Interested in adding a TEC to my setup :)

[Kai Robinson]

Okay - having read up, i'm interested in bunging a TEC onto my setup. Not after massive subzero temps or anything, just some nice chilly temps

I'll be using a Zalman CNPS9700 - yes, i want to use this with an aircooler!

CPU in question will be a Core 2 Duo E7200, TDP of 65watts

The TEC in question is the CP1-12726, 243.5watts @ 15Vdc, 68* Tmax

So, the calculation works out at:

(1 - (65/243.5)) * 68 = 49.84*c

That should be adequate, no?

[Sam__]

That air cooler would not be able to handle 65 + 243 = 308W of heat. Hence why this setup would not work. sorry.

This is why people use TECs when they watercool.

[n00b 0f l337]

And also why they find there waterblocks really cannot handle large peltiers. Its too high a heat density. It's almost like dealing with Phase cooling however we have latent heat on our side for helping transfer.

[Kai Robinson]

Dammit....watercooling is just far too expensive

I remember in the days of old, Heatsinks with pelts on without any of this watery lark

[STEvil]

Just need a proper waterblock.

Either way I dont believe TEC's are usefull for high heat loads anymore. Use them on northbridges or as an assisting device but thats about it.

[Uncle Jimbo]

Okay - having read up, i'm interested in bunging a TEC onto my setup. Not after massive subzero temps or anything, just some nice chilly temps

I'll be using a Zalman CNPS9700 - yes, i want to use this with an aircooler!

CPU in question will be a Core 2 Duo E7200, TDP of 65watts

The TEC in question is the CP1-12726, 243.5watts @ 15Vdc, 68* Tmax

So, the calculation works out at:

(1 - (65/243.5)) * 68 = 49.84*c

That should be adequate, no?

I have to disagree with the posts above - with the lower power of the Core 2 duo chips, TECs can provide an efficient cooling solution using air coolers.

The Zalman CNPS9700 is a nice cooler - about .1 C/W with fan on high, .125 C/W with fan on low. Your E7200 is low TDP. So let's look at some likely outcomes.

First, you don't want to run the TEC anywhere near max ratings - you lose efficiency and don't need the heat transfer with that CPU. You can use any of the PCM thermal fan control circuits around, with a MOSFET power booster, to drop the voltage and maintain the CPU temp you want - that will also maintain a consistent CPU temperature, and run the power up if things start to get warm.

Looking at the 70W transfer curves for a 12726, 8V will put you at the 14A point with a differential of 25C. That is power in of 112W for a total heat load of about 180W under max load. Efficiency is a CoP of .625, which is the better end of normal commercial CoP for big TECs. The E7200 at Idle has very low TDP so you will see very low (sub-zero) temps if you don't control the TEC power, another reason to control power in.

The Zalman CNPS9700 will have a rise of 18C with that load. Depending on airflow ambient in the case will be higher than room ambient, but with a decent setup, you should be able to supply the Zalman with near-ambient air - so if we use 28C as the nominal temp for the heatsink, you will see 46C at the hot side of the TEC.

That will give you 21C at the CPU under max load. Just running the cooler alone with that heat load and ambient assumptions would give you 34.5C at the CPU, so the TEC gets you a big 13.5C drop.

If you use a 12730 TEC, the 70W transfer point for a 30C differential is 7.5V at 14.5A for total power in of 109W. That gets your CPU to 16C for about the same heat load and CoP.

If you are content with just below ambient CPU temp, you can get much better efficiency. Using the 12730 and shooting for 24C (which means you don't have to worry much about condensation), you can target 5V and 9A for total power in of 45W and a differential of around 15C for 70W transfer. Now your CoP is 1.5, about the best you can do with that differential, and total power used is 115W. Using the cooler assumptions above, you will see 39C at the hot side of the TEC and something like 24C at the CPU. So using the bigger TEC at much lower power gives about the same temps as your 12726, but using about half the power.

Using a bigger TEC yet, say the 19933 offered as the '437W' on FrozenCPU, you can get even better performance. The increased number of elements moves the voltage / amp curves up - less amps for a given voltage. 6.2V at 7A gives a 70W transfer across 18C with input of 43.5W, for a CoP of over 1.5, but with 3C better differential. Total power is 113.5W so the hot side is 39.5C using our assumptions - putting the cold side at 21.5C and wasting only 43.5W to get there.

If you later decide to run higher CPU power, all of these arrangements will scale - you will run more power, but should be able to maintain temps without much problem.

That should give you an idea on the range of options. I highly recommend controlling TEC power - that CPU has a big swing from idel to average to max power, and without control you will see a big swing in temps both at the TEC and CPU, which is bad for component life.

[Vinas]

You will not be able to OC much if at all with that setup... Also you'll need to insulate your socket. All of this for little to no OC benefit. At 3.0GHz your CPU will be near 100watts of heat and sorry, but no, the Air Cooler will not be able to dissipate 345 watts of heat adequately enough. However Jimbo is correct you could do it, but doubtful you'll get the results your looking for.

I say still do it and tell us how it works out.

[Kai Robinson]

Ideally, if i can get the load temp to say....25*c as stated, that would be fine - i'm not looking for condensation or 'extreme', just a nice cool chip...

Obviously the TDP is 65watts at stock, so i'll be overclocking to about 3.2 or so...as it is, i have an E4300 at 3.2 and when i feel the Zalman, its not even vaguely warm, turning the fan upto max makes next to no difference to the temps...CPU is lapped, too...

Question is though, i'm guessing i'd have to make a coldplate to sandwich the peltier between, right?

[Vinas]

You can purchase a cold plate from frozencpu for about $6... If you have some copper stock maybe consider making your own. The coldplate will stabilize your temps and I wouldn't run a TEC without one TBH... Ideally the coldplate should be tightened down to your TEC somehow at about 150psi. I think this usually requires creative thinking and is different on every setup. For me I just sandwiched the thing between my water block and was able to tighten it down with my regular mounting hardware. Without a strong backplate and bolts I would not have been able to tighten my setup good enough (MB would bow under the pressure).

Looking forward to your results!

[Kai Robinson]

I can get hold of copper stock easily enough, problem is how to shape it, i have no tools like a pillar drill or anything like that. I'd like to make one that would be universal enough to switch between boards, although LGA775 sockets are all the same sort of shape/size/clearance, so that shouldn't be too hard...

[Uncle Jimbo]

I can get hold of copper stock easily enough, problem is how to shape it, i have no tools like a pillar drill or anything like that. I'd like to make one that would be universal enough to switch between boards, although LGA775 sockets are all the same sort of shape/size/clearance, so that shouldn't be too hard...

The cold plate at FrozenCpu is 50mm square by 3/16 - exactly the size of your TEC. http://www.frozencpu.com/products/24...old_Plate.html
For the power levels you intend to run, that is adequate. I have not used these but have been told that they need a bit of polishing - there are good guidelines on this forum and it's not hard to do.

But with a cold plate the same size as your TEC, the heatsink mount will be clamping everything together, and the most you can put on your CPU is 70 PSI, with something a bit lower being recommended.

With the relatively low power you are running, that may be adequate. Be sure to use AS Ceramique or similar NON METALLIC grease, and use a very thin film.

I can't recall if the base of the Zalman CNPS9700 is big enough to cover a 50mm TEC - if not, you will need the same plate on the hot side. The whole TEC needs to be covered, otherwise you will see an early failure as the uncovered elements will be overstressed.

An alternative is to make a set of cold plates yourself. It is really not too hard and can be done with household 'handy man' tools and a little patience. For high power I use 3/8 copper plate stock, but for lower power, 1/4 or 3/16 is adequate. If you go that route, you can tap one of the plates and get exactly the pressure you want. The commonly quoted 150 PSI is not ideal, it is the minimum at rated power - I usually do 250 PSI. Max for most TECs is 300 PSI.

All of that may create challenges in actually mounting the whole rig. Have you lined up the TEC and heat sink to see if the mounting clears?

[Uncle Jimbo]

Here's a shot of a blue orb cooler mounted to a 4 inch square by 3/8 copper plate. I just mounted the hardware that was supposed to secure the orb to the socket, and instead locked it onto the plate. That forms the hot side plate. There is an identical plate mounted over the CPU - I drilled it to match the MB holes, and tapped it to take 3mm hardware to secure the upper plate. Using the stock MB back plate, I torqued the cold plate to 50 PSI, put the TEC and a neoprene gasket on, then torqued the hot plate to about 200 PSI.

Since it is a moment's work to remove the orb, it was easy to mount and torque the copper plates nicely, and they also protect the CPU from any stress.

I should have used 3 inch square plate but this was a spare parts deal and it's what I had around, and I was too lazy to saw that big chunk of copper. Sorry, I didn't keep any photos of the rest of the setup...

I tested this with a CPU that was putting out about 100W. Using a 19933 TEC at 12V, drawing around 12A, we drove the cold plate to 17C... with what I know today, I would have used 2 of those TECs at 6V each, which would drop the heat load and give much colder temps.

The point though is to show one way to get a good hot or cold plate without much effort.