Return to Level1Techs.com

Why are Peltier CPU cooler not more popular?


#1

Are they cooler than basic-air cooling?
Has anyone used them?
Are there any downsides?
Whats holding them back?
Has anyone used them?

AGT3_1_20180708726953040


#2

Have a friend who has played around with them. They are actually less effective than air cooling. I don’t know the specifics of why, I’ll have to ask him about it. But I imagine it’s something along the lines of them being better at lowering things below room temperature, and worse at actively keeping a heat-source close to room temperature.


#3

The heat still has to go somewhere. You have to have a heatsink on the hot side to dissipate the extra heat from the cold side, so there’s really no point. @gigabusterEXE used them, or tried to at least, he could tell you more.


#4

It’s inefficient because the energy required to drive the reaction is triple or more the energy cooling capability being output.

der8auer recently did a video on it:


#5

Cooler Master had a commercial peltier CPU cooler a while back. It was crap, mostly due to how CM designed it.


#6

Seems like these could be useful in niche applications that need extensive cooling in tight spaces for short periods of time.

Overclocking quad CPUs in a 2U, then filling a rack with them? I dunno.

Pretty neat either way


#7

Using a peltier for getting the CPU sub-ambient requires loads of energy. They are very inefficent.

However, if you were to use them as thermo-electric generators, you could make use of some of the heat. Power some case lighting or small fan to keep the VRM-temps in check :stuck_out_tongue:


#8

I forgot what aerospace camera manufacturer it was, but they used those types of cooling things in their cameras.


#9

most peltier coolers mean you need 2 lc cooling systems and the liquids uses are really bad for normal cooling. And most people never will need sub dew point cooling. very few systems will take advantage of that level of cooling. its fun but never needed.
having used them and i am going to use them on a version of my “flex tape” builds so if you are interested i will be doing a build with em.


#10


if you want air cooling but, personally 2 loops transfering heat in a antifreeze bath would be better similar to LTT and the ac unit for tr2


#11

You need more than double the power to cool as good and even more if you want sub ambient temps
For reference I had a time keepimg a 35w gt 1030 subambient with a 65w peltier

Plus to do it right, a lot of the times you need expensive custom mounts and heatsinks and it’s messy because you have to slather everything with dielectric shit like art eraser or petroleum jelly

You need a lot of time and resources
Also a REALLY big power supply since it takes double the juice of whatever you’re cooling


#12

I’m actually working on a pelt loop now, there’s a lot to it, so I’ll try to be brief.

Think of pelts (TECs) as heat pumps, rather than something that cools something. It moves a certain amount of heat from one place (the cold side) to another (the hot side). That said, just like a pump, it takes power to run, the more heat you move, the more power it takes. This is where the inefficiency comes from. Lets say you have 50w to cool. A tower cooler can do that just by sitting on the heat load, it takes no power, creates no extra heat, very efficient. Lets take a custom loop for example (an AIO is similar), the pump will actually create a small amount of heat in moving the water through the loop, as well as require power to move the water, so less efficient, despite cooling better. TECs require pumps to move the fluid, but more importantly power to run the TECs themselves, which in turn creates a LARGE amount of waste heat, least efficient. Check this playlist out to learn a lot more: Video

With that in mind, lets talk about what TECs can and cant do. Each can move a certain amount of heat, but that’s a simple way of looking at it. Lets presume we have a 200w unit, that is, at its max power it can remove 200w of heat. At that heat load, however, the temperature difference in the sides will be 0c. If the hot side is 50c, the cold side will be too. Now, lets say that at a 100w load the delta is 20c. That means that if the hot side is 50c, now the cold side is 30c. This is detailed in the specifications sheets specific to that unit, so it’s important to pay attention. This is one of the considerations you need to make for a loop, but we’ll come back to that.

The next aspect to consider is cooling. The hot side cooling needs to be able to handle not only the heat from the CPU (or other source) but also the heat from the TEC. Lets use our imaginary TEC again, and say we have a 100w load from the CPU, and were using the TEC at full power. Lets add another stipulation, that at 200w it’s drawing 12v and 20A, for a total power draw of 240w. The cooler needs to be able to handle a 340w load. In this case, with a 100w load, the delta between the sides is 20c, and that wont change depending on how hot/cold either side gets (that delta is based on the heat being moved by the TEC). That means that if your cooler can keep up, and lets the hot side get to 60c, then the cold side is only 40c. For puling liquid/CPUs down below ~23c, that obviously wont work. Lets assume now that the hot side cooling keeps that hot side at 30c, in that case the cold side would be sitting at 10c, in which case performance will exceed typical watercooling. This should highlight the inefficiency argument, a 100w CPU can easily be handled by a 120mm AIO, keeping a 340w load at around a ~10c dt over ambient is a much harder task, likely requiring a custom loop of its own. This is another one of the main considerations that needs to be made when looking at a TEC loop, which we will come back to.

The last aspect deals with the design of the cooling. There’s 2 major types of pelt TEC cooling, direct and chiller. Direct is as it sounds, with the TEC smack dab on top of the CPU, whereas the chiller uses blocks to cool the liquid that goes into the CPU block. Direct will perform better, but you have to make sure that the TEC can handle the heat coming into it. If you have a TEC capable of moving 200w with a 0c dt between sides, and throw a 250w load on it, you’re going to have a very bad time. So, for direct you have to be sure the TEC can handle the load, while also still having overhead (for reference, that number is typically 2 times the heat, ie. for a 250w load, have a unit that can move 500w of heat). The size of CPUs means that you’ll really only get one, maybe 2 units on there if you’re lucky, and with a custom waterblock, limiting the total amount of TEC power you can put in the loop. However, with chillers there is no such limit. You can have 1, 2, 8, even 200 TECs if you want, you’re only limited by your power draw and space. Chillers are not as effective, as they cool the water that cools the CPU, but allow you to more easily add and build up more cooling. Instead of 1 500w unit, now you can use 3 200w units.

Revisiting the considerations:
1 - make sure you have enough “TEC power” that is, that your TECs can pump enough heat out. Typically 2 times the load is the goal, as this should provide a decent enough delta between the two sides.
2 - make sure your cooling can handle the load of both the component and the TECs. This means that for a 250w CPU, like we have been considering, and then 3 200w TECs (drawing their hypothetical 240w each) you need to be able to cool ~1000w with the lowest delta possible.
3 - more implied than specifically mentioned, always overestimate. There will be inefficient parts everywhere, TECs do not scale perfectly, they do NOT typically like to run below 40% or above 80% (far less efficient), there is thermal resistance in waterblocks (for pulling heat off of CPU or cooling water in chiller blocks), and so on. In our case, we should be safe with the estimated 600w of cooling for the 250w load, even up to 300w, and we should plan to be able to dissipate at least 1000w with a ~10c dt, mandating 2 360mm rads with push pull fans, or 3 360s and a more reasonable set up.
After all that, it becomes clear why TECs are not more popular. They require a lot of power, they require a separate cooling loop, with a ludicrous amount of cooling. There is far more to consider than a typical water cooling loop. The cost can quickly get out of hand. That said, compared to other sub ambient choices, they have advantages. The only other sub ambient cooling options that come to mind are stand alone chillers and phase change, which are both loud, bulky, expensive, and feature their own list of issues. The other issue to consider, as mentioned by some earlier in the thread is condensation. When going below ambient, you run the risk of condensation, which in turn can fry your board. The only way around this is to cover that nice new expensive motherboard in liquid electrical tape, Vaseline, and many other horrible, unremovable things. There are ways to control this, such as varrying the voltage coming to the peltiers depending on the temperature, temperature controllers with a hysteresis band, and so on, but realistically that isn’t always a safe bet.

tl;dr
They’re extremely hot, require 3-4 times the cooling to be more efficient than water, can require custom hardware, their own power supply, and are far more complicated than even a custom watercooling loop.

There’s of course a lot more to it than than, but I wanted to try to keep things short and simple. Let me know if you have any questions, about my info, or in general. I’ll try to explain as best I can. :slight_smile:


#13

Very popular around diy brazilian groups a little while ago, before AIOs took the market by storm