Anyone using Peltiers on their res to chill coolant?

[Negatron]

I've done some browsing of the pictures thread and I've been a little surprised that I haven't seen any metallic reservoirs covered in peltiers for chilling the water. Does anyone do this? I guess it would be especially convenient for systems with externally mounted reservoirs since the heat sinks on the hot sides of the pelts could be easily cooled (maybe even passively) with unlimited flow of room air. This kind of a setup would even create some cool opportunities to use crazy custom/hand made copper heat sinks.

Also how come I don't see more people using copper automotive heater cores as radiators for exchanging heat from liquid coolant?

 

[Bobnova]

To the peltier on reservoir idea:
It does work, but there are some issues.

#1 is that if the pelt is big enough to be useful it will freeze the water in the reservoir over time and eventually plug it. The workaround is to use antifreeze in the water and to control the power the pelt puts out with a microcontroller.

#2 is that you can only use that to get down to ambient temps, the moment the pelt tries to get the water under ambient the radiators start collecting heat and putting it into the coolant instead of the other way around. The radiators fight (hard!) to equalize air and water temps.

#3 is that peltiers are woefully inefficient.

All told it's doable, but expensive and inefficient and will only gain you a couple degrees if you're adding it to a properly designed loop.



On the automotive heater cores: People used to, back before WC radiators were cheap. Now they are fairly cheap and they fit into cases, which is something heater cores don't do very well. Plus radiators are easy, they bolt in, no modding needed. People are lazy.

 

[S_I_N]

did the peltier route before while it was fun its kinda expensive for the minimal gain. They for the most part require their own power source so thats extra to the light bill again for a minimal decrease in temps they offer. Now the peltier cooling I did was via pelt to die and water block to remove the heat but again minimal decrease in temps. It did however tame the Prescott P4 a degree.

 

[Negatron]

did the peltier route before while it was fun its kinda expensive for the minimal gain. They for the most part require their own power source so thats extra to the light bill again for a minimal decrease in temps they offer. Now the peltier cooling I did was via pelt to die and water block to remove the heat but again minimal decrease in temps. It did however tame the Prescott P4 a degree.

I remember having to underclock a couple P4s back in the day. I can't imagine a Peltier being an anywhere near practical option for cooling one directly!

 

[Negatron]

I guess it is probably most practical to use Peltiers on reservoirs on HTPC type applications where OTF trans-coding and the like is needed, for keeping fan speeds and temps down at the same time.

Eliminating liquid to air heat exchangers it would be possible to get sub ambient coolant temps. It would be even less energy efficient than using a regular radiator in conjunction with peltiers would be but this way there would be no radiator to take on energy from room air and exchange it back to the coolant. Cooling the liquid directly with pelts and cooling the pelt hot sides with room air would surely take a lot of power but keeps below room temp coolant. Has anyone done this specific type of setup?

 

[hokiealumnus]

To keep fan speeds & temps down at the same time, use water; add more radiator if you're having to run fans too fast. It feels like you're forgetting you'll have to cool the hot side of the pelts, and doing so will not be anything near quiet and definitely not quieter than a slightly larger water loop.

In this day and age, pelts are nothing more than a throwback to a bygone era when chips put out a lot less heat than they do now. Unless you're just looking to play around with them, when you add up their power draw and potential condensation problems, it equals a practically useless solution.

 

[Negatron]

To keep fan speeds & temps down at the same time, use water; add more radiator if you're having to run fans too fast. It feels like you're forgetting you'll have to cool the hot side of the pelts, and doing so will not be anything near quiet and definitely not quieter than a slightly larger water loop.

In this day and age, pelts are nothing more than a throwback to a bygone era when chips put out a lot less heat than they do now. Unless you're just looking to play around with them, when you add up their power draw and potential condensation problems, it equals a practically useless solution.

The thinking behind being able to reduce sound pressure is that, when moving heat away from coolant using pelts, then moving heat from pelts with heat sinks, semi-passive cooling can be employed by using lots of pelts and a huge passive heat sink which wouldn't normally be practical with a standard case. Using liquid to move heat away from the CPU and then using peltiers to move heat away from the liquid means it might be practical to use giant heat sinks that need no fans or that, being so large, can be fitted with very low speed fans which certainly isn't 'practical' using traditional liquid to air heat exchangers. Of course, being so exotic and complicated, none of this is what I would really call practical but can probably be accomplished using mostly off the shelf parts.

It's something I would like to try but I really have no actual use for at the moment. I would need access to a pretty well equipped machine shop (not to mention a whole bunch of disposable income) to do it the way I imagine it should be done. someone send me pictures of something similar if you know of any!

 

[Bobnova]

It'd have to be huge.
If you have a 100w CPU you're going to need ~150-200w of pelts to keep it decently happy, that means that your pelt cooling loop is going to need to get rid of 250-300w.
That takes some doing, quite a bit more doing than 100w would have in the first place.

 

[hokiealumnus]

Low-FPI radiators + low speed fans = low noise. Any pelts, plus any size heatsink = more noise than a solid loop with properly designed radiators for that application.

I admire your gumption and inventiveness, but the practical side of this just isn't there. At least IMHO.

 

[Negatron]

It'd have to be huge.
If you have a 100w CPU you're going to need ~150-200w of pelts to keep it decently happy, that means that your pelt cooling loop is going to need to get rid of 250-300w.
That takes some doing, quite a bit more doing than 100w would have in the first place.

I think the one that you are describing is different than my most recent imagination of this kind of setup. Heat would be exchanged directly from the coolant by peltiers which would then be cooled by heat sinks. In this way it would be possible to use a nearly unlimited number of peltiers and heat sinks. Presumably more peltiers and heat sinks means a relatively low difference in temperature between hot and cold sides.

I don't have the best understanding of the use of Peltiers for any purpose. That said I do have pretty good understanding of most practical and some theoretical physics and nothing I have learned in those areas is really telling me that this kind of thing couldn't be done with success and without a lot of hassle. Practical is a matter of opinion, I guess, and is probably not the best word for describing taking on something that can probably be accomplished with much less trouble and time. I think I still want to try it!

 

[hokiealumnus]

Well in that case....

 

[RnRollie]

well 100 watts of cpu would still need +/- 200 wats of pelts, which means that you have to dimension your passive heatsinks to dissipate +300 watts (by convection).
Depending on the material used (copper, silver, stainless,...) you are looking at a few (heavy) square meters...
there's a formula hidden somewhere: http://www.engineersedge.com/properties_of_metals.htm

Pratically, in a passive water loop , the size of the passive rads would be smaller that the size of the passive rads in a loop + pelts

And if practicality is relative... why not just use a geothermal loop?

But, please go ahead, i like to see these things happen

However, -dont take this wrong, but- in your understanding of peltiers, you seem to be overlooking that the use of magic to create or destroy energy is not accepted in physics.

Anyhow, success and post pics

 

[Negatron]

well 100 watts of cpu would still need +/- 200 wats of pelts, which means that you have to dimension your passive heatsinks to dissipate +300 watts (by convection).
Depending on the material used (copper, silver, stainless,...) you are looking at a few (heavy) square meters...
there's a formula hidden somewhere: http://www.engineersedge.com/properties_of_metals.htm

Pratically, in a passive water loop , the size of the passive rads would be smaller that the size of the passive rads in a loop + pelts

And if practicality is relative... why not just use a geothermal loop?

But, please go ahead, i like to see these things happen

However, -dont take this wrong, but- in your understanding of peltiers, you seem to be overlooking that the use of magic to create or destroy energy is not accepted in physics.

Anyhow, success and post pics

I have quite a good understanding of physics for someone without a phd in it; better than most people. I disagree that it would take several square meters of radiator for this application, magic or no magic, and I think an actual physicist would agree with me. I would even go so far as to say that statement is incorrect. The type of energy transfer we are talking about is conduction which is many times more efficient than convection for this application and most applications in fact. Square meters of surface might be needed with an average material and poorly designed/implemented form factor but one square meter of surface area with the right material using the right form factor can be made to exchange enough energy for 300-400 watts without forced air. It can be done but it's something I will get in to some day when many other projects are finished, within budget.

Geothermal, as cool as it would be to use for a computer, is far too complicated, expensive, and physically large for my taste. Being that it would cost 50 times more than what I'm thinking of I'm afraid it's something I would never try for a fun project. If I was already using geothermal to heat and cool my house I most certainly would have already added another water line to the loop for my gaming rig! I wonder if anyone has ever actually done that....

 

[wagex]

 

[Bobnova]

A square meter is plenty. A square meter is also expensive.

 

[RnRollie]

I have quite a good understanding of physics for someone without a phd in it; better than most people. I disagree that it would take several square meters of radiator for this application, magic or no magic, and I think an actual physicist would agree with me. I would even go so far as to say that statement is incorrect.

If you take apart a fairly standard "tower" heat sink (eg a CM Hyper EVO), or a simple 120.1 radiator and put the (unfolded) fins next to each other, you end up covering quite a surface. Several square cm, feet. To (passively) dispense of heat into air, its surface that counts, whether its "flat" or "folded".

The type of energy transfer we are talking about is conduction which is many times more efficient than convection for this application and most applications in fact. Square meters of surface might be needed with an average material and poorly designed/implemented form factor but one square meter of surface area with the right material using the right form factor can be made to exchange enough energy for 300-400 watts without forced air. It can be done but it's something I will get in to some day when many other projects are finished, within budget.

Fair enough, if you bolt a copper sheet to a concrete floor or submerge in a pond/river. If not, and your end stage is to transfer the heat to air... its no longer conduction. While i agree that the correct material and the correct form will help a lot, but unless you use a different end stage, you'll end up with the less efficient hand-off to air. Where surface counts.
Alphacools Cape Cora comes close to what is very doable in that case, pity its Alu
http://www.performance-pcs.com/cata...e=product_info&cPath=59_310&products_id=30156

I'm sure its all well thought out, but sofar, i haven't seen any other other end-stage as xfer to air.
Yes you can cool a res to freezing if you want, but the hot side of the pelt needs to be cooled. So far i see the heat being transferred from CPU into waterloop and then being sucked out by a pelt. From the pelt it has to go somewhere and you are proposing passive to air.
And i say to get rid of +300 watts by passive convection cooling, you need surface. Show me where i am wrong, and i'll shut up.

Geothermal, as cool as it would be to use for a computer, is far too complicated, expensive, and physically large for my taste. Being that it would cost 50 times more than what I'm thinking of I'm afraid it's something I would never try for a fun project. If I was already using geothermal to heat and cool my house I most certainly would have already added another water line to the loop for my gaming rig! I wonder if anyone has ever actually done that....

nah, its in a way no more complex as a bong cooler

If you think about it, its a closed loop. The most simple approach is to bury 30 feet of 10mm rolled up copper pipe 4-6 feet deep, slip some tubing over the ends, fill up with distilled, done.

Its only relative expensive - 30 feet of copper doesn't come for a nickel+dime and you need a decent pump to compensate for the loss caused by distance.
It scales up wonderfully... nobody is stopping you from burying +100 feet of 3/4 pipe The only other substantial "cost" is that digging a trench and displacing several cubic feet of soil with a spade leads to the discovery of muscles you never knew you had
Of course, going geo is not always practical, besides the digging & some drilling, the infrastructure must be present. One of the reasons why i had to shelve my plans again: i moved to a 7th floor penthouse this year.


PS: i guess you want to achieve the pic on the left, using ordinary HS.
The one on the right would be easier to do without bolting a lot of EVo HS to the contraption... you could just bolt a Cora to the side of your PC case.
It still wont be very efficient thou, because off twice the losses on the contact area between water/res/pelt/hotplate-block/water
Dropping the res and using a plateHX would be more efficient

 

[Negatron]

If you take apart a fairly standard "tower" heat sink (eg a CM Hyper EVO), or a simple 120.1 radiator and put the (unfolded) fins next to each other, you end up covering quite a surface. Several square cm, feet. To (passively) dispense of heat into air, its surface that counts, whether its "flat" or "folded".

You hadn't specified having meant surface as opposed to just square meters of heat sinks so it was unclear what you meant. For the configuration I have in mind it might well take a couple square meters of surface but that can be achieved in a pretty small overall volume.

Fair enough, if you bolt a copper sheet to a concrete floor or submerge in a pond/river. If not, and your end stage is to transfer the heat to air... its no longer conduction. While i agree that the correct material and the correct form will help a lot, but unless you use a different end stage, you'll end up with the less efficient hand-off to air. Where surface counts.
Alphacools Cape Cora comes close to what is very doable in that case, pity its Alu
http://www.performance-pcs.com/cata...e=product_info&cPath=59_310&products_id=30156

Energy flowing from one material to another material, such as copper to concrete or copper to water, is conduction, just the same as energy flowing from copper to air is conduction. Convection does take place in passive cooling but it is a secondary function to conduction, a virtue of the condition of two volumes of gasses of different density and buoyancy, which will cause the lower density air to move and, eventually, transfer its energy to some other material or throughout the gas which surrounds it, causing equilibrium in densities (theoretically).

When energy moves from a copper heat sink to the gas which surrounds, and makes physical contact with, the heat sink that is conduction. Convection would only, technically, occur if the gas which was heated by the heat sink moved, made contact with other matter, and conducted heat energy to that. Most of this usually occurs as the transfer of the heated airs energy to the unheated air which surrounds the heated air which is still conduction. You can think of it this way; If there is a contained volume of gas which is (magically) made not to move at all in any direction and heat goes from one end of this volume to the other end of the volume, this is still conduction, even though the volume is gas and not solid matter. If the same volume is heated, moved to a new location, and the volumes heat energy is transferred to another object, convection has occurred.

That case is pretty badass and probably really a much better solution than what I'm talking about but I'll be doing it mostly for fun and screwing around with peltiers and hand made copper parts and whatnot should be pretty fun! I'll probably end up using manufactured heat sinks for their ease of use and efficiency but mostly for their flat, easily-mountable-to-another-flat-surface, surface (and to save quite a lot of money probably).

I'm sure its all well thought out, but sofar, i haven't seen any other other end-stage as xfer to air.
Yes you can cool a res to freezing if you want, but the hot side of the pelt needs to be cooled. So far i see the heat being transferred from CPU into waterloop and then being sucked out by a pelt. From the pelt it has to go somewhere and you are proposing passive to air.
And i say to get rid of +300 watts by passive convection cooling, you need surface. Show me where i am wrong, and i'll shut up.

It's definitely going to take quite a bit of surface. I never intended to mean that it wouldn't take surface area but, as I said above, it wasn't clear that you weren't talking about square meters of heat sinks as opposed to square meters of surface area. I don't think it would take more than a couple square meters but this remains to be seen without doing research and math and whatnot which I like to avoid. I'll probably end up learning just how much surface I need by starting with a reasonable amount and moving up as needed.

Your first picture is pretty close to what I have in mind, except for I will use a much larger reservoir and many peltiers. By using more, lower power, peltiers and multiple heat sinks I can get the temperatures of the coolant and cpu down while still achieving passive energy transfer between the heat sink and surrounding air. It definitely won't be efficient but it will be fun to do and pretty quiet - maybe even nearly silent if I do the pump right.

nah, its in a way no more complex as a bong cooler

If you think about it, its a closed loop. The most simple approach is to bury 30 feet of 10mm rolled up copper pipe 4-6 feet deep, slip some tubing over the ends, fill up with distilled, done.

Its only relative expensive - 30 feet of copper doesn't come for a nickel+dime and you need a decent pump to compensate for the loss caused by distance.
It scales up wonderfully... nobody is stopping you from burying +100 feet of 3/4 pipe The only other substantial "cost" is that digging a trench and displacing several cubic feet of soil with a spade leads to the discovery of muscles you never knew you had
Of course, going geo is not always practical, besides the digging & some drilling, the infrastructure must be present. One of the reasons why i had to shelve my plans again: i moved to a 7th floor penthouse this year.

The basic idea of using a geothermal loop to cool a computer is pretty simple but it's all of the 'extra' stuff that would need to be done that deters me from trying it. There is all of the bending and fitting of the pipe to actually get the liquid flowing past the CPU, and through the walls of my house, and to the under ground loop. It would be a pain in the *** if I wanted to move the computer to another part of the house or whatever since I would have to run more pipe, etc. There's also the fact that I live on top of a huge block of limestone and it's nearly impossible to dig. The geothermal system that I am familiar with in my area is the kind that is buried in a shallow trench like you are talking about but it takes a huge amount of pipe to get a usable amount of heat transfer this way as opposed to vertical well loops which, like I said, requires digging through solid limestone. If I were super rich I would be using geothermal to heat my house anyway and I would have a dedicated auxiliary well and loop installed for computers and other fun stuff. Basically I would be looking at about $75,000 US to achieve a cost effective solution so I'm gonna stick to peltiers haha.

If I was willing to spend the money and time buying all the copper water line and drilling holes through the side of my house to get them outside I think I would just lay the pipes on top of the ground and shade them from the sun and achieve about the same result.

It still wont be very efficient thou, because off twice the losses on the contact area between water/res/pelt/hotplate-block/water Dropping the res and using a plateHX would be more efficient

I'm afraid I don't quite understand what you are trying to say. I guess you are saying it would be more efficient to use one of those passive aluminum radiation cases than to use the peltiers and passive heat sinks like I'm talking about? If so, I agree, but I can get lower temps using the setup I'm thinking of.

Really the only thing I'm not sure about is how many peltiers of which design to use and just how much copper conducing heat sink I will need but finding out is part of the fun I guess.

 

[SeeThruHead]

Why not just get a really large reservoir and chill it with an aquarium chiller....

 

[RnRollie]

i'm saying that using the passive radiators (loop) would be more efficient & more compact than cooling the pelts hot sides with "ordinary" Air heatsinks.

There are a few other concerns also, see below

As for geo, i've got a design which involves a lot less digging, drilling & bending
i'll (re-)post it someday


A few issues... math is inevitable

No matter how "conductive", we are actually talking about termal barriers.

die > tim (or solder) -> ihs -> tim (eg pk1) -> copper base of waterblock -> water

water -> res wall (copper?) -> cold block -> pelt -> hot block -> tim -> HeatSink base -> heatpipe -> alu fins -> air

Ideally one should work out each layer to find the maximum thermal transfer possible - this is where material properties & C/W values & joules, etc come in
Unfortunately, some of these values are either unknown or unclear or too variable.
Acutally, all of these values or the calculations to achieve these values can be probably found on the net somewhere, but since there is no big "X marks the spot" & i dont feel like doing all that work for you ...
The only thing we know for sure is that the stuff used between the die & ihs on a ivy bridge is "less than optimal"

Therefore, i'm going to commit a few sins & risk a spanking by Conumdrum

The first sin is that i'm going to ignore the whole die -> tim -> ihs -> tim -> block limits and consider it one whole item and ASSUME 100% efficiency
The second sin is that i'm going to forfait C/W, joules , ... and just gonna use WATTS.

Doing so simplifies the whole issue and allows some abstraction , while it increases the error and the (considerable) risk to my own life

Anywas, because of simplex model, we can assume that 1 WATT produced at the die goes straight into the Water.

So, 100 Watts produced goes into the water and means 100 watts has to be taken out by the pelt(s)

Meaning that the pelt(s) have to take out the 100 watts in the same timeframe/cycle at the "cost" of using 150 watts of power. Meaning that 100+150=250 watts will need to be dissipated into the air by the passive heatsink(s)


Here comes the issue:

Assume you've got quite a "beefy" waterloop which contains 1/4 gallon water
Assume you've got a good 1 - 1.5 GPM flow, which is not unreasonable

At 1 GPM flow, it suggests that ALL the water in your loop runs by the CPU & through the res 4 times per minute

So, in order to dissipate 100 watts produced / min, the water has to pick up 25 Watts at each pass AND the pelt(s) need to be able to pick up 25 Watts at each pass. The passive HS need to dump 250 watts/min into the air.

All this to stay close at ambient

So, it all seems to work out on paper (when roughly sketched), given that the thermal capacity of Water is 4.1 kJ/kg °C0 , and 1kg water = 1 liter which is close enough to 1/4 gallon (what a coincidence) to marginalise our ... margin of error... then the water itself does not seems to be the limiting factor.

A few questions remain
Is the waterblock *really* good enough to transfer the equivalent of 100 Watts to the water/min? (given limited contact surface -> not sure)
Are the pelts able to suck the equivalent of 100 Watts/min out of the water? (Stick on enough of them to have enough surface contact -> doable)
Is water able to pick-up & dump the equivalent of 100 watts / min?
Is air capable of picking up the equivalent of 250 watts /min from your passive heatsinks? (i suspect you need to bolt on at least 4-6 evo's to keep it under control)
Will it all be "portable"? (whats the weight of one evo?
What Will be the cost of powering the pelts

And since i am sure that besides my sins, i made some terrible logical errors, will i be able to suffer the consequences with dignity?

 

[RnRollie]

Why not just get a really large reservoir and chill it with an aquarium chiller....

yes, THAT would be lot easier to do
only need a chiller, a loft tank (+20 gals), a plateHX and a simple eheim 600