TRUE TEC Module (with pics)

[ZOMGVTEK]

I dont know if i would want to do this, or just get the $300 OCZ phase setup...

[scifikg]

scifikg, check post#70, that'll be a real cooling test for you and us
Yep, 32M only loads one core.

Is the Cu you used Alloy 11000?

It is Cu alloy 110.


I ran the OCCT v2.0.0.0a auto test for 1 hour.

The highest temperature I saw is below. It occurred about half way through the test.

Temperature cold plate = -0.2C
Temperautre hot plate = 46.4C


This is closer to what I was expecting to see and a pretty big difference between the superpi test.

[STEvil]

Use Real Temp, Coretemp does not work very well with 45nm cpu's.

[KTE]

Pretty good temps those, it did work to some degree of decent efficiency there - it can be improved though. Thanks for the tests
110 has a conductance of 388 W/m-C at 20C BTW.

[wittekakker]

As Vapor said, each heatpipe has a load for it, most around here are 35W or 50W. Past that they become quite useless.

Useless to say the least, overloading the heatpipe might make the heat exchanger go into error, you might burn a thing or two. This has not been an issue yet, but I do remember seeing a chart which proves this.

[Clue69Less]

Excellent performance!

[Sam__]

you really should think bout manufacturing on a commercial basis if you can produce them for a reasonable price.

[evil-98]

you really should think bout manufacturing on a commercial basis if you can produce them for a reasonable price.

i see to many complications.. i see it being a cool custom thing but if your going to try and mass produce your going to face too many issues.

most obvious issue is it wont fit on all boards.. every layout is different

none the less still think the design is brilliant

[ZOMGVTEK]

What about getting some huge copper sinks, say 3 of them. Then put 3 300w pelts on 1 layer, and then a copper coldplate, and then 2 more 300w pelts into another copper coldplate, wich then cools the cpu. I have bunch of 150w pelts, i could mess around with those and see what i can do.

[scifikg]

i see to many complications.. i see it being a cool custom thing but if your going to try and mass produce your going to face too many issues.

most obvious issue is it wont fit on all boards.. every layout is different

none the less still think the design is brilliant

This is very true. There is another big problem. It won't fit in most cases. It just barely fits in my coolermaster case. As it is right now, it will only fit on some Intel DFI boards and only in some coolermaster or Lian-Li cases.

[systemviper]

Love the work with the TEC's, It's nice to see some real world application and numbers.

I have always been excited by TEC's but i haven't been around here to see much about them but warnings and negative comments.
And I know that most people are speaking from direct expierience, but it's cool to see something from concept to Trialz...

Great job!

[Jorlen]

What capacity does you fans run at to get the current temps?
And where on the coldside and hotside are your temperature probes attached?

Your results are really good, and the heatpipes seem to be working just fine.

[scifikg]

What capacity does you fans run at to get the current temps?
And where on the coldside and hotside are your temperature probes attached?

Your results are really good, and the heatpipes seem to be working just fine.

The fans are at 12V, but I'll be dropping them all to 7V when everything is said and done. When I was testing the module with no load, the hotplate temerature went up 3C as I dropped the fan voltage to 7V.

You can see the cold plate probe in the insulation picture. The hot plate probe is in the middle of the plate just between the corner edges of the TRUEs

[Duh]

I am glad it performed so good.. the usual negative comments seem to be vanishing after your results

[Jorlen]

I say but the probe on the CPU heatspreader, on the side not sandwiched of course. I imagine thats the hottest place you can measure without modding to much.
I just think your results are to good to be true.

[littleowl]

his numbers are right where they should be. He has done his homework and knows a lot!

[ArtosDracon]

Nicely done, glad it worked out. And congrats on a very nice, clean build.

Only last word of caution, when dropping the fans down to 7v be careful to keep hot side @ or under 50C. Going above that is going to start affecting your peltiers efficiency pretty quickly.

[Hazaro]

his numbers are right where they should be. He has done his homework and knows a lot!

Nicely done, glad it worked out. And congrats on a very nice, clean build.

Only last word of caution, when dropping the fans down to 7v be careful to keep hot side @ or under 50C. Going above that is going to start affecting your peltiers efficiency pretty quickly.

[STEvil]

I think you need to move your probe on the cold side, but otherwise turned out pretty good

[Utnorris]

I must have missed the answer, but how much power is required for that TEC?
Thanks,
Utnorris

[Uncle Jimbo]

The fans are at 12V, but I'll be dropping them all to 7V when everything is said and done. When I was testing the module with no load, the hotplate temerature went up 3C as I dropped the fan voltage to 7V.

You can see the cold plate probe in the insulation picture. The hot plate probe is in the middle of the plate just between the corner edges of the TRUEs

Using 12730 TECs at 9.2V, you should be pulling about 18A with a 50C hot side and 50C differential. That's a heat load of about 160W from the TEC itself. The other data you provided suggests you are pumping about 50W with each TEC, for a total heat load of 210W per TEC.

Based on what I can see from other performance reviews, the TRUE has a thermal resistance of about .1C/W with a good fan, so with 210W load, I would expect a 21C rise above ambient. You indicated the probes were at 23.5C with no power, so if we take that as ambient, we would expect the hot side at 45C, so your second test seems about right.

But if the comments by the other posters about the capacity of the heat pipes are correct, then you may be approaching the max for those coolers. I think that you might not be able to put much more heat load on - but I don't have that cooler so I really can't verify experimentally.

This data also suggests the CPU heat load is only 100W - otherwise you would see a bigger heat rise. Your first test had a 60C differential, and with your voltage, you could only have been moving 70W or so and achieve that differential.

All in all your work is impressive, and certainly very pretty. Getting close to 0C is an achievement, and you can play with TEC size and voltage if it turns out you need more cooling with different CPU settings. That would also give you a way to see how close to the edge you are with the coolers - if you increase the voltage and the heat runs way up or the differential does not increase, you may be at the limit.

[Vinas]

That's the best TRUE setup I've ever seen! Great work, Sir.

[scifikg]

I must have missed the answer, but how much power is required for that TEC?
Thanks,
Utnorris

Power Consumption ~= 2 x 9.2V x 18A = 331.2W

[m4gg0t]

im going to do something like this with a 168W TEC.

[desone]

One small suggestion. I would try to eliminate as much heat transfer as possible to the cold plate. One way to achieve this is to insulate the clamping bolts. Why not try some plastic transistor insulating mounts and some slightly smaller diameter bolts. The plastic mounts could insulate both plates.