Peltier (TEC) + Liquid Cooling Project

[Bobnova]

Oh they're not good at all for CPU cooling, unless you're in Marketing and trying to sell them. They're horrendously inefficient, as mentioned at least a couple times on the first page of this thread.

The only way TECs are efficient is in space required, for their cooling power they are very very small.

To get more effective at moving energy around than water you're looking at liquid ammonia, something you don't want to muck with.
Water beats all the various liquid coolants out there.

The pump is your enemy, it's putting as much heat into the loop as the TEC is pulling out.
A beefier TEC will solve that problem, leaving just the question of whether the cooling for the TEC can deal with the new improved higher power TEC.

 

[myom]

Would this TEC show significant improvement over the current TEC?

Reported Qmax of 138.6W compared to the original's reported Qmax of 58-65W. Does that allow it to remove ~2x the amount of heat from the loop (as well as generate twice as much heat on the hot side)?

 

[Bobnova]

Probably. It depends on whether the specs are real or made up on the spot.
If they're real, you'll be able to move significantly more heat, at the cost of producing significantly more heat.

 

[myom]

Cool, I'll give it a shot. Any suggestions for a CPU heatsink+fan? Or would you still suggest going with a 2nd loop on the hot side (+ radiator)

 

[Bobnova]

Big.
A TRUE, a Noctua D14/D15, or a Phanteks big twin tower thing.
Something in that class, anyway.

Now that said, you probably could use a stock AMD PhII 125/140w TDP CPU heatsink/fan. I used one with a ~136w TEC and was able to freeze water, if slowly.

 

[RnRollie]

for what its worth... (dont know if they are any good) these guys seem to offer some "complete" solutions
Controller: http://www.quick-cool-shop.de/en/peltier-controller/peltier-contoller-qc-pc-c01c.html
Display piggyback: http://www.quick-cool-shop.de/en/peltier-controller/display-qc-pc-d100.html
And the TECs of course: http://www.quick-cool-shop.de/en/peltierelements/one-stage-peltier-module/

 

[myom]

This giant setup is still not working.. The water block at the TEC+heatsink junction (on the bottom right) only gets cold when the pump is off though (same issue as last time). So it seems the pump is generating too much heat, but it only consumes 5W (pump specs). Pumping up the TEC to 15-17V doesn't work. Also, leaving the pump on by itself (TEC off) doesn't really "heat up" the water, so I don't understand how it can counteract the cooling power of the TEC so well.

I can't seem to find a pump that consumes any less energy either so... I'm stumped. Any ideas?

 

[Bobnova]

How is the waterblock being held to the TEC and heatsink? They need to be pressed together quite tightly with a decent TIM.

 

[myom]

The waterblock + TEC + heatsink are held together by an arctic silver thermal adhesive paste

 

[Silver_Pharaoh]

The waterblock + TEC + heatsink are held together by an arctic silver thermal adhesive paste

Try a bunch of tie straps on it.

Like, wrap then around the TEC and the heatsink, top to botton style.

 

[myom]

I don't think the problem is the contact point - when the pump is off, the waterblock gets quite cold.

 

[Bobnova]

The waterblock + TEC + heatsink are held together by an arctic silver thermal adhesive paste

That's probably a large part of your problem. That stuff isn't up to high load applications.
You need some sort of clamping arrangement to actively press them together, and a non-adhesive thermal paste of some sort.

 

[myom]

Is that really the problem though? The water block directly in contact with the TEC becomes quite cold in a short amount of time when the pump is off. When the pump is on, however, not so much. Perhaps the block would get even colder even faster if I used nonadhesive paste?

Anyways I was able to detach the TEC from the heatsink, but probably won't be able to detach the TEC from the waterblock, so I'll apply some nonadhesive paste to the heatsink-TEC side at least. (It took quite a bit of effort! Isopropyl alc, freezing, wedging a blade into the tiniest crevice and hammering it in... it barely popped off!)

 

[Bobnova]

From what I've seen of people using the adhesive on CPU/Heatsink interfaces, yes.
The CPUs overheat quickly, and that's about the same sort of load you're looking at with the TEC.
That stuff is designed for holding RAM heatsinks on and that sort of thing, RAM chips generate a fraction of a watt rather than 100+w. Watts/surface area wise the TEC is >10x the watts per square whatever of contact area compared to a fermi estimated RAM chip.

 

[myom]

How important is it that my TEC is entirely in contact with a heat sink? The surface of the heat sink can only make contact with 80-90% of the TEC. But it's a pretty fancy heat sink with heat pipes and all.

 

[Bobnova]

Fairly, any parts that aren't touching the heatsink are going to get hot. Exactly how hot is hard to say.
Try to center it as best as possible so it's a thin band around the outside that isn't touching.

 

[myom]

I've got the below contraption going (using ArcticSilver thermal paste), and it doesn't seem any better than when I used the thermal adhesive. Again, with the pump off, the water block gets pretty cold, so I guess the biggest issue is with the pump even though its wattage is pretty low... I'm quite lost now

 

[Bobnova]

How long are you letting it run? Is the aluminum base on the heatsink getting warm?

 

[myom]

I've let it run for a good twenty minutes and nothing's happening. The heat pipes get a little warm but the aluminum base does not. I guess that means I can use an even stronger TEC, but it's already blasting at 75W.

 

[Bobnova]

Do have any temperature sensors you can attach to the hot and cold bits?
Numbers would be really useful I think, especially in looking at the water temp.