Peltier Questions

[Winflakes]

Unfortunately, from my trolling of various forums, I assume a lot of people will read the first word in the title and come in to tell me how dumb of an idea and bad it is.
That really isn't needed here as I am not trying to get on a pedistal or anything.

The ultimate goal here is knowledge. I have been doing some research on peltiers and they seem interesting for sure! I was hoping I could get some expert information, preferably in the form of math equations.

Here is my projected set up, please weigh in with numbers or advice on any part.

I would like to fabricate a waterblock that simply pumps the liquid from CPU to block. This would be mounted with a couple, I imagine 2 on each side, peltiers to cool this liquid below ambient, with the goal of getting below 0C if possible. I know condensation will be a problem, but nothing insulation can't fix. And if I DO ruin a computer from condensation... that's fine, at least I made something and tried it out for myself. Moving on!

I know that the Peltiers would need to be kept cool, so I would also fabricate a waterblock for each of the peltier sides that would go through a normal rad system to bring that to ambient.

When I try to calculate the hot side of the Peltier, so I can get to the cold side, the equation I am using asks for a Heatsink efficiency ( Degrees C/Watts). I think I understand what this is, the heatsink setup's ability to get the hot side to ambient. But I am not 100% that that is correct, nor do I know how to calculate this number. First inclination would be, "If my WC setup can cool X watts to 1C above ambient, ambient @ 23, then Loop/amb - 1 = C/W? Anyone have better way to come up with this number, because I am not too comfortable with it.

That was a heck of a first post, but I am really hoping that I can learn something. Heck, if it comes out that a sub-ambient peltier setup is not possible, then fine. I just will not accept this conclusion on the basis of, "Pelts suck, just do WC."

Thank you for your time, and I look forward to hearing your input and contributing to the forum in the future.


TLDR: What are all the equations behind how Peltiers actually work.

 

[Bobnova]

It's possible, but there are issues. Fairly serious ones.
The first is that the pelts will go cold enough to freeze the water in the blocks, blocking flow and causing overheating.
The second is that you'll need a lot of pelt, for a 100w CPU you need at least 200w of pelt to do something genuinely useful.
So for a core2duo it's not too bad.
For a heavily OC'd 2600k you're looking at more like 300-450w of pelt. That means an extra 300-450w of PSU as well of course, and your secondary loop has to be able to deal with the 300-450w of pelt as well as the CPU's heat output.

It's totally doable, don't let anybody tell you that it isn't, but it is an awful lot of work for very little actual performance gain.
As to equations, I wouldn't worry about them as there are too many unknown elements to consider for them to be useful.

Really my first question for you is, why do you want subzero?

It may well be cheaper and easier to mod a window AC unit into a water chiller, and it'll almost certainly give lower temps.

 

[Winflakes]

Yeah, sub zero on 100% distilled = my compy laughing at me as it dies. I would run a mix of one of the -nols or anti-freeze. I haven't delved into that yet as it would be a non-factor if sub-zero is not even possible.

Yeah, In my calculations I am assuming 200W CPU. And throwing 4 150W Pelts running at 75% power to counter this. I read that Pelts perform better when drawing less than max power. If I do this, my loop to cool the pelts will for sure be able to cool all the extra heat, as I know there will be a lot.

I understand that this would take a second PSU, Perfectly fine with that.

Why? Been building pc's for 12 years, and I wanted to do something new. I know sub-zero, to that extent, doesn't give much bang for the buck... But I still like the sound of running super cold, even just intermitantly.

The last part you said, would this be more or less like a DIY Phase change? I've read into that a bit, and it does sound cool and I will probably do it eventually. But I just wanted to go slightly sub, not 20-50 sub... for now.


I guess the end goal is I want to push a machine... hard for 24/7 stable conditions. I don't have a real reason, I don't do editting or any other hardcore applications. I just want to "see what this baby can do". I am inquisitive by nature, and have alloted money in my budget to feed that, and this time I want to spend it on computer fun.

Anyway, thanks for the reply and the questions, I am interested in hearing what you mean by modding the window AC unit.

 

[Bobnova]

Here ya go: http://kingpincooling.com/forum/showthread.php?t=490

If you read through there you'll see even glycols freezing out of solution. Whether your pelt setup would have this issue or not I don't know, pelts haven't been done much lately. Nine CPU water blocks (cold+hot per pelt, plus CPU) sounds horrendously expensive to me.

 

[Winflakes]

Oh my lordy no! Not nine blocks! That would be way too much money for a pet project and an Idea.

What I invision is 3 custom fabricated blocks for the CPU cooling solution. One block runs to the CPU itself, and is filled with a liquid that has a low freezing temp. Then 2 peltiers on each side, cold facing in. And one Water block on each of the 2 peltiers.

I jsut read that article, and I must say... However, I am really wanting to build a closed system that is contained within a single case. I am going to be getting a big case, but I doubt it will have room for an AC chiller unit, regardless of how much time and effort is spent tidying it up.

 

[ssjwizard]

Sounds like a pretty standard 2 stage pelt water chiller your aiming for here. The one thing I will say is that these kinds of setups are usually large and mounted externally. For an internal setup your generally speaking looking for a peltier on die solution, which with a custom made block its easy to get 2 or more pelts directly mounted to the CPU with no cold side water loop involved.

Check this thread about peltier powered water chillers

Also see mindchills Dual TEC LGA 1155 block
AND
mindchill's Quad TEC LGA 1366 block

 

[Winflakes]

Yes, that first one is Very similar to what I am envisioning. Except slapping TEC's on the other side and adding another Water block to cool the hot side.

I considered mounting it directly to the CPU as Mindchill did in his 2 and 4 tec solution, the 4 tec one looks like harware sex btw, but I don't like it.

I like the idea of 1) running liquid to the CPU, it would mesh with the image I have in my head, and 2) I dont want the tec directly touching the CPU incase of extreme failure and massive overheating.

Reading through Mindchills posts, on the 4-tec cooler got me pumped. I think I am just gona have to start buying and fidgeting. Nothing like some good old hand on science!

Is mindchill or Elloquin on this forum? If not I will probably hop over there and pay them a visit.

 

[ssjwizard]

The first one I posted is an inline water chiller. It is composed of two water blocks. A hot side block which cools the pelts and carries the heat off to the radiator, and a cold side block that pull heat out of the water flowing over the cooled components. Pretty much what you asked about no?

 

[Winflakes]

The first one I posted is an inline water chiller. It is composed of two water blocks. A hot side block which cools the pelts and carries the heat off to the radiator, and a cold side block that pull heat out of the water flowing over the cooled components. Pretty much what you asked about no?

Yes, sort of. The example uses 4 tecs, one cold side one hot side. My idea would be like a sandwich. 2 hot sides as the bread and both cold sides touching the one single meat. That was probably really dumb...

I guess like this...

Waterblock to rads || Hot - Cold || Waterblock to CPU || Cold - Hot ||Waterblock to rads.

Does that make more sense? Would having the Tecs on both sides of the block make it any more efficient? Instead of 4 tecs on one side, 2 tecs on each side?

 

[ssjwizard]

Total surface contact area is all that matters not the shape.

 

[Winflakes]

Total surface contact area is all that matters not the shape.

This doesn't quite sound true to me...

Example - If I have a block of copper that is 4x1x2 And I placed 2 tecs on one side it would chill the block. but if I put one on each side, it would get the center colder faster.

If hot water (in respect to tec temps) is running from the CPU to the block and all the tecs were on one side, then the side without any tecs would be warmer, right? It may be by a small degree, but I don't see how it wouldnt be that way.

 

[Bobnova]

To an extent perhaps, but the copper on the cold side will be colder than it would be with the TECs spread out, too.
From that standpoint, a sandwich is a good idea. You don't really want the copper super cold or it'll freeze water out of whatever solution you use in that loop.

Here is what I recommend doing:

Buy a couple TECs, they're cheap.
Buy/find/use some lousy cheap PSU (if it dies, who cares?) and spend some time playing with the TECs. Use 'em for cooling something, strap a TRUE to one side and a water loop to the other and see what happens. All that sort of stuff.
Basically, familiarize yourself with the TEC.
We can really only offer you theory by and large because we haven't done what you want to do, specifics will only be had via hands on R&D

I have used a TEC for cooling, but I put it directly on the CPU.

 

[ssjwizard]

You must consider several factors here. One of the most important ones when dealing with liquid cooling is as the water flows you are generating micro-turbulance that stirs the water as it flows through the block. Secondly water is one of the most thermaly conductive materials in existance and holds a very high specific heat also. The goal is not for the water to freeze instantly because in a cooling "loop" your going to reach an equilibrium point that is completely reliant on 2 things and 2 things only. 1 total heat removal capacity, and 2 total heat generation. Volume of water position of tec color of the radiator and the OCZ sticker dont have any relevance to thermal equilibrium. You can somewhat improve total heat removed in the equation by increasing the total surface contact area that the water has to the copper, where these surfaces are located is irrelevant. More surface area doesnt increase the TECs capacity but it does ensure your maximizing your heat exchange efficiency.

 

[Winflakes]

Gotcha. So doing basically exactly what was in that guide would offer the same results as two tecs on each side?

On to another question... I have been looking around for tecs and I find them with DeltaT of 65-72, with one company claiming 79. Is there not any that are more powerful in that aspect?

 

[ssjwizard]

The problem with tecs is that total thermal capacity scales inversely to deltaT. I have seen tecs with a deltaT of 120C but there thermal capacity was only 10-20w at that delta.

 

[Winflakes]

The problem with tecs is that total thermal capacity scales inversely to deltaT. I have seen tecs with a deltaT of 120C but there thermal capacity was only 10-20w at that delta.

Gotcha. So the bigger the W the less potential delta?

Would it be possible to string like 10-20 of the one you mentioned to get a total large delta?

 

[OptyTrooper]

Not to mention condensation issues. This alone makes Pelts. a no go for 24/7 use. I have seen attempted again and again down through the years on numerous forums and almost always the results in the end are disastrous.

Cost is a factor as well considering you will have to buy a power supply designed for similar purposes and they tend to be expensive. Do a bit of googling and you'll see.

If benching is what your after which it seems you not but I'll throw it in anyway a single stage phase change unit is far and away superior to any pelt set-up and when all is siad and done not much more considering you'll likely end up buying new components when condensation kills it (in addition to the PSU, pelts etc.).

Not for nothing but I grow tired of pelt questions. A bit of research will tell what I and the other posters have already said.

 

[Winflakes]

The ultimate goal here is knowledge. I have been doing some research on peltiers and they seem interesting for sure! I was hoping I could get some expert information, preferably in the form of math equations.

I know condensation will be a problem, but nothing insulation can't fix. And if I DO ruin a computer from condensation... that's fine, at least I made something and tried it out for myself. Moving on!

You do with your money and time how you please, and I will do the same. If I want to ruin $10k worth of components what is it to you? Do I harm you in any way? If everyone took the stance, "Something already exists, don't try it on your own or you might break something" we would have a lot less innovation. I know about phase change machines, both prefab and DIY. Some day I will do a DIY one, but my current project is not something I want to have a phase changer for.

I always find it funny that people can say "condensation is a problem" and then follow that with "just get a phse changer"... Phase changers have to combat condensation a lot more... so why cant a different set up combat condensation?


Just so everyone is on the same page as me...
Yes I know condensation will be a problem, I will handle that.
Yes I know it will require a second PSU, already planned for
I want to learn about this technology

 

[OptyTrooper]

Look if you had done the research you would know that what you a proposing is impractical to say the least. It has been explored over and over. Don't get pissy b/c I am being straight with you.

Your right though if you want to dump large amounts of cash on a set-up that will ultimately fail that's your business. After all I am only a top notch bencher so what do I know?

 

[Bobnova]

Don't really need a second PSU, just a big one.
400w of pelts is nothing to a modern beefy PSU.
Now if you're using 24v pelts that's a different story of course!

Condensation issues make any subambient cooling a dubious idea in my book, but if you know what you're up to I certainly don't care what you do with your time/money

Plus, I'm curious to see how this works out.