them.

100A of power for $50.00. Wow.

[TheDude]

Ingenious setup! :D

:toast:

[gouda96]

once i get my pelt setup done i am going to have to look into water chillers, they are so awesome!

[zippyc]

I finished the second radiator box today and am happy to report that my hot loop temps look like they dropped four degreees, and my maximim load cold loop temp has dropped to around 18C.

My maximim CPU temp read today on MBM was 66F during very heavy use and with high voltages! (18.8C)

For reference my hot loop now uses three MJ1200's in the series that goes like this:

5 Gal Open Res. (Ice Capable!) MJ1200 Pump# 1 to top rad> bottom rad (both in Heat Exchange #1) return inlet to pump 2 (underwater in res. to cool and in case of leaks, but still pretty tight seal) pumped back out of res. to top rad > bottom rad (H.E. #2) then THROUGH WALL to y and hot block's A and B, reverse y, return back through wall to res and suctioned by MJ1200 #3 into res.

I have two dangerden blocks on order and two extra geminicool spiral (2" x 2")blocks on my shelf (and I have another 226W pelt and four 25A 12V power supplies on the way....), so I'm sure I'll be busy this weekend. (I'm thinking... four pelted cold blocks in parallel, combined into one high flow for the CPU/GPU ...hmmmmm)

I guess I have gone a little insane, but it sure is fun!

[zippyc]

Update:

I ran four blocks in parallel with excellent results. The water velocity was still pretty high through the main line to the whitewater, and with 800w watts of pelts I had much faster cool down and recovery times (down to 2C in eight min). 12C under full load was as hot as it ever got.

Insulating and hose routing was a mess, but even the seven block exchange worked great (not having any better solutions to choose from).

Then I recently came across a guy who was selling his custom peltier compression blocks (a pair), and after some haggling he was nice enough to work with me and build a massive third chiller block for me to match the other two he was selling. Long story short, we had a new block built for me in a week that matches my high-flow requirement and fits the other two blocks perfectly.

My previous three block sandwich was highly efficient but had a flow bottleneck in the center block, and the blocks I had ware made out of poorly anodized aluminum and began to react in my water.


No problems like that with these babys!!! ). These suckers are made of solid copper (they weigh almost 30Lbs together). They include eight custom bolts that hold them together after applying clamping pressure, although I will probably build three more permanent larger external clamps so I can apply more massive clamping pressure (they have been used/tested at 300Lbs external clamping force). The center block will be covered in pelts on both large mating sides, and up to six full-size (50MM) pelts will fit this set.

The center block is thicker than the other two (more volume/better flow).

I polished 'em and as they will be stuck in a case out of sight soon, this is their moment to SHINE! I put them on a beach towel for color and pattern variety, and you can actually see the reflection of the reflection of the label on back of the tape measure, as well as easily read the single reflection of the inches and increments. They look like little gold bars on the floor, ‘cept they are copper. I like ‘em anyway! ;)

[saaya]

i think youll get better water temps when you sandwich the fans and the rads directly onto each other. if you leave some room ebtween the rads and the fans like its now youll get turbulences that slow down the airflow quite a bit.

get one for each room and you dont need to worry about cold winter days anymore :D

[wymjym]

By allowing that extra 'space' he has built a non restrictive shroud. Playing with some diverters might make a difference but one would need to measure the flow at different points within the tower'. Higher flowing fans would help (at the cost of more noise). Putting a fan directly on a radiator is not as effective. It only pulls (or pushes) air thought that particular area.
wj

also
I am really glad that I have dual monitors, how else can I see these rather large photos that ZippyC shares with us?

http://westech.home.mindspring.com/chiller/bigpicture.jpg

[zippyc]

If I put these two 120MM fans right up against the heater core, there would be about 40 pct. of the core surface area that would not get air circulation, not a good thing.

If the fans covered the entire surface area of the core, the design may have been different.

As it is with a tight seal at each "shelf" of my design, 100 Pct of the air that blows through the fans has passed but the first heater core already, and 100 Pct of the air will pass through the top core on the way out, and all fins of both cores have fairly even airflow for best efficiency. Thanks for the suggestion though.

Wynjym, you don’t like the widescreen format? If the site supported it I’d put together some surround sound music and sound effects (rushing water maybe?) so you could get the full home theatre experience (lol).

My pictures always appear larger when I post than when I edit them (sorry). At least I got rid of the ugly blank white borders around this one!

[greenBean]

Has anyone tried or even considered using aquarium water chillers? some models have their own temp control and they can be easily hooked up to your water cooling rig. they can run 24/7 without any problems and their quiet too. i was considering buying one of these things to use in my aquarium, then i had an idea of using one for a water cooling rig. i'm still searching for a cheap used one so i can try it.

an alternative would be those drinking fountains or those hot & cold water dispensers. although this would take a lot of work but it's cheaper. u also have to figure out how to control temp coz most of this things don't have any built in temp control.

[eweast]

Originally posted by zippyc
OK. It is beginning....

So I have done the ambient air/water cooling thing with excellent results, with a parallel setup cooling waterblock on CPU, GPU, Northbridge, and a custom water cooling loop for my video memory.

Of course this is not enough for me, so I am going down the road of chilling my water.

zippyc

I wouldn't suggest using TECs for water chillers. They don't work well.

[(TECs cooling system) - (CPU+GPU+NB+Other+any TECs)] / [4.18*(ml of water)] = number of degrees (C) per second * 60 = “” minute

This doesn't include any inefficiency.


You can get better results using phase-change and use less electricity.

[zippyc]

Phase change is way more efficient for the electricity $$$ for sure. I have put this and all my watercooling projects on hold for a while as I have begun a new career and bought a "New to us" house that is consuming all my free time for the next few months.

But, I think a combination of phase change AND pelt cooling might be the way to go for me. And that will probably consist of a big pelt block on the cpu, a smaller one on the gpu and northbridge, and phase change superchilled fluid out of a big freezer chest/holding tank.


But I do not want to design any system that has a compressor running in the same room as my computer as silence is also a goal for me.

[wymjym]

Originally posted by zippyc
is consuming all my free time for the next few months.



too bad

wj
http://westech.home.mindspring.com/pstuff/notime.jpg

[zippyc]

That river looks like a nice watercooling system to me!

Ah, I long for those idle days again...

Thanks for rubbin it in wymjim! :p:

[wymjym]

it also has really good flow numbers!!
wj

[eweast]

Originally posted by zippyc
Phase change is way more efficient for the electricity $$$ for sure. I have put this and all my watercooling projects on hold for a while as I have begun a new career and bought a "New to us" house that is consuming all my free time for the next few months.

But, I think a combination of phase change AND pelt cooling might be the way to go for me. And that will probably consist of a big pelt block on the cpu, a smaller one on the gpu and northbridge, and phase change superchilled fluid out of a big freezer chest/holding tank.


But I do not want to design any system that has a compressor running in the same room as my computer as silence is also a goal for me.

Afraid you can't go overboard with a compressor and TEC. Solid state devices, like TECs, stop working correctly at about -40C. Since you heat load is huge, you may be able to use a compressor to cool your water and just remain well above the -40C mark. In fact, that wouldn't be a bad idea seeing as how TEC water chillers are pretty awful efficiency wise.

For noise, I suggest throwing the compressor in a box of some kind and adding some noise insulation, like they do for sound rooms.

BTW, if you're married, which I assume you are... how does you wife put up with you attaching pumps and hoses to your computer?! And where do I find women like that?

[zippyc]

There was a guy I saw in the forum that combined cold water from a freezer at like -15C, and used it to cool the hotsides of pelts as I described above. Worked pretty well for him, so I thought I'd give it a shot.

I met the wife at a college social over the pool table while I was going to engineering school at The University of Texas. When I saw she was pretty AND could play pool, I became interested. Then I found out how damm smart she was. She IS very understanding...let me drill holes in the apartment walls to put my radiators outside my office on my last water rig.

So...maybe go to University of Texas or Sixth Street in Austin, you'll run across many pretty and smart ones there. Thats if you're not like 60 YO, lol.

[Jabo]

eweast, where did you get this data that TECs stop working correctly at -40C?

[eweast]

Originally posted by Jabo
eweast, where did you get this data that TECs stop working correctly at -40C?

A long time ago when I wrote my TEC guide (http://www.xtremesystems.org/forums/showthread.php?s=&threadid=38367). I've heard others quote the exact same figure and over at EOCF a guy named Chong noticed that when his coolant temps fell below -40C, his CPU temps started to go up.

The TEC doesn't stop functioning all together, it just slowly gets warmer as you go below -40C.

Also, it's kinda hard to get coolant temps cooling a TEC around -40C. Something like what zippyc is trying, -15C, is more common and very doable.

------------


Originally posted by zippyc
So...maybe go to University of Texas or Sixth Street in Austin, you'll run across many pretty and smart ones there. Thats if you're not like 60 YO, lol. [/B]

Ahh... I can't go to UT Austin, they don't have a great Neuroscience program, unless you're a grad student. Till then I'm stuck at UTD. :( Well... that and I can't stand Austin's traffic. :cussing:

[Jabo]

K eweast, I can tell you that it is the truth.
It is difficult to get any empirical data to back it up but supported by Kryotherms simulation software 172 Drift-08 module stops working at -110C coolants temp- 'cold' side starts to show hihger temps than 'hot' side.
For perfect insulation the above mentioned modules sweet spot is between -40 and -20C (-35C is the best imho).

I cannot comment of Chongs results since I have no details on it (linky would be good) :)

I will give your guide a go when time allows