Question Cooling CPU's by thermoelectric cooler or TEC

[iamgenius]

I was reading about generating electricity by converting heat into electricity using thermometric generators:

Thermoelectric generator - Wikipedia

en.wikipedia.org

and I came across thermoelectric coolers which uses the same concept but in reverse. I saw them being used as coolers for even cpus:

Thermoelectric cooling - Wikipedia

en.wikipedia.org

10 FERROTEC 9500/127/085 B THERMOELECTRIC PELTIER COOLER 12V 8.5 A TEC1-12708 | eBay

Find many great new & used options and get the best deals for 10 FERROTEC 9500/127/085 B THERMOELECTRIC PELTIER COOLER 12V 8.5 A TEC1-12708 at the best online prices at eBay! Free shipping for many products!

www.ebay.com

I just thought it is interesting. I have been overclocking cpus since probably 2002 and never saw this before. Is it relevant? Or is it something to consider? I don't think it will outperform watercooling.

https://www.ebay.com/itm/Titan-AMAN...612621?hash=item4d92dfa70d:g:TGcAAOSw9N1V0tg2


Anybody who used them?

Something to discuss !

 

[lightmanek]

I used them, but they are not good for high wattage CPUs so their usefulness is limited to old processors up to 50W or so. You do pretty much need watercooling to cool peltiers as they double heat output from cooled component.
I had fun playing with them, but as a long term CPU radiator they're not good.

 

[AdamK47]

Why it's bad...

 

[DrMrLordX]

TECs have a bad rep thanks to various eBay TECs that have been known to fail at inopportune times. I think there's a solid argument to be made for TECs when your heat loads are below 200W - and even for that, you would have to size a massive TEC. The other problem is that despite various lab advancements in materials science, the common availability of next-gen TEC products hasn't happened yet. In other words, you can expect to see more bismuth telluride than anything else, and that's been the standard for awhile.

I think there is an argument for next-gen TECs on 7nm-and-smaller CPUs. It's seriously getting hard to dissipate massive amounts of heat from anything except Threadripper-type monstrosities, even if you attempt to overclock. I don't think my 3900X has ever pulled more than 190W. An eBay bismuth telluride TEC could be sized for that application, and I could even cool it with my custom wc loop. The only thing stopping me is possible condensation around the cold side of the TEC. Well that and having to resize my PSU for the 320W+ TEC I would have to size for the application.

 

[lightmanek]

TECs have a bad rep thanks to various eBay TECs that have been known to fail at inopportune times. I think there's a solid argument to be made for TECs when your heat loads are below 200W - and even for that, you would have to size a massive TEC. The other problem is that despite various lab advancements in materials science, the common availability of next-gen TEC products hasn't happened yet. In other words, you can expect to see more bismuth telluride than anything else, and that's been the standard for awhile.

I think there is an argument for next-gen TECs on 7nm-and-smaller CPUs. It's seriously getting hard to dissipate massive amounts of heat from anything except Threadripper-type monstrosities, even if you attempt to overclock. I don't think my 3900X has ever pulled more than 190W. An eBay bismuth telluride TEC could be sized for that application, and I could even cool it with my custom wc loop. The only thing stopping me is possible condensation around the cold side of the TEC. Well that and having to resize my PSU for the 320W+ TEC I would have to size for the application.

Yes, to cool Ryzen you would have to get the most area efficient TEC available as otherwise area of 200W tec is usually greater that the whole heatspreader of a modern CPU. Better application would be to use them as a heat exchanger for liquid cooler to keep water at a low temperature. You then avoid problems with condensation as controlling large liquid volume at a semi constant temperature is easier than with CPU which can fluctuate heat output in seconds. Ideally, you would have to have a proper controller for TEC elements reading temperature form cold side and adjusting current to keep them from getting too cold.

As I sad, it was nice to play with them back in a day, bit there are easier ways to get your CPU cold enough and way way cheaper

 

[DrMrLordX]

Better application would be to use them as a heat exchanger for liquid cooler to keep water at a low temperature.

Yes. You're making a DiY chiller. You could shoot for decent sub-ambients if you engineer around condensation though. I would have to see some significant advancements in TEC materials before I would even think about something at that level.

 

[iamgenius]

Thanks for the great info guys. I was actually reading about it so that I can demonstrate how heat can be converted into electricity. Now that I think about it, I don't think it will be efficient enough to be really worth it as an electricity generator. I was considering an application where you can use excess heat for extra free energy. There are some papers that talk about new materials but it didn't really happen yet.

 

[LightningZ71]

Back when my dad was still working somewhere that there was a machine shop, I did some experiments with a rather modest TEC setup on some older CPUs. I learned that you had to do a lot of active monitoring of the TECs as the failure modes could be quite undesirable. You could easily do sub-ambient with them, with all the issues that you face there, but, it's one thing to get the heat off of the CPU, its another to get the heat out of the other end of the TECs. I did find a setup that worked quite well, using what I called a thermal surface area expander, which is just essentially a solid metal chunk that was shaped kind of like a square funnel, with the smaller end on the CPU, and the larger end on top with multiple smaller TECs attached to it. Then, a hot plate with embedded tubes in it that I had plumbed to an external radiator with a sandwich of fans on it. It was FAR, FAR from pretty, using what was essentially a radiator that you'd use for the transmission fluid cooler on a car, with a sandwich of electric fans on either side of it. The water pump was repurposed from an automotive application. It worked. It was loud, big, power hungry, and not exactly a clean or neat installation, but, it did get me another 20% more overclock on the processors.

I later performed a similar experiment with a dorm fridge. It was MUCH quieter. However, dorm fridges don't have the thermal capacity that you'd think. They also weren't made to run the kinds of constant thermal loads that cooling a CPU put on them, and you quickly wear out the compressor. Lesson learned, use a continual duty class chiller...

 

[LightningZ71]

I looked at what you are referring to in an automotive application while I was in college. The amount of electricity that you could generate from waste heat was, at the time and with the TECs available then, not worth the extra weight that the system would require. We basically concluded that efficiency would have to be roughly ten times what it was for it to make sense, and the vehicle would also have already had to have been a hybrid to overcome the weight penalty required to convert the electricity to motive force. It might make a bit more sense with modern hybrids, but they're so good at running with the engine off that they wouldn't generate enough waste heat to matter. Even if you included using the waste heat from the electric motors, the charging systems, and the discharge of the battery pack, you still wouldn't get enough to make it make sense with current, commercially feasible TECs.