TEC/Peltier CPU Chilled Water Cooling

[4Ryan6]

OK for the record I've covered in this thread that using a peltier and a radiator in the same loop basically cancels each other, if you're after cooling below ambient performance it is a bad idea.

Using a peltier attached to a radiator does not work and many have tried that approach and come away disappointed.

 

[4Ryan6]

Present cooling setup passes 1 year of operation, News at 11!

 

[4Ryan6]

It's really amazing to be able to run my 3770K at 5ghz 24/7 and the CPU not delidded, many wonder why even do it but the desktop and application response is amazing at 5ghz, it tends to seriously spoil you.

Once you get spoiled by the response you don't want to go back, it's kinda like comparing the performance of a Volkswagen Beetle to a Chevy Corvette, until you actually drive the two, you just don't know, you can imagine what you want, but once you know, well there it is!

I'm only running on one peltier energized unless I'm gaming, then I energize the 2nd unit and that comfortably drops my water temperature about 10c below ambient room temperature, which is gaming cruising for a 5ghz clock.

I am not running hyper threading on the 3770K, for the applications I am running it has proved to do more harm than good, there is almost zero of the games I play it could possibly help, and it actually cuts the audio/video encoding frames per second performance in half, somewhere in the vicinity of 650fps cut down to around 240fps, and that's totally unacceptable.

In all my testing the only advantage I've seen hyper threading gain is in Futuremark benchmarks, there it does show score improvement but that doesn't make it useable when it takes more CPU voltage which equals more heat to dissipate, to stabilize the CPU running hyper threading, for no other gains.

Passing as I said above 1 year in operation with this present setup, with zero problems so far!

My best to everyone! Ryan

 

[4Ryan6]

http://valid.canardpc.com/show_oc.php?id=1091644

Your CPU-Z validation still stands as one of the best anyone achieved back in the day, overclocking the Q9550 was a serious challenge going past 4ghz, there were only a handful that came near that success overclocking a core 2 quad.

You should be proud of what you accomplished! Ry

 

[a123b]

Hello there, long time lurker here. (Since ~2004)

Apologies for arriving late to the party but here's some random musing.
I have a theory why direct peltier mounting to the CPU will basically guarantee a "fail".

1) Peltiers don't have actual "cold" & "hot" sides.
The way you pump the power in is what determines the cold & hot side.

2) Peltiers are also generators that convert heat differentials into electricity.

3) One would pump electricity to the cooler in a way that cools the CPU side while cooling the "hot" side of the peltier.

4) CPUs get very hot, very quickly. (Hot enough to cook actually http://www.youtube.com/watch?v=7uBNCN6v_gk)

Putting those together makes it easy to see why it fails:
A) System starts at ambient temp. (Both "hot" & "cold" sides are at equal temps)
You set up the rig & turn on the system.
Peltier is now cooling the CPU.

B) CPU quickly generates massive amount of heat.
Peltier now sees the CPU as the hot side & the coolant side as the cold side.

C) Peltier starts generating a large electrical current >>counter<< to the power you're attempting to pump in = terribad efficiency while CPU continues to heat up further going against the input.

4Ryan6's set up circumvents this issue by working with the peltier rather than against it by actually having the cold & hot side in the way we want it rather than flipping it with a direct CPU mounted setup. (By allowing the water to dissipate the heat to a level the peltiers can manage, thus minimizing the counter current).

Or in other words:
4Ryan6's setup = "Full-Power Super Saiyan"
http://www.youtube.com/watch?v=xGsUN4do5mE

Direct CPU mount = "Ultra Super Saiyan"
http://www.youtube.com/watch?v=0J_u97yaJEQ

 

[4Ryan6]

@ a123b

Interesting theory!

:lol: Frying bologna on a CPU, that was interesting!

Most all that direct mount peltiers to a CPU run independent power supplies to supply the max voltage the peltier can handle which normally is at least 15v ~ 16v DC and some up to 24v DC, with the peltier powered separately from the PC itself.

In those circumstances all they have to do is power the peltier before the PC is actually powered up, then the peltier is producing enough cold way below ambient to override the CPUs heat, then they're back to the condensation and ice forming, that they insulate the motherboard against.

Your theory is actually exactly what happens with a direct mount when both are powered at the same time from an ambient temperature, but the direct supplied voltage will overcome the generated voltage effect rather quickly, but there will be a cold build up stall taking place, that's probably why they power the peltier separately.

Very Interesting!

Thank You for your input!

 

[kauedg]

Guys, I'm doing some research and I'm setting up something amazing. On top of Ryan's great solution I'm setting up an Arduino controller to read reservoir's water level, temperature and water flow. This way it's possible to achieve a fully auto controlled system. Here are some things that are possible:

- reservoir's water temperature;
--> Automatically make the second peltier module kick-in;
--> Warn when the water hits a defined temperature with the two peltier modules working;

- reservoir's water level;
--> Warn when the water level is low;

- water flow;
--> Warn when the water flow has stopped (eventually turning off the computer when the reservoir's water reaches like 30C, for example, so you have time to save stuff);

- peltiers' temperature;

- cooler fans' speed;
--> Change their speed to reach a desired peltier cold side temperature;
--> Warn if they stop working;

... and many more.


I'm making some tests with an arduino simulator until my modules aren't delivered and I start using the real thing.

 

[4Ryan6]

@ kauedg, I am definitely interested in your results!

 

[arthurh]

Finally I have my cooling system up and running. This is basically a clone of Ryans cooling system. The important components, not the case. It took 30 minutes to go from ambient to 10C with my i7 2700k CPU @ 5.0GHz. It idles @ 17.5C and during IBT run @ 55C avg.

This is a shot of my "Cooling Tower" using an old Antec case I modified for this purpose.



This is a shot of my PC showing the cooling lines going into the back of my case via a notch I cut out so when the side panel is removed the lines can be moved around for easier service. That suggestion came from Ryan as well and I really like it.



The next two shots are of both my PC and cooling tower as I have them set-up and running.





I have not gamed yet with this set-up but think it will do just fine. I will report back how my gaming session went when I get a chance.

 

[kauedg]

Ryan have you experimented with smaller reservoir size? It seems too big sometimes I check the pictures again.

Also, is the use of CM 212 necessary or was just something you had around?

 

[arthurh]

I am posting some more details about my set-up with pics.

This is a shot of my Swiftec CPU water block black top modified with a dremel tool for more water flow.



This is a close up of my peltier between a 1/8th" alum. block, Thermalright Venomous X-rt CPU cooler and modified Swiftec water block.



This is a shot of the whole peltier assembly.



This is a close up of my lapped alum. block. Lapped on both sides.



More to follow.

 

[arthurh]

This side panel had a clear window in it. I removed that and using the same mounting hardware for the window installed "Hardware Cloth".



This next shot shows my modified case before wiring and insulation.



This is the back of my modified case. Although you can not see it too well the cooling units are offset so I would have room to reach and adjust the pump without having to take the side panel off.



A closer peek at my pumps speed regulator.

 

[arthurh]

I got a request to show some more detailed shots.

These first two shots are with my removable plug removed.





This next shot shows a mounting bracket I made from 1/2" angle alum. to mount that assembly on and to my case.



This shot is not easy to make out but I used 1/2" U channel alum. to make a mount for my pump.



Still have not gamed enough to give accurate temp readings. I will post them once I have them.

 

[4Ryan6]

Nice Arthur, That's an interesting reservoir mount, can you explain what we cannot see in the pictures as to how you mounted it?

I know you've got a lot of testing to do, but in your opinion is your results so far meeting your expectations?

Please share everything you learn, It's very exciting to me another is running this cooling, Thank You for having the guts to take the plunge! Ry

 

[4Ryan6]

Ryan have you experimented with smaller reservoir size? It seems too big sometimes I check the pictures again.

Also, is the use of CM 212 necessary or was just something you had around?

At the beginning of the thread is a link to the entire journey that led to this setup, and I have experimented with a smaller reservoir size down to a dual bay XSPC reservoir, which was not an adequate cold storage collection reservoir, that's why I made my own using 4" PVC pipe, and insulated it.

The heat pipe cooler is not a CM 212, as it will not handle the heat load to properly cool the peltiers being used, the heat pipe coolers I am running are Thermalright TRUEs, which are still available today at.

http://www.performance-pcs.com/cata...oduct_info&cPath=46_450_504&products_id=26116

And all of this is covered in the threads, either this one or the linked thread, this thread is to share exactly what works should you decide to duplicate this cooling solution yourself.

 

[arthurh]

Nice Arthur, That's an interesting reservoir mount, can you explain what we cannot see in the pictures as to how you mounted it?

I know you've got a lot of testing to do, but in your opinion is your results so far meeting your expectations?

Please share everything you learn, It's very exciting to me another is running this cooling, Thank You for having the guts to take the plunge! Ry

First question. No.

Just joking. I had a 4" coupler that I glued to the 4" pipe then cut it off near where the pipe is, giving me double the normal thickness to use to drill holes in it without compromising the pipe in any way.

This system so far meets my expectations completely. I am very happy with the results I have been able to achieve thus far.

I have to add that Ryan really nailed this in his explanations of what works and what did not. I followed this to the letter buying the recommended parts he referred to in his thread and they work as stated. This is a fully reproducible cooling solution for those wishing to have a below ambient cooling solution that does not require extreme insulation of any other part of your computer, excluding of course the cooling lines going to your CPU water block. I did mount things some what differently and used a different case but the concept is "exactly" like his.

 

[kauedg]

Ryan have you experimented with smaller reservoir size? It seems too big sometimes I check the pictures again.

Also, is the use of CM 212 necessary or was just something you had around?

At the beginning of the thread is a link to the entire journey that led to this setup, and I have experimented with a smaller reservoir size down to a dual bay XSPC reservoir, which was not an adequate cold storage collection reservoir, that's why I made my own using 4" PVC pipe, and insulated it.

The heat pipe cooler is not a CM 212, as it will not handle the heat load to properly cool the peltiers being used, the heat pipe coolers I am running are Thermalright TRUEs, which are still available today at.

http://www.performance-pcs.com/cata...oduct_info&cPath=46_450_504&products_id=26116

And all of this is covered in the threads, either this one or the linked thread, this thread is to share exactly what works should you decide to duplicate this cooling solution yourself.

Thanks, I just lost the counting of how many times I've read the post. I'm following your setup and I'll make some measures with the hw I got (pelts' sweet spot, water flow speed and so on).

Sorry for asking about stuff already there, sometimes I miss a topic or another. I started to learn Arduino boards (just got it to work with Eclipse) and I think I'm starting a thread of my own, to document the work. By monday or tuesday I'll buy some parts to build a 3-pin fan controller (to control the CoolerMaster Blizzard T2 I have) and a temperature probe. The probe will read the water temperature and increase/decrease fan speed to maintain a certain temp level. Hope it works.

In the near future I'll make a Java program wich communicates with the Arduino board, showing the readings and allowing to configure the system's parameters. It's going to be awesome =D

 

[kauedg]

First problems that raised: this cooling solution takes a lot of space (another tower just for it) and the price... it's really expensive the get all the parts.

 

[4Ryan6]

First problems that raised: this cooling solution takes a lot of space (another tower just for it) and the price... it's really expensive the get all the parts.

A below ambient cooling solution is definitely going to cost you more than the investment an elaborate radiator cooling system that can barely match ambient room temperature will actually cost you.

The cheapest below ambient is simply a drink cooler with ice, pump, tubing, and a CPU water block.

Even that below ambient cooling solution requires a certain amount of space for it to be setup.

I never claimed this peltier cooling was a cheap cooling solution, it is guaranteed to allow you to run below ambient, but it is up to the user as to whether it is worth the expense to them or not.

 

[kauedg]

Have you considered using other heatsinks? Like this little monster:

http://www.ebay.com/itm/Thermalright-Macho-120-Rev-A-CPU-Kuhler-/400610689404?pt=DE_Computer_CPUs_K%C3%BChler_CPU_Zubeh%C3%B6r&hash=item5d4642017c

 

[4Ryan6]

Have you considered using other heatsinks? Like this little monster:

http://www.ebay.com/itm/Thermalright-Macho-120-Rev-A-CPU-Kuhler-/400610689404?pt=DE_Computer_CPUs_K%C3%BChler_CPU_Zubeh%C3%B6r&hash=item5d4642017c

I have used other heat sinks but not the one you linked, will it work?, I don't know, you'd have to try it yourself and report your results with it.

The parts assembly of this thread are guaranteed to work for you, and the setup has been in daily operation over a year now.

Now that another Toms Hardware member has built his on setup from the design and has confirmed it works, it is no longer my word against the world, and soon others will duplicate the setup themselves and discover it all works as well.

After all the trial and error experimenting, once I arrived at a successful cooling setup, I stopped experimenting and decided to run the system for longevity and dependability and so far so good.

 

[arthurh]

I just finished playing Crysis for a little over an hour @ a temporary clock of 5.08GHz.

Started at a mim of 19.75C CPU (according to Real Temp) and reservoir was reading 12.5C. My max temp was 40C according to Real Temp. I went back to 5.0GHz as I did not care for the 1.44Vc needed to run @ 5.08GHz with stability.

My bad on OC. Meant 5.08GHz. My BLCK is 99.76 default and I leave it there.

 

[4Ryan6]

I just finished playing Crysis for a little over an hour @ a temporary clock of 5.88GHz.

Started at a mim of 19.75C CPU (according to Real Temp) and reservoir was reading 12.5C. My max temp was 40C according to Real Temp. I went back to 5.0GHz as I did not care for the 1.44Vc needed to run @ 5.88GHz with stability.

If you're going to claim 5.88ghz stable overclock lets see some CPU-Z validation, and some completed IBT and benchmark scores to ensure total stability, and why worry about 1.44v when the Sandy Bridge max tested Intel voltage was 1.520v, you are still within 1.520v.

You should post an overclock like that in the Intel Overclocking Club thread, following all the requirements.

Now that you have this cooling in operation you need to handle what it can do with wisdom or else get ready to replace your hardware.

Once you damage the CPU Arthur, there is no undoing it, there's good reason my setup has lasted over a year without damaging my 2700K and 3770K.

Be Careful Sir! Ry

 

[4Ryan6]

My bad on OC. Meant 5.08GHz. My BLCK is 99.76 default and I leave it there.

Thank You for that clarification I can now go pull my head out of the toilet!

 

[toolmaker_03]

Well I starting to get there, I have finished the radiator module upgrades, it now has all metal fitting and strong hoses so that it can withstand the extreme temps of the TEC’s



The only concern I have left with this unit is weather the pump handle the temps, or if I will need to upgrade the pump to a high temp chemical pump. There not too bad on price, only $180 so it is doable, if this pump can’t cut it.