Convection + 2 x Heatpipe + TEC

[Jorlen]

Hi this is my first post in this excellent forum. I have not been an active overclocker for many years but have as of lately gotten interested in cooling again.

I have read a few threads here about TECs and I notice a lot of people think they they are of little use (reffering to the "Is T.E.C. Dead?" thread).

I have made a quick sketch (in paint, hehe) to show one way in which to use TECs focusing on their strong points:
No noise
Small size

While trying to minimize the weak points (compared to phase change mostly):
Low to horrible COP, Coeffecient of performance
COP= Q/Pin
Q energy in at coldside
Pin = Electrical energy draw
At the hotside of a TEC the energy dumped will be Q+Pin.

It is of some importance to maintain decent COP within you cooling solution. The final heatsink/radiator will grow larger/louder the worse our COP gets, given a wattage to cool.
The more heat you dump on your rad/sink the warmer it gets, as does you CPU as a result.

How to achieve decent COPs with TECs.
Low heatflow density, which means low amount of energy transported per TEC.
Low temperature difference between hotside and coldside.

If you translate that into application you need to use a large nr of peltiers to spread the heatload and run them at low voltages to keep temperature difference (and power consumption) down.

Now from another viewpoint, the one which I started from. How do I cool my computer passivly and effectivly. The size of a heatsink needed to passivly cool a quad to 10-20 degrees above ambient would not be practical.
To more you want to convect from a heatsink passivly the larger the heatsink or the higher the temperature of the heatsink. This is where the TEC comes in. The TEC will be able to push more heat through a smaller heatsink by raising its temperature. This will benefit a passive heatradiation and passive convection cooling solution more then a solution based on forced convection. The beauty of passive cooling is that it has a C/W that goes down as the temperature of the rad/sink goes up. It is true for forced convection cooling too, but it happens at lower temperature for passive cooling. Think of heat radiators in houses they can move 2000W through passive convection, thats what happens when size and temperature difference to ambient goes up. We need less then that, I am shooting for a COP of 0.5 or higher Which means I would dump approx 3 times the heat of the CPU into the passive rad.

The reason for me posting this i because I want someone to build it to see how well it works when implented.

I have gone on long enough now. I hope some interest has been sparked with this post. Questions and suggestions are welcome.

[Jorlen]

I made some rough calcs using the "400W !!!" peltiers that you can buy off ebay.

Using 4 of them running at 6V you will achieve a 40 deg C temp difference drawing 300W and moving about 140W of heat from coldside. Dumping 440W on the Radiator.

using 6 pelts at 4V we get 140W and 210W dumping 350W.

The heatpipes makes it all possible since they enlarge the cooling surface of the CPU to fit this many TECs.

In thermal practice it uses the same principals as chilled liquid cooling with heatpipes instead of water and pump. And from what I understand heatpipes contend very well with watercooling.

[littleowl]

how are you planing on getting rid of the heat from all them tecs?

[hollo]

hey, nice
this was my variation on your idea: i was considering making a sub-ambient (preferably sub-zero) peltier cooler that would be aircooled (preferably with quiet/low-speed fans). i thought about doing it passive, but very low-speed fans are still a huge improvement in cooling-power per surface-area compared to passive. of course passive would be possible but you'd probably need to make a very very large home-made heat-pipe array... i figured it might be possible to make a fairly cheap but massive gravity-fed heatpipe-radiator with aluminium pipes, but back on topic..

clearly i'd need to use a whole bunch of peltiers. i figured i could make a fairly cheap heat-pipe tower by simply welding 6 plates of metal together, fill it with a little butane (or propane), and then mount low-powered peltiers all over it. so this design, say it had 2 rings of 4 peltiers on the square heatpipe-tower + 1 on top would be a total of 9 peltiers (5 peltiers are shown since its a cross-section). if each peltier consumed 50 watts and moved 25 watts that'd be 225 W of cooling power, which should be enough to get any dual-core sub-zero, and maybe even a 45nm quad. a $5-10 s939 or AM2 cooler should be able to handle that heat-load with a fan running at low speed, 5 V, maybe 3.3 V. of course in the end it'll be noise-level vs cooling power, as it often is in this area from silent passive air-cooling to very noisy phase change

and you can get 10-packs of peltiers on ebay pretty cheaply - say you got 10 120W 12V peltiers and ran them at 5 volts from old computer power supplies...

i like the idea of using large numbers of small air-cooled peltiers, i hope someone does it, although personally i've been thinking more along the lines of making a large, flat, several-metre square heatpipe array with a rotating desk-fan to blow across it, and using it to cool a couple of large (~500W) peltiers, one slapped on either side of a CPU (one on the CPU, one behind the mobo) with the CPU-socket soaked in oil
but that's a story for another thread if i ever do it

[STEvil]

You guys need to look at the air cooled TEC heatsinks like the Titan Amanda then go to the custom heatsink thread in Air Cooling and look at my modified Thermaltake Big Typhoon.

Hint hint.

[Jorlen]

Getting rid of the heat?

The heattransfer of 4x10mm heatpipes will be able to transport the heat of 4 TECs suberbly, energy flow up to 500w-700w should be no problem. Those pipes need to be between 500mm to 700mm long to reach outside the Computers case and to house a nr of thick 1-3mm cooling fins spaced 20-40mm apart.
The size of this convection radiator would be 6-20 dm3 (litres) and would sit on top of the computer.
Its no small thing thats for sure.

[littleowl]

steve is saying the heat transfer of a heat pipe is not as good as you may think. your talking about a lot of heat and it has to be transfered very fast.

[Jorlen]

The idea of a squareheatpipe is great. It would be one massive sucker of a pipe. But it would be a great transporter of heat energy.

There would be temperature differences and pressure differences when running and turning the system on and off. With such a huge area, the pressure will be able to mount a great deal of force so the walls would be quite think compared to the more pressure resiliant geometry of round pipes. Which would make it heavy and a bit more expensive. Other then that its really nice way of expanding the surface area of the CPU to allow for multiple TECs.

I would prefer a horizontally oriented motherboard so that the column could shoot straight up. It would look nice with a bunch of old Zalman heatsinks, cnps 6000-6500. With their V-shape they would cover the pillars sides.
Would cost quite a bit though.

[Jorlen]

I am not guessing the capacity of heatpipes. They are well documented. And for the system to work you have to do your homework.
What I propose should work well according to all the data and formulas provided by the manufactures of CPUs TECs and heatpipes. Heatpipe capacity increases greatly with size and 4x10mm heatpipes should be enough to transfer that amount of heat according to heatpipe manufacturers. And the thermal resistance of the pipe is not greatly effected by length and a pipe of length up to 2 meters should work almost as well as a short 0.2m pipe.
The numbers are all good.

Now I want someone to make one to prove me right.

[littleowl]

I want to see the #'s!!!! but I am still wondering what size tec's your planning on using.

[Jorlen]

http://www.enertron-inc.com/enertron...-selection.php
http://www.norenproducts.com/Heat_Pi...sign_data.html
According enertron a 9.5mm heatpipe would be able to transport roughly 200watts at 30c working temperature. Maybe 12mm heatpipes would be safer, they have more margin.


http://new-electronics.gr/peltier/12726-12730.pdf the "400w" TEC
The TEC nr came from the diagrams in this document. And I believe the this document to credible, I have DLed Aztec, the peltier selection/modelling program from Melcor.com and their peltiers show the exact same scaling. But they dont have the same highpower models.

Try 4v or 6v in the diagrams to see what kind of Qc you get.

[littleowl]

I don't know what it is yet but I don't think this is going to work like you want it to. I cant put my finger on it but I don't think it will work really well.

The Tec specs you linked me to are from thermal enterprise and there #'s are always wrong. when they say 400w tec it is really a 320w tec. Just so you know to watch that.

[ArtosDracon]

I don't know what it is yet but I don't think this is going to work like you want it to. I cant put my finger on it but I don't think it will work really well.

The Tec specs you linked me to are from thermal enterprise and there #'s are always wrong. when they say 400w tec it is really a 320w tec. Just so you know to watch that.

It is a 400W tec, just not a 400W Qmax tec, there are two wattage ratings on TECs the ammount of power they draw and the ammount of heat they can transfer. They draw 400w to transfer 320W.

Besides that though, the think thats bothering you and the reason this isn't going to work as well as one might think is the thermal resistance of the materials involved and worse yet the thermal resistance in the joints between the different transfer mediums.

Beyond that, only powder metal heatpipes can operate with any degree of efficiency below ambient temperatures and they'll have to be insulated between the TEC and the CPU and the entire socket area will have to be insualted if dropped below the dew point which can be as high as 3C below ambient.

Beyond that even, to get effective thermal transfer you'll need a bare minimum of 4x8mm powdermetal heatpipes from the CPU to the TECs just to properly disipate the thermal load from the cpu, if you want below ambient, I'd say 10mm minimum probably 12mm. Then, you're going to be stacking heat loads so, take the heat load of the processor, lets say 200W for S&G, add that to whatever the DRAW wattage of the TECs are and you'll have to disipate that much heat from the hot side of the TECs. so assuming you used a smaller 220W deltaT tec, your draw wattage will be about 300W per tec so you've got 200W+(4x300W) for a total of 1400W of heat that you'll have to disipate from the hotside of the four tecs. Thats 350W per heatpipe. You're looking at 16mm groove or fluid heatpipes, 14mm powder metal heatpipes just to keep the TECs from burning up, bigger would be mandatory if you want to run them at their peak efficiency of below 30C. Now, you have to factor in the thermal resistance of each material being used, if you can even find them. Lets say you get lucky and you're looking at 80% thermal efficiency per material. thats a 20% resistance at each joint and you have 3 joints before you get to the disipation surface, probably 4 by the time you get to the actual disipation fins.

Sorry to burst your bubble.

[Jorlen]

The thermalenterprises data might be decieving but unless they flat out lie, this is the output of their data. The COP running the TECs at 4v is between 1.6 to 2.0 with a 10 degree delta and 0.5-0.7 when running a 40 degree delta. That means load on final stage of dissipation will be anywhere between 150% of the CPU to 400% of the CPU output. 300w-800w. With big enough heatpipes and big enough radiator you will be fine. And yes 12mm is propable needed if you wanna go substantially below ambient.


You might be right about the coldsides heatpipes, non-water based heatpipes which are required for the coldside have 30%-50% of the capacity of a waterbased pipe of same size. I dont think I took that into account.
But it is only necessary if you wanna go below 10c on the coldside.

Its not supercooling just really good noiseless cooling.

[ArtosDracon]

The thermalenterprises data might be decieving but unless they flat out lie, this is the output of their data. The COP running the TECs at 4v is between 1.6 to 2.0 with a 10 degree delta and 0.5-0.7 when running a 40 degree delta. That means load on final stage of dissipation will be anywhere between 150% of the CPU to 400% of the CPU output. 300w-800w. With big enough heatpipes and big enough radiator you will be fine. And yes 12mm is propable needed if you wanna go substantially below ambient.


You might be right about the coldsides heatpipes, non-water based heatpipes which are required for the coldside have 30%-50% of the capacity of a waterbased pipe of same size. I dont think I took that into account.
But it is only necessary if you wanna go below 10c on the coldside.

Its not supercooling just really good noiseless cooling.

I must be missing how it's going to be noiseless. Even assuming you are "only" going to need to dissipate 800w of heat how to intend to do it noiselessly? I thought I saw that this had intentions of being air cooled. Look at the heatsinks it takes to cool just the processor itself, you'll need 4 TRUEs to cool this with heavy duty fans. and if water cooling why even bother with the second set of heatpipes, just water cool the tecs. 800W will only require a pair of mcr320s or a pair of PA120.3s.........

[Jorlen]

I think you missed the part where I explain that you need a radiator on the final heatpipes the size of a small computer case. Which is many times larger then 4 TRUEs.
Look at a standard heating radiator for a water carried heatsystem for a normal house. Each of the those radiators dissapate between 200-2000w per radiator. passivly and quietly. By being seriously big and having a temperature of 20-40 degrees above ambient and taking advantage of convection.
So no it wont be pretty. But it would (should) work.

[ArtosDracon]

I think you missed the part where I explain that you need a radiator on the final heatpipes the size of a small computer case. Which is many times larger then 4 TRUEs.
Look at a standard heating radiator for a water carried heatsystem for a normal house. Each of the those radiators dissapate between 200-2000w per radiator. passivly and quietly. By being seriously big and having a temperature of 20-40 degrees above ambient and taking advantage of convection.
So no it wont be pretty. But it would (should) work.

i don't know that it would need to be that big, and yes, I did miss that, still can't find it but, it doesn't really matter. were it that big and there was a fair ammount of airflow in the room it could be cooled passively. 4TRUEs *should* be sufficient with enough airflow. Regaurdless even that puts this relatively far outside the realm of plausible/reasonable does it not?

[Jorlen]

I dont think its something anyone can build, I know it would take me a few tries for sure before I got it right. But someone familiar with heatpipes that have access to cheap copper copper and a workshop could propably toss one together at the cost of a high end watercooling system. And it would perform like one too. And the space it would take up, well nothing is free. If you want less noise, it takes more space.

[littleowl]

It is a 400W tec, just not a 400W Qmax tec, there are two wattage ratings on TECs the ammount of power they draw and the ammount of heat they can transfer. They draw 400w to transfer 320W.

Besides that though, the think thats bothering you and the reason this isn't going to work as well as one might think is the thermal resistance of the materials involved and worse yet the thermal resistance in the joints between the different transfer mediums.

Beyond that, only powder metal heatpipes can operate with any degree of efficiency below ambient temperatures and they'll have to be insulated between the TEC and the CPU and the entire socket area will have to be insualted if dropped below the dew point which can be as high as 3C below ambient.

Beyond that even, to get effective thermal transfer you'll need a bare minimum of 4x8mm powdermetal heatpipes from the CPU to the TECs just to properly disipate the thermal load from the cpu, if you want below ambient, I'd say 10mm minimum probably 12mm. Then, you're going to be stacking heat loads so, take the heat load of the processor, lets say 200W for S&G, add that to whatever the DRAW wattage of the TECs are and you'll have to disipate that much heat from the hot side of the TECs. so assuming you used a smaller 220W deltaT tec, your draw wattage will be about 300W per tec so you've got 200W+(4x300W) for a total of 1400W of heat that you'll have to disipate from the hotside of the four tecs. Thats 350W per heatpipe. You're looking at 16mm groove or fluid heatpipes, 14mm powder metal heatpipes just to keep the TECs from burning up, bigger would be mandatory if you want to run them at their peak efficiency of below 30C. Now, you have to factor in the thermal resistance of each material being used, if you can even find them. Lets say you get lucky and you're looking at 80% thermal efficiency per material. thats a 20% resistance at each joint and you have 3 joints before you get to the disipation surface, probably 4 by the time you get to the actual disipation fins.

Sorry to burst your bubble.

exactly all this is true.

but that is not why I don't think it will work.

[Jorlen]

Your a born sceptic, I get it.
But I am getting a bit annoyed with people saying that a system like this COULD fail in achieving what it is intended for. Of course it could! Just do a mistake in the calcs or in the manufacturing and its gonna fail.

What I am saying is that it could work, and it wouldnt be guess work getting it right. All the information is available. And the resources needed to complete a project like this are on the same scale watercooling and phasecooling.

[ArtosDracon]

Your a born sceptic, I get it.
But I am getting a bit annoyed with people saying that a system like this COULD fail in achieving what it is intended for. Of course it could! Just do a mistake in the calcs or in the manufacturing and its gonna fail.

What I am saying is that it could work, and it wouldnt be guess work getting it right. All the information is available. And the resources needed to complete a project like this are on the same scale watercooling and phasecooling.

I am a born skeptic and a trained cynic but, I'm not just saying it could fail. I'm pretty confidently saying that unless this is built by someone with a vast knowledge of heatpipes, TECs and thermal transfer and with a vast ammount of experience in milling, copper welding and fin design that the chances of it performing up to expectations are pretty slim. IF a person or better yet a small group of people properly designed, machined and assembled this then yes, it would cost similar to high end water cooling and could perform in the area of well, other well built TEC chillers, 0C load at best probably, depending on the TECs used(just an estimate) and at that it's going to either need a lot of airflow or be quite large or a combination of midpoints of the two. I'm not saying it's impossible or even implausible, simply impractical.

[littleowl]

Your a born sceptic, I get it.
But I am getting a bit annoyed with people saying that a system like this COULD fail in achieving what it is intended for. Of course it could! Just do a mistake in the calcs or in the manufacturing and its gonna fail.

What I am saying is that it could work, and it wouldnt be guess work getting it right. All the information is available. And the resources needed to complete a project like this are on the same scale watercooling and phasecooling.

I am no sceptic but if you think I am that is fine also. If you feel so strongly that I am wrong in saying it will not work then please just build it and show the #'s to prove it. You keep saying that you cant do it because you don't have the equipment, There are machine shops all over the place and I am sure your home town has one in it.

[systemviper]

I just saw the new Thermalright Ultra-120 Extreme Heatpipe CPU Cooler, if this was put on one TEC hotside, would it be able to dissapate the heat of one of those 226w tec's running at full potential.

Is there somewhere that shows the optimum heat transfer in watts that these CPU coolers can handle?

[littleowl]

I am not sure systemviper. I will try and get some time here in a little bit to see what if I can answer your question.

@ Jorlen
If you want to prove me wrong and show me that this idea of yours will work then contact this guy in IM and have him make you the heatpipe

[hollo]

so where were we..

ignoring you skeptics, since you're not working through the whole system properly, or even trying it seems-

suppose, pessimistically for safety, CPU gives off 200W
4 6mm heatpipes can handle about 400W, no problems there.

peltiers running @ 4V with 40 degree celcius dT have a COP of .5, so you need 400W of peltiers

The COP running the TECs at 4v is between 1.6 to 2.0 with a 10 degree delta and 0.5-0.7 when running a 40 degree delta.

so total heat output on the hot side of the peltiers is 600W (pessimistic)
and you have a theoretical maximum heat capacity of 800 W with the 4 larger heatpipes

and assuming the power-draw of the 16V 400W TEC at 4V is 1/4 of its power-draw at 16V (i'm not sure if that's gonna be high or low, Jorlen you seem to have checked that out more than me) that means you'll need... 4 of them
as drawn

what was so wrong with the plan again

edit: bunged the numbers here - at 1/4 the voltage you'll get 1/16th the power draw because P=V^2/R
so from my rough working it'd seem 6 peltiers at 4V should be able to handle 150W
or

using 6 pelts at 4V we get 140W and 210W dumping 350W.