Post Rad Chiller Concept

[Xeon th MG Pony]

you want lots of low wattage peltiers, this spreads the load out and improves efficiency (As much as one can using a pelt any way) All so to keep Amperage reasonable wire them in such a way as to use 120V, and do a direct rectifier circuit so you don't have to mess around with low voltage high amperage ( this will net you a tad bit of efficiency as one less conversion loss ), Low wattage pelts should be cheaper too so make it with two hot sides with all the cold sides on your chiller block, that will all so give capacity control as you simply switch on the second bank if you need more cooling!

Pick up 2 snap disk thermostat set for 60C or 70C=off for both the hot side of the peltiers and the rectifier, this will give you some safety preferably manual reset and of course a fuse. All ways keep in mind with pelts ther is all ways a risk of thermal runaway and can resualt in a fire! All so make sure all wires you use have an insulation rating of or better then 90C

[Xeon th MG Pony]

rofl pete!

@ that voltage and those tec's my case would be an AC! thanks to the radiator.

Pelts are no where near that efficent, be more like an excellent blast furnace on the other side, which of course then the cold side starts warming, rins and repeat!

[NaeKuh]

Pelts are no where near that efficent, be more like an excellent blast furnace on the other side, which of course then the cold side starts warming, rins and repeat!

True... lol... the hotside wouldnt have enough cooling. But if it did...


Anyhow... i gave martin to okey to start. Unless someone has last min changes he should do on the design, i think its set.

Im going to go with 3 x 80W TEC's. nicely downvolted i hope to pull around 50W each for a total of 150W off my cpu loop. That would hopefully then mean my entire cpu loop system is only drawing ~80W of heat.

mmmm might be major overkill to keep my PA120.3 on that.

[Duh]

True... lol... the hotside wouldnt have enough cooling. But if it did...


Anyhow... i gave martin to okey to start. Unless someone has last min changes he should do on the design, i think its set.

Im going to go with 3 x 80W TEC's. nicely downvolted i hope to pull around 50W each for a total of 150W off my cpu loop. That would hopefully then mean my entire cpu loop system is only drawing ~80W of heat.

mmmm might be major overkill to keep my PA120.3 on that.

serve yourself a solar panel and ask the guys round here if your cooling system is efficient

[littleowl]

True... lol... the hotside wouldnt have enough cooling. But if it did...


Anyhow... i gave martin to okey to start. Unless someone has last min changes he should do on the design, i think its set.

Im going to go with 3 x 80W TEC's. nicely downvolted i hope to pull around 50W each for a total of 150W off my cpu loop. That would hopefully then mean my entire cpu loop system is only drawing ~80W of heat.

mmmm might be major overkill to keep my PA120.3 on that.

what changes if any have you decided to do on the block?
Is martin going to make the block big enough to take a 62mm tec if you ever decide that you want one on there?

[Martinm210]

what changes if any have you decided to do on the block?
Is martin going to make the block big enough to take a 62mm tec if you ever decide that you want one on there?

Yeah, the block will be big enough to take on 3 x 62mm TECs, so it'll be flexible for later changes if needed

[Maenor]

Looks awesome so far, I only have a couple questions about the design:

How thick is the base between the pelts and water (both sides)? IIRC, 3/8" is about optimal for a plain copper plate, so maybe have a thicker base for the cold side and obviously thinner on the hot side.

That brings me to my next question, the height/thickness/overall contact of the fins. The current design looks really free flowing, but I don't see enough of the cold dispersing through the loop. Machining and materials cost are a concern, but you might want to (somewhat) follow Swiftech's GTX and especially MCW6500 designs and reduce the height past the inlet/outlet, or angle the inlet down. Just a thought there from someone who's taken approximately 0% of the thermal/fluid dynamics classes you have.

And just to clarify for some other people, the purpose of this isn't to get subzero (even subambient) temps (at load at least ) but to effectively cut away some of the heat load on the actual CPU loop. Like Naekuh said, if he has 240 watts of heat, 180 watts of (theoretical) cooling from pelts would give him a CPU loop that only needs to dissipate 60 watts.

[littleowl]

Yeah, the block will be big enough to take on 3 x 62mm TECs, so it'll be flexible for later changes if needed

hey martin will you make the fins inside a little bit taller so the will get more contact to the water. I think that will help with heat transfer for the bigger wattage tecs. What do you think?

[Duh]

stupid question : I read carefully the 1st post and after some days I realized that obviously you should do a sandwich thing right??

sandwich: 2x devices made by martin :
1) to cope with the hot side
2) to get the cold that they deliver and to be able to get it into the loop.

In case of making it very long ( martin s device) : shouldn't it work in a bad way as the last tecs would be having water that had been already heated by the tecs before them?

[littleowl]

stupid question : I read carefully the 1st post and after some days I realized that obviously you should do a sandwich thing right??

sandwich: 2x devices made by martin :
1) to cope with the hot side
2) to get the cold that they deliver and to be able to get it into the loop.

In case of making it very long ( martin s device) : shouldn't it work in a bad way as the last tecs would be having water that had been already heated by the tecs before them?

that is no different the people using a maze 4-1 for tec cooling.

[NaeKuh]

stupid question : I read carefully the 1st post and after some days I realized that obviously you should do a sandwich thing right??

sandwich: 2x devices made by martin :
1) to cope with the hot side
2) to get the cold that they deliver and to be able to get it into the loop.

In case of making it very long ( martin s device) : shouldn't it work in a bad way as the last tecs would be having water that had been already heated by the tecs before them?

martin is building 2 blocks for me.

And they are getting sandwitched. The hotside will have a PA160 DDC-3.2 EK-Xres

The cold side will connect to my main cpu loop with a RD-30, PA120.3, EK Supreme, MCW30, Martin's heat exchanger, EK 250 res.

[Duh]

In case of making it very long ( martin s device) : shouldn't it work in a bad way as the last tecs would be having water that had been already heated by the tecs before them?

I forgot to refer this was to be considered to the hot side of HX ( heat exchanger)..

looks promising.. when will you have it done??

J

[systemviper]

Hi, I have been thinking of the same thing for a while, but since i don't have equipment, i have been looking for some kind of equipment to mod to do kinda the same thing.

With that said, wouldn't it be more efficient to run some kind of dual pass over the cold side of the loop, like a radiator flows and have as much "cooling water time" on the TEC's cold side as possible, why just run the water down the shoot once, loop it back and have both the input and output on one side, again just like a radiator. I think you will add a little more flow resistance to the loop but the gain from using the TEC's cooling more efficiently by giving the water more cold face time would be much more beneficial.

BUT on the hot side keep it just the way it is to keep the heat flowing off the techs as fast as possible.

Does this make any sense?

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

That would be the way I would design it. I would also look into a triple pass also, especially since it's just going to be on a single CPU loop and have a good pump or 2 to keep it moving....

Well that is my

PS, Martin I would be using the 62mm TEC's and be looking for the most cooling I could get. This is not for 24/7but for some fun benching, the techs will be run by some Variable Power Supplies so Voltage and Amps are not a concern.

How could I get some of your time to get one of these too

I will shoot you a PM....

[Duh]

Hi, I have been thinking of the same thing for a while, but since i don't have equipment, i have been looking for some kind of equipment to mod to do kinda the same thing.

With that said, wouldn't it be more efficient to run some kind of dual pass over the cold side of the loop, like a radiator flows and have as much "cooling water time" on the TEC's cold side as possible, why just run the water down the shoot once, loop it back and have both the input and output on one side, again just like a radiator. I think you will add a little more flow resistance to the loop but the gain from using the TEC's cooling more efficiently by giving the water more cold face time would be much more beneficial.

BUT on the hot side keep it just the way it is to keep the heat flowing off the techs as fast as possible.

Does this make any sense?

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

That would be the way I would design it. I would also look into a triple pass also, especially since it's just going to be on a single CPU loop and have a good pump or 2 to keep it moving....

Well that is my

PS, Martin I would be using the 62mm TEC's and be looking for the most cooling I could get. This is not for 24/7but for some fun benching, the techs will be run by some Variable Power Supplies so Voltage and Amps are not a concern.

How could I get some of your time to get one of these too

I will shoot you a PM....

I like your ideas but such as my ideas they lack of scientific backup. would be nice to do some floworks reproduction of it .

anxious to see how this turns out

[littleowl]

With that said, wouldn't it be more efficient to run some kind of dual pass over the cold side of the loop, like a radiator flows and have as much "cooling water time" on the TEC's cold side as possible, why just run the water down the shoot once, loop it back and have both the input and output on one side, again just like a radiator. I think you will add a little more flow resistance to the loop but the gain from using the TEC's cooling more efficiently by giving the water more cold face time would be much more beneficial.

not really. the water will only pick up so much heat. so if you do a double pass or even more you will just carry all the heat from one pelt to the other pelt. if you look at the maze4-1 from danger den you will see the same thing in a way. but only for 1 tec. The reason the maze was never a very good tec block was because one side of the tec got cool while the other side stayed hot.

[systemviper]

not really. the water will only pick up so much heat. so if you do a double pass or even more you will just carry all the heat from one pelt to the other pelt. if you look at the maze4-1 from danger den you will see the same thing in a way. but only for 1 tec. The reason the maze was never a very good tec block was because one side of the tec got cool while the other side stayed hot.

I'm not sure what you mean, I would have a separate heat dissipation loop with say 3 - maze 4's, 1 on the back of each of the 3 pelts running to a seperate loop of a 120.3's or 2.....to remove the heat with out impacting the cooling of the water in the "Martin rig" so the water is only coming in contact with the cooling side of the pelt and making the second pass just gaining more cooling as it moves thru the martin rig. My objective is to get the water as cooled as possible....

[NaeKuh]

the cold side is going to be connected to an RD-30, PA120.3, EK Supreme, MCW30.

The hotside will be a very small loop with only a EK-XRES PA160 and a DDC3.2

I highly doubt i'll be lacking in the flow department for either side of the loop.

[systemviper]

the cold side is going to be connected to an RD-30, PA120.3, EK Supreme, MCW30.

The hotside will be a very small loop with only a EK-XRES PA160 and a DDC3.2

I highly doubt i'll be lacking in the flow department for either side of the loop.

Yea, sorry to be a little off topic, i was thinking of making it better for my use with a dual pass cooling loop, and not for your system.

Sorry for the hijack, now back to your regularly scheduled thread.

[littleowl]

I'm not sure what you mean, I would have a separate heat dissipation loop with say 3 - maze 4's, 1 on the back of each of the 3 pelts running to a seperate loop of a 120.3's or 2.....to remove the heat with out impacting the cooling of the water in the "Martin rig" so the water is only coming in contact with the cooling side of the pelt and making the second pass just gaining more cooling as it moves thru the martin rig. My objective is to get the water as cooled as possible....

that is fine on the cold side but what about the hot side? the hot side has to be cooled and that is what I was talking about. he may gain some if he did multi pass on the cold side loop idk that he would gain enough for the extra time involved

[systemviper]

that is fine on the cold side but what about the hot side? the hot side has to be cooled and that is what I was talking about. he may gain some if he did multi pass on the cold side loop idk that he would gain enough for the extra time involved

Yea, but I was more referring to MY idea and not thinking about his needs,

For my needs i would mount a maze4 on the back of each pelt (hot side) in a separate loop and solve the heat issue so the cooling would be sweet on the dual pass (cold side), i would even think of going triple or quad pass myself, just to keep the water on the cold side of the pelts as long as possible, then run it straight into the cpu petl block, but we have 2 different objectives, he is looking to drop his temps by a bit and run 24/7; I am looking to get my CPU as low as she can go for benching numbers....

Again sorry for being off topic...

[Martinm210]

FYI
Fabrication is in progress, this must be the biggest workpiece I've milled yet...

RAW Materials...



Lathe put away, and chips are flying on the mill...


I'll keep taking photos of the progress and post them here for fun..

I'll try to answer a few questions:

The base is being constructed from 1/4" copper. I think one thing that is being confused here regarding thickness is a cold plate vs a CPU block design. You NEED thickness when you're trying to transfer heat being collected from one area to another. Because this block has full channel coverage, extra thickness is not necessary and will only act as more of an insulator than anything. I'm purposfully leaving the pin hieght .100" to allow a very small amount of water to pass over the top of the pins as well. Overall a simple channel design like this will approximately double the heat transfer area. This will leave about .150 thickness for the base. I think a combination of .150 base thickness and the .750 delrin should provide an adequate amount of bending resistance to sandwich the TEC. The block will be big enough for either 4x 40mm TECs or 3X 62mm TECs.

Naekuh and I discussed keeping the fin design simple, so they are going to be 1/8" channels wide, this will also minimize the pressure drop in the actual chilled CPU loop, so you are not taking a big hit there. It could very well be a pin matrix design, but youre all aware of the pressure drop of a standard Apogee GTX, the distance from barb to barb is only about 1.5 inches or so, imagine the pressure drop having 5X that length. I think you need to be careful about pressure drop on something this big especially on the cold side were you still need good flow rates to keep the CPU block cool. Also milling diamond pins is extraordinarily time consuming, I've spent as much as 4 hours milling a small diamond pin matrix for a CPU block. I'm not a CNC machine...

[systemviper]

Excellent work guys, i think since it's not going directly on the cpu and has plenty of transfer area, especially at the thickness you are talking about, a diamond cut barb pattern would be overkill anyway. This is an exciting project and i am looking forward to seeing the finished project!

[Martinm210]

Small update:
Base shapes all milled, working on the milling down the perimeter of the first block. That's alot of copper!

I'm sure glad I finally got some good spring collets, it's amazing how much faster you can cut with a solid collet.



Back to it...

[littleowl]

Looking great martin. Naekuh and I have also talked about the design of the channel. That is why I was saying I think same channel but on the hot side plate maybe make the fins a little higher. I do agree you want to watch the resistance. But also if you look at most of the water blocks the pins/rows always go clear up to touch the top. I don't feel that we want that but I think the water may get in the slots and not really slosh out much if there is too much water flow. I think on this design we need to have the fins up high enough the water at least hits them instead of flowing right over them. Tec makes a bunch of heat and you don't want to let any heat sit. I know you know all of this.


It is looking great!!!

[Pete]

Im mega impressed lads! I'd love one (well 2) and use 3 437w tecs all at there full voltages etc etc and see how cold i can get the water to go! Just for fun