Thermoelectric CPU cooler?

[Clechay]

I was just thinking that using thermoelectric heat pump to cool CPU would be an interesting continuation to "PC Build in a Fridge" video and I've never seen any attempt to do such a thing. What I find interesting about this idea is that if it works it would increase rate of heat dissipation without generating any sound.

By thermoelectric heat pump I mean Peltier cooler that is well described on Wikipedia (https://en.wikipedia.org/wiki/Thermoelectric_cooling). Placing such a thing in between CPU and it's radiator should decrease temperature of CPU and increase temperature of radiator. While it cannot be as effective as water cooling, it sill can be a huge thing for extremely quiet passively cooled PCs because it's a way to dispatch more heat without any fans or other moving elements.

All that I expect to be necessary is putting thermoelectric heat pump between CPU and CPU cooler. The easiest way to power it is using external power source but with some research, match and little amount of work it should be possible to use main power supply.

Does it make any sense? Is it possible for thermoelectric heat pump to work effectively enough? Is there any chance of LMG video with that kind of stuff?

Sorry for my English, I'm not a native English speaker

[Moonzy]

well from my understanding, you'll need a lot of wattage to power this thing to move heat away from the cpu, which means the heatsink have to remove the heat from this plate + the cpu

which will worsen the condition you have to begin with?

 

but yea, interesting tech `-`

[Windows7ge]

Peltier plates can operate off of 12V and computer power supplies have more than sufficient amps on the 12V rail to power it. A concern will be if moisture forms around the peltier but in theory if the heatsink can dissipate the heat away from the hot side sufficiently then it might work. I just ordered a 72W peltier. I can mount it to my 3930K CPU and cool the hot side with an H100i liquid cooler. If I have a way to prevent moisture build up (Provided the peltier actually remains below ambient) I'm willing to test this to see if a 72W peltier can keep a 3930k cool under desktop load. I don't expect good results under CPU stress.

[suchamoneypit]

1 minute ago, Windows7ge said:

Peltier plates can operate off of 12V and computer power supplies have more than sufficient amps on the 12V rail to power it. A concern will be if moisture forms around the peltier but in theory if the heatsink can dissipate the heat away from the hot side sufficiently then it might work. I just ordered a 72W peltier. I can mount it to my 3930K CPU and cool the hot side with an H100i liquid cooler. If I have a way to prevent moisture build up (Provided the peltier actually remains below ambient) I'm willing to test this to see if a 72W peltier can keep a 3930k cool under desktop load. I don't expect good results under CPU stress.

That would definitely be interesting to see though, regardless if it works or not.

[Windows7ge]

Just now, suchamoneypit said:

That would definitely be interesting to see though, regardless if it works or not.

I'll keep a link to this thread and post my results when I have them. The peltier is coming from over seas so it's going to be a while before I can perform the experiment.

[DragonTamer1]

Peltier coolers draw massive amounts of power to operate and have a very high thermal dissipation on their hot side. The heat of the hot side would be the peltiers own thermal output (based on it's power draw and efficiency) plus the heat of the CPU that is being moved.

 

The thermal energy from the CPU is not being canceled, it is simply being moved from one place to another. In this regard a passive solution is impossible because the amount of power being dissipated by the system has increased not decreased.

 

When done properly it can indeed lower the CPU temps, however the overall heat that your system puts out has gone up as well and by quite a lot. In order to constantly cool the CPU the peltier will be running almost constantly and if my memory serves they have a power dissipation between 40 and 60 watts.

[porina]

1 minute ago, Windows7ge said:

Peltier plates can operate off of 12V and computer power supplies have more than sufficient amps on the 12V rail to power it. A concern will be if moisture forms around the peltier but in theory if the heatsink can dissipate the heat away from the hot side sufficiently then it might work. I just ordered a 72W peltier. I can mount it to my 3930K CPU and cool the hot side with an H100i liquid cooler. If I have a way to prevent moisture build up (Provided the peltier actually remains below ambient) I'm willing to test this to see if a 72W peltier can keep a 3930k cool under desktop load. I don't expect good results under CPU stress.

The biggest problem here is they need massive amounts of power to create a temperature differential. Then you have to get rid of both the original heat AND the heat from the peltier.

 

The 3930k is rated at 130W TDP without overclocking. Stick a 72W peltier on it, and you're looking at shifting 200W. That should give the water cooling a test. The thing to understand is the peltier creates a temperature differential across it. You need to keep the hot side cool, to lower the cool side. It isn't good if the hot side goes up! I suspect a 72W model will not be powerful enough to do any good in this case.

 

I can't remember exactly when or the details any more, but I did try it myself in the distant past. I don't remember the outcome being useful and I sold my peltier onto someone else who wanted to cool an image sensor. That application makes more sense since they're much lower power devices.

[Moonzy]

2 minutes ago, porina said:

and you're looking at shifting 200W.

still lower than an FX9590, so watercooling should do just fine

 

my main concern is how many watts can it move from the cool to hot side? is it the same amount of wattage it's using or... something else

meaning if you wanna cool a 100W CPU, you'll need to input 100W into the plate? (assuming 100% efficiency) or something else? so a total of 200W on the hot side? or im wrong

[porina]

1 minute ago, Moonzy said:

still lower than an FX9590, so watercooling should do just fine

 

my main concern is how many watts can it move from the cool to hot side? is it the same amount of wattage it's using or... something else

meaning if you wanna cool a 100W CPU, you'll need to input 100W into the plate? (assuming 100% efficiency) or something else? so a total of 200W on the hot side? or im wrong

I have to admit it is too long ago since I last looked at this. My memory suggests that people were buying peltiers rated at a higher power than the CPUs, but it is so long ago I can't remember the reasoning.

 

I suppose another way to look at it is, if it was usefully effective, why aren't there more products off the shelf for doing it? The 24/7 sub-zero overclockers went to phase change coolers and everyone else made do with air/water.

[Weak1ings]

Speaking of all of these watts, It might make a custom liquid cooling setup practical with 3 or more radiators since it can dissipate that heat over a longer "chain" before coming back for more.

[Windows7ge]

6 hours ago, porina said:

The biggest problem here is they need massive amounts of power to create a temperature differential. Then you have to get rid of both the original heat AND the heat from the peltier.

 

The 3930k is rated at 130W TDP without overclocking. Stick a 72W peltier on it, and you're looking at shifting 200W. That should give the water cooling a test. The thing to understand is the peltier creates a temperature differential across it. You need to keep the hot side cool, to lower the cool side. It isn't good if the hot side goes up! I suspect a 72W model will not be powerful enough to do any good in this case.

 

I can't remember exactly when or the details any more, but I did try it myself in the distant past. I don't remember the outcome being useful and I sold my peltier onto someone else who wanted to cool an image sensor. That application makes more sense since they're much lower power devices.

As I stated previously, I'll see what happens when I idle at the desktop, I predict some degree below ambient temperature (not much) when the CPU is underclocked and undervolted but I also predict when I pin the CPU at 100% load it will probably fail.

[Stefan1024]

The pelier needs to have about double the power of the CPU, so about 260 watts (without OC). At gthe rated wattage the temperature difference is 0C.

I thougth about peltiers when designing my fanless PCs (look in my signature), but the 10 to 20 degrees C gain isn't worth the tripled heat output.

I ended up using peltiers but not to cool but not to cool. I used them as a thermoelectric generator to use the heat output of the GPUs.

 

Peltiers are nice to play with and for a test go aheard, but it is a very bad solution for 24/7 operation with some serious drawbacks.

[Windows7ge]

51 minutes ago, Stefan1024 said:

Peltiers are nice to play with and for a test go aheard, but it is a very bad solution for 24/7 operation with some serious drawbacks.

Even if it does go well I plan to disassemble it immediately after. The peltier I ordered has another purpose, this is just for the sake of the experiment. Honestly, I don't expect good results once the CPU heats up but I wanna see what'll happen anyway. Just have to address the potential condensation issue somehow...not sure how I'm gonna do that.

[Stefan1024]

8 hours ago, Windows7ge said:

Even if it does go well I plan to disassemble it immediately after. The peltier I ordered has another purpose, this is just for the sake of the experiment. Honestly, I don't expect good results once the CPU heats up but I wanna see what'll happen anyway. Just have to address the potential condensation issue somehow...not sure how I'm gonna do that.

There are a lot of guides how to do this online, but the methodes are usually quite messy. But i hink you won't be able to cool it below 10 C so there shouldn't be to much consensation.

[Clechay]

22 hours ago, porina said:

The biggest problem here is they need massive amounts of power to create a temperature differential. Then you have to get rid of both the original heat AND the heat from the peltier.

 

The 3930k is rated at 130W TDP without overclocking. Stick a 72W peltier on it, and you're looking at shifting 200W. That should give the water cooling a test. The thing to understand is the peltier creates a temperature differential across it. You need to keep the hot side cool, to lower the cool side. It isn't good if the hot side goes up! I suspect a 72W model will not be powerful enough to do any good in this case.

 

I can't remember exactly when or the details any more, but I did try it myself in the distant past. I don't remember the outcome being useful and I sold my peltier onto someone else who wanted to cool an image sensor. That application makes more sense since they're much lower power devices.

We all know that peltier can only transfer heat from one side to another, not  make it disappear. However the hotter radiator is the more heat it's gonna dispatch to air, the hotter air around radiator becomes and finally airflow generated in passively cooled system by heating up air increases. Situation where we could actively transfer heat from CPU to radiator and make radiator hotter than CPU would be prefect, the only problem is how much heat is generated in transfer process. I'm afraid I misunderstood the way peltier works and miscalculated amount of heat it's gonna create. Doubling amount of heat that needs to be dispatched is to much for any radiator.

[Shoggy]

I have the feeling this idea is the same like the immersed oil cooling: once in a while it always comes back

 

Like many already stated, it does not make much sense and comes with several problems. If that would be a good thing you can be sure it would be already on sale but it is not.

 

We also already played around with this stuff several years ago.

[Windows7ge]

Unfortunately my order for the 72W peltier has been cancelled. They're out of stock so the experiment will have to be postponed indefinitely. I have a 42.5W peltier also on the way but that stands no chance of cooling a 3930K...or could it at least handle it at desktop load?...I might go through with it anyways. Absolute max the Corsair H100i will have to dissipate 172.5W. Still quite a bit of power. Don't know if 42.5W could fight a 3930K and keep it under control even at idle. What do you guys think?

[Windows7ge]

On 5/2/2016 at 6:11 PM, suchamoneypit said:

That would definitely be interesting to see though, regardless if it works or not.

Peltier came in, tested it, complete failure lol. It seems the peltiers aren't so good at making things cold when what they're trying to cool radiate heat. At least not a 42.5W (The 72W I received an email saying they were out of stock but I can't say it would have fared much better)

[suchamoneypit]

15 minutes ago, Windows7ge said:

Peltier came in, tested it, complete failure lol. It seems the peltiers aren't so good at making things cold when what they're trying to cool radiate heat. At least not a 42.5W (The 72W I received an email saying they were out of stock but I can't say it would have fared much better)

thats a shame, thanks for coming back and letting me know the results though!

[Windows7ge]

2 hours ago, suchamoneypit said:

thats a shame, thanks for coming back and letting me know the results though!

I've got one more thing to try. I have this extremely low power 25W CPU on a Mini-ITX board. Functionality wise I'd gain nothing by making it colder but for the sake of seeing if 42.5W can cool something putting out 25W I'm willing to give it a shot. Stefan1024 mentioned that the peltier needs to be about twice as strong as the CPU so this has a much higher chance of succeeding. I tested the peltier, when nothing is acting on it it goes down to -7.8°C. So with 25W of heat trying to be dissipated it should be a fair bit higher but if it stays below ambient I'll call it a success.

[Windows7ge]

20 hours ago, suchamoneypit said:

thats a shame, thanks for coming back and letting me know the results though!

I've got good news and bad new. Good news is the peltier was more than strong enough to cool a 25W CPU, bad news is it cooled it so much so that it dropped below zero of which this motherboard apparently doesn't like negative number so all it gave me was : N/A (As seen below, I watched the temp drop from about 20°C down to N/A so this isn't just some error with my hardware, I even turned off the peltier and watched the temperature slowly climb back up but it's safe to assume the temperature is somewhere in between 0°C - (-7.8)°C) if the CPU's FSB and multiplier weren't locked I'd overclock it and see how far it can go but that's just not an option. Seems all that thermal paste didn't go to waste after all, the not very impressive sized heatsink it's attached to I have to say is quite hot so I don't think I'll be trying again on my old 3930K. Worried I might break the peltier.

[thekeemo]

Below ambient = condensation

That is why you need a controller

Really annoyed that I have one that is made to handle 250w CPUs with a heatsink built for it but cant use it because:

1. Pump is dead

2. Im a sissy and dont want to destroy my system