Heat Transfer Modules

[njordan0526]

I had an idea a little while back about using a peltier unit in place of a radiator to cool the liquid cooling solution. The peltier units temperature is dependent on the amount of electricity being fed to it as well as the temperature of the hot/cold side so with some tinkering it should be pretty easy to find the right balance (I'm sure someone else has approximated this).

 

Essentially you would have your heatsink mounted either outside or near the exhaust area of your case and attached to the "hot" side of the peltier unit. The "cold" side would be attached to a custom fin to cool your solution before it heads back to your reservoir. 

 

Since my computer was stolen a few months back I can't post designs or even hope to put this into action any time soon so I'd love to see this put to practice.

 

Here is the math involved and the basic concept. https://en.m.wikipedia.org/wiki/Thermoelectric_cooling

 

They can be bought online fairly cheap.

[Enderman]

Uh, yes, everyone already did this 10 years ago.

Except people did this with liquid cooling instead of a heatsink because a peltier powerful enough to cool a CPU uses about 500W so you need more than a heatsink to cool the peltier+CPU.

On top of that, they fail easily, and when it fails it becomes an insulator and makes your CPU isntantly overheat.

Also, the more heat you apply to the cold side the more inefficient they become, until you CPU is only a few degrees cooler than it would be without the peltier, assuming you have the radiator cooling capacity for the massive amount of heat created by the peltier, and the PSU capacity to support the peltier.

 

So basically people don't do this anymore because it sucks.

[Deous]

On 11/14/2017 at 9:52 AM, Enderman said:

Uh, yes, everyone already did this 10 years ago.

Except people did this with liquid cooling instead of a heatsink because a peltier powerful enough to cool a CPU uses about 500W so you need more than a heatsink to cool the peltier+CPU.

On top of that, they fail easily, and when it fails it becomes an insulator and makes your CPU isntantly overheat.

Also, the more heat you apply to the cold side the more inefficient they become, until you CPU is only a few degrees cooler than it would be without the peltier, assuming you have the radiator cooling capacity for the massive amount of heat created by the peltier, and the PSU capacity to support the peltier.

 

So basically people don't do this anymore because it sucks.

That's BS. Nowadays we have powerful much better peltiers that won't fail and besides if CPU overheats the new motherboards with shut power off immediately with no risk of damage.

I am surprised that people don't build peltier cooled PC systems. In fact I am about to build one myself.

There are two ways of cooling - cool down water with peltier or cool directly the CPU with surface to surface Peltirer/CPU mount. This system requires very advanced thermostatic temperature control for minimum 15A/12V and you do not need 500W as today's CPUs use less than 100W. All you need is very efficient water flow. I believe such system can overclock to speeds comparable to refrigerated or liquid nitrogen systems

 

[Enderman]

9 minutes ago, Deous said:

That's BS. Nowadays we have powerful much better peltiers that won't fail and besides if CPU overheats the new motherboards with shut power off immediately with no risk of damage.

I am surprised that people don't build peltier cooled PC systems. In fact I am about to build one myself.

There are two ways of cooling - cool down water with peltier or cool directly the CPU with surface to surface Peltirer/CPU mount. This system requires very advanced thermostatic temperature control for minimum 15A/12V and you do not need 500W as today's CPUs use less than 100W. All you need is very efficient water flow. I believe such system can overclock to speeds comparable to refrigerated or liquid nitrogen systems

 

I don't think you understand how a peltier works....

In order to get lower temperatures, the peltier needs to have a much higher heat transfer capacity than the object it is cooling.

If you have a 100W peltier and stick it on a 100W CPU then your temperature drop is going to be pretty much 0 under load.

The more heat you apply to the cold side, the less the temperature delta is going to be.

Even with a 500W peltier (~250W heat capacity) you will see very little temperature drop under load.

And on top of that, not only does your liquid cooling system have to cool 100W from the CPU, but now it has to cool 600W from the CPU+peltier.

In which case you would have probably gotten lower temperatures by simply cooling the CPU directly without the peltier if your loop can cool 600W.

 

If you want to do it then go ahead, I'm just telling you why it sucks and why people stopped doing it.

Peltiers are not magic. You won't get much if any temperature improvement, at the cost of money, heat, and power.

[Deous]

1 minute ago, Enderman said:

I don't think you understand how a peltier works....

In order to get lower temperatures, the peltier needs to have a much higher heat transfer capacity than the object it is cooling.

If you have a 100W peltier and stick it on a 100W CPU then your temperature drop is going to be pretty much 0 under load.

The more heat you apply to the cold side, the less the temperature delta is going to be.

Even with a 500W peltier (~250W heat capacity) you will see very little temperature drop under load.

And on top of that, not only does your liquid cooling system have to cool 100W from the CPU, but now it has to cool 600W from the CPU+peltier.

In which case you would have probably gotten lower temperatures by simply cooling the CPU directly without the peltier if your loop can cool 600W.

 

If you want to do it then go ahead, I'm just telling you why it sucks and why people stopped doing it.

Peltiers are not magic. You won't get much if any temperature improvement, at the cost of money, heat, and power.

Man, where do you get that assumptions from? Peltier doesn't transfer heat externally but makes the charge carriers in the material to diffuse from the one side to the other causing temperature difference or temperature/heat separation.

You don't have to cool the object in 100/100% ratio.  200W Peltier which costs $5 can use 100-150W power to cool down

cpu to very low temps. You probably need only around 20C temps which is water condensation point.

The problem is removing the heat from hot side or it will punch thru the module and disturb the 'Peltier effect'

 

 

[Enderman]

9 minutes ago, Deous said:

Man, where do you get that assumptions from? Peltier doesn't transfer heat externally but makes the charge carriers in the material to diffuse from the one side to the other causing temperature difference or temperature/heat separation.

You don't have to cool the object in 100/100% ratio.  200W Peltier which costs $5 can use 100-150W power to cool down

cpu to very low temps. You probably need only around 20C temps which is water condensation point.

The problem is removing the heat from hot side or it will punch thru the module and disturb the 'Peltier effect'

Just as I thought, you don't understand how a peltier works...

https://thermal.ferrotec.com/technology/thermoelectric-reference-guide/thermalref09/

If you have a 100W peltier, and you stick it to a 100W CPU, your delta T is going to be 0.

Aka you get NO temperature improvement, and you have to dissipate 300W of heat instead of 100W.

I suggest you try doing more research on thermoelectric cooling, or you're going to be severely disappointed when you try it.

[Deous]

1 minute ago, Enderman said:

Just as I thought, you don't understand how a peltier works...

https://thermal.ferrotec.com/technology/thermoelectric-reference-guide/thermalref09/

If you have a 100W peltier, and you stick it to a 100W CPU, your delta T is going to be 0.

Aka you get NO temperature improvement, and you have to dissipate 300W of heat instead of 100W.

I suggest you try doing more research on thermoelectric cooling, or you're going to be severely disappointed when you try it.

That's why you don't want to load all transfer on Peltier only. That would be killing the poor module) Just read about thermoelectric effect instead of some delta measurements. Cooling system must have much heat looses during the exchange path which is no problem with good radiator. Also CPU must be connected to two surfaces conveying heat - one of them is peltier. People building the layout you are talking about were meant to fail or didn't understand thermodynamics. 

 

[Deous]

Besides talking about Peltier - I suggest doing small experiment - cool down one side of Peltier and keep other warm - then measure voltage

The module can also be used as a temperature compensation control circuit.

I am not sure if some of that heat can be transformed into voltage energy that can be discharged instead of dissipated as heat.

[Enderman]

1 hour ago, Deous said:

That's why you don't want to load all transfer on Peltier only. That would be killing the poor module) Just read about thermoelectric effect instead of some delta measurements. Cooling system must have much heat looses during the exchange path which is no problem with good radiator. Also CPU must be connected to two surfaces conveying heat - one of them is peltier. People building the layout you are talking about were meant to fail or didn't understand thermodynamics. 

 

*facepalm*

Yes, the way a peltier works is you put the cold side on CPU and the hot side on the waterblock.

Why are you explaining to me how a peltier works, when you're the one who doesn't understand that the peltier Qcmax needs to be much higher than the CPU TDP??

Please go read the link I posted so that you can hopefully learn why "some delta measurements" are literally the reason why peltiers exist.

The purpouse of a peltier IS to create a temperature delta between the two plates..........

 

1 hour ago, Deous said:

Besides talking about Peltier - I suggest doing small experiment - cool down one side of Peltier and keep other warm - then measure voltage

The module can also be used as a temperature compensation control circuit.

I am not sure if some of that heat can be transformed into voltage energy that can be discharged instead of dissipated as heat.

The amount of electricity produced by using a peltier as a source is hundreds or thousands of times less than what it consumes to transfer heat from the cold to hot side.

And "energy that can be discharged" is called heat. Unless you want to attach a motor to it and turn it into motion, which would be a pretty bad method of dissipating energy.

Also, this would mean that you're not cooling the CPU, you're insulating it because in order to produce electricity a peltier needs a high temperature delta between the two sides.

[Deous]

Check this picture.

The effect I was talking about is called Seebeck Effect and used in Thermoelectric generators.

I will repeat again - Peltier should never be used in a cooling loop as a heat exchange device - it should be used as

the support of heat transport or/and its separation. For example if you put peltier on top of a running cpu waterblock and a good fan on top of the module - you will get better results than putting it between cpu and block.

CPU will freeze easier and the power consumed will never be as high as you say is used in retarded Peltier cooling systems.

 

[Deous]

Well, anyways I think people gave up on Peltier cooling too fast because of some unsuccessful experiments that were using them. Thermoelectric effect is not implemented properly - they use PT as the heat element that generates whole thermal delta by electricity. That's insanity and power used for that is super high because they use it to power Peltier and to remove that power as the heat from the circuit! Heh I am just shocked what some people say about cooling systems.

Peltiers should be used as a heat transporters and not heat exchangers where separated heat dissipates during travel path inside the module. I am building WC system that will have two cpu water loops and between them PT.

How they built TEC systems for pc was just wrong

[njordan0526]

My method wouldn't be to attach the peltier unit directly to the CPU (Sorry if I wasn't clear about that in the original post). The unit would be use to transfer heat using a waterblock FROM the water TO a heatsink on or near the exterior of the case.

 

 

[njordan0526]

I just got a better laptop that will actually run a CAD program so I'll make a sketch and get it posted so you can see what I'm talking about.