My pelt experience

[Phrenetical]

Picked up two very cheap 110W pelts off ebay, basically i got them to play with them and test out making a small AC unit that could fit into two 5.25 drive bays

This is what i found. Pelts are very crap in general.

I did a fair bit of testing last night with a 110W @ 12v made my own little passthrough power connector for it.

And hooked the hot side up to a stock intel cooler with an 80m fan blowing air away from pelt. used AS5 to stick it to HS, and turned it on, ran fine but seriously even a 110W pelt drawing 10 amps or so, doesn't even get cold enough to buildl up ice on the cold side.

I was able to leave it running for ages, like 30 minutes and the cold side was just cold enough to build up condensation, which you could feel when touching it.

What i found amazing, was that the cold side was pretty cold, but the HEatsink was burning up, i recon it was maintaining 35 - 40 degree's the whole stock intel hsf was very hot even after only a few minutes of running.
Without the fan, i couldnt run it for more then 3 - 5 minutes without the cold side starting to warm up.


I then thought ok maybe this HSF setup sucks, so i grabbed my old Hyper 6 and hooked up the fan to it, hooked this up to the pelt, and left cold side bare. Still no ice, bit colder to the touch, Heatsink wasn't as hot cause it can dissipate the heat better.



Step2.

attach heatsinks of various sizes to cold side with a fan, to simulate AC.

Tried with a second intel HS barely even got cold.

Tried with a small intel pentium 233 HSF small copper heatsink with 40m fan in middle. The sink got pretty cold but the air once it gets accellerated seems to lose it coolness and was no better then the fan not attached to the cold sink.



So even without a CPU under the pelt, a 110W tec was producing enough heat to warm up a hyper 6, and not enough coolness to even chill a small heatsink with a 40m fan.


Long story cut short, tec's are extremely innefecient.


I thought a 110W tec would at least build up ice on the cold side, guess i was incorrect.

ambient temps were probably 20-23C in my room.

 

[jivetrky]

I have NO clue about TECs...but I wonder if you stacked them up....would the bottom one then ice up? seems like the problem is just that the heat isn't getting away fast enough for the cold side to get sub freezing

 

[GigaForce310]

Peltiers are fun little devices. They are very inefficient except when used properly. Even then they are still pretty inefficient compared to phase change.

Your experiment showed exactly why pelts are not good on air at all. The issue is that a 110 watt pelt puts out 110 watts plus any load is put into the device. The simple fact that the cold side's temperature is below ambient means that there is a load put on the heatsink. Heat from the air transfers onto the plate.

Another issue is that typical CPU heatsinks can't handle the 150-200 watts that your expecting it to. Even large ones don't perform too well. They reach their limit and their temperature rises. The idea of a pelt running full voltage is that if the temperature differences between plates is 0, then it's tranfering the full cooling wattage. If it's about 67-69C then the plates transfer no heat. When the heatsink warms up, then the minimum possible cold side temperature is increased.

Since your pumping in 110 watts of heat, the heatsink will warm up regardless of how little load you put on the cold side. A hot heatsink kills cold side potential. Say you have 50C difference between plates and the cold sides happens to be 5C. Your heatsink is 55C. Say you are able to lower the temp of the heatsink to 45C. Then you will form frost on your cold side. Frost insulates and your delta temperature will increase since there is less load. The opposite is true, since there is no frost, they heat is readily tranfered to your heatsink.

Water cooling is the only way to cool a CPU with a pelt. The reason is that a large radiator and good heat transfer of water is able to handle both the CPU and the pelt heat and keeps the temperatures closer to ambient. Given a good water cooling system with 30 degree water block temp and the same 50C delta temp, then the cold side would be -20 degrees.

Because the delta temperature derates linerally as the heat load increases, you need much more Pelt wattage than CPU wattage to get a good delta temperature. The rule of thumb is that a good pelt setup will have 2-3 times the wattage over the CPU wattage. For example, if your processor puts out 100 watts then you would want a pelt setup with 200-300 watts total capacity. This means that you would have 300-400 watts to deal with in the water cooling.

To get good results with pelts, you need a lot of electricity and very powerfull water cooling.

 

[GigaForce310]

I have NO clue about TECs...but I wonder if you stacked them up....would the bottom one then ice up? seems like the problem is just that the heat isn't getting away fast enough for the cold side to get sub freezing

That works, but only on very small input power setups. Given a 110 watt pelt and a 60 watt heat load on it means that the second will have to take in 160 watts. Say the second happens to be a 226 watt pelt. Then the heatsink would have to handle 386 watts.

Next take a 100 watt CPU with a 226 watt pelt. That is 326 watts... 2 437 watts pelts in parrellel will do it. That mean you now have 1200 watts of heat to cool. You would need a car radiator to handle all of that heat.

A good single phase change setup costs less, use much less electricity, gets better temperatures, and would be better looking too.

 

[Phrenetical]

Yes i totally understand the math and physics behind the tec's, but i had to test it out for myself, plus i wanted to test other things and it was a good experience and taught me alot.

I posted up my results to highlight how badly an air cooled tec performs, as the heatsinks just can't transfer the heat quickly enough, and it highlights how very very innefecient they are. So i hope all the people who post about air cooling a tec read this and take some knowledge away.


so water cooling is really the only option with pelts, and even then you need much more then just a tec, its very complicated and your right they just really are not worth it, may as well go to phase, it looks difficult but i guarantee it's no worse then trying to use a tec setup.


also ii'm using 110 W tecs because the older PSU's i have for testing purposes such as this supply about 15 - 17a on the 12v, and using a 176W or 220W would probably fry them.

 

[Nautica]

I have NO clue about TECs...but I wonder if you stacked them up....would the bottom one then ice up? seems like the problem is just that the heat isn't getting away fast enough for the cold side to get sub freezing

wondered the same thing about stacking the pelts and thought that might be a idea but like giga mentioned youd then have a tremendous amount of heat to deal with in the end so after some research I just decided to get a mach II

fun idea though I do recall coming across a picture of a cpu heatsink with alot of pelts cold sides touching it then those cold sides being water cooled which was pretty wild Ill have to see if I cant find a link for it

 

[praxis1452]

I don't think it's the fact that you it's not powerful enough but that you couldn't cool it well enough...

 

[RIG RIDER]

I don't think it's the fact that you it's not powerful enough but that you couldn't cool it well enough...

?

 

[jivetrky]

I think he's saying that if there was better cooling on the thing that it may have gotten coldside may have gotten cold enough to ice up.

 

[praxis1452]

?

He wasn't actively cooling anything but the air and air is a very poor conductor so therefore its basically like there's almost no heatload so it'd be pretty close to the maximum Delta T correct me if I'm wrong so therefore if you just used an even smaller pelt with a better heatsink it would have frosted.

 

[jivetrky]

Well it was being actively cooled by the Stock HSF, but that still isn't really good enough for TECs

 

[greenmaji]

How much preasure did you apply between the heat sink and the cold plate?

 

[Phrenetical]

Tied down on each side with a thin wire loop, then twisted loop ends together to create force.

So there was definently enough force to achieve proper transfer. (This is on stock HSF)

On the Hyper 6 i just had them sitting on each other, so pressure was weight of sinks, but this was no good anyway because it proved the amount of Heatsinks you'd need to come close to a nice AC unit just wouldnt fit anywhere in your case

lol...



And like i said, a 110W is feasible to plug into your sdtandard PSU's, anything above that your looking at a re-investment in a much beefier PSU or a second one.

So i proved that the feasibility of tecs is total crap. NOt worth the effort on AIR in any way shape or form.


I will post a few pics today, of the setup, and you can make your own judgements.

 

[ziggo0]

Haha, I bet that was fun, did you burn the pelts out?

 

[greenmaji]

Tied down on each side with a thin wire loop, then twisted loop ends together to create force.

So there was definently enough force to achieve proper transfer. (This is on stock HSF)

You might want to head over to procooling's articals section and read the last few TEC articals.

 

[praxis1452]

Tied down on each side with a thin wire loop, then twisted loop ends together to create force.

So there was definently enough force to achieve proper transfer. (This is on stock HSF)

On the Hyper 6 i just had them sitting on each other, so pressure was weight of sinks, but this was no good anyway because it proved the amount of Heatsinks you'd need to come close to a nice AC unit just wouldnt fit anywhere in your case

lol...



And like i said, a 110W is feasible to plug into your sdtandard PSU's, anything above that your looking at a re-investment in a much beefier PSU or a second one.

So i proved that the feasibility of tecs is total crap. NOt worth the effort on AIR in any way shape or form.


I will post a few pics today, of the setup, and you can make your own judgements.

I really don't think that's enough force.... I think the reccomended amount of force is a couple hundred lbs

 

[Pf.Farnsworth]

This is what i found. Pelts are very crap in general.

If you read the forum a little you would have found that out prior to your experiment. Pelts are garbage when it comes to cooling and are not a reasonable solution by any means. There are simply to many safer and more efficient ways to do the job and do it much better.

 

[ls7corvete]

Read what people are telling you and you will get better results.

Of coarse if you just want to add to the bashing of TECs from people who do not know what they are doing and do not care to listen then you have succeded.

Pelts work great, the knowledge to use them right is not there. Period.

 

[greenmaji]

Reaserching projects FTW!

Seriously, this would have turned out better.