Peltiers - dead and buried ? Or still viable ?

[aussie-revhead]

Guys, has anyone played with Peltiers recently ? Or is this an old idea that has been and gone ? If anyone has any results to share then please do.

The reason Im asking is AMD cpus seem to really like the cold but I dont want to go LN or phase, so its going to be either water with pelt or water without pelt.

I did see the thread here and found it both interesting and controversial. Could you use an over sized heatplate with two pelts, two water blocks ? I guess it would be hard to mount etc and just get messy ....

Or what about under an mcw60 on a 5870 ?

Any input would be great.

[Ultrasonic2]

yeah TEC's are still alive and kicking .. actually i think they are making a come back

heres a 62mm water block



and results below

How well does it work ?


I have a Q6600 overclocked to 3600@ 1.41Vcore
Please note the Core temps doesn't go below 11,4
Ambient 18.6
Run at 5 volts off the PC's PSU
CPU at idle
Please be aware the below numbers were tested on one of my older prototypes, so the newer blocks will perform better.



Full load 3600mhz 40 minutes (yes i know it says 2.4 but that's a win7 thing)
Mean well PSU @17.2v

[Mescalamba]

Thats nice, but is there anything able handling i7? C2Q is "cold" CPU compared to i7 (9xx series).

[Ultrasonic2]

Thats nice, but is there anything able handling i7? C2Q is "cold" CPU compared to i7 (9xx series).

Well the one above (62mm) may very well cool a i7 .. no one has actually tried to TEC cool a i7 properly yet


but if you want more power then there is this block

2x50mm TECs which will be able to cool a i7 with out a doubt

[craigbru]

Nice looking 2x50 block Ultrasonic! What kind of power draw would a typical pair of 50mm TEC's require?

[zipdogso]

Nice looking 2x50 block Ultrasonic! What kind of power draw would a typical pair of 50mm TEC's require?

Having 2 TECs like this is attempting to benefit from increased Q so staying in that vein of thought you wouldn't want a typical pair of 50's, you'd would want a high power pair of 50's. With a high power CPU and running the TECs at a reasonable power level you could hope for a COP in the region 0.8 - 1.
With lower power CPU's using less TEC power could get you COP of 2 or maybe more.

I haven't attempted to specify CPU's as there is some ambiguity about the actual power output of i7's and 4 core in general. I think a lot depends on whether or not one has many or any multithreaded programs. With the increase in the future of multithreaded programs it may transpire that i7s have a greater output than they do at present and require more cooling in the future.

[craigbru]

Interesting. I had a Wintsch Labs block in my last rig, and have been toying with the idea of TEC in my new one as well (i7 920 on a Rampage GENE II). I'm very limited in space, and have no room for an aux. PSU. Both of those negatives really tell me that I should stick with plain old WC, but I'm not one to play it safe. If I could find a workable solution, I just might do it...

So, what would you consider a reasonable power level?

[aussie-revhead]

A dedicated TEC psu is much smaller than trying to fit a second ATX psu in your case, and they werent badly priced from memory, maybe thats what you need.

http://www.meanwell.com/

[zipdogso]

Interesting. I had a Wintsch Labs block in my last rig, and have been toying with the idea of TEC in my new one as well (i7 920 on a Rampage GENE II). I'm very limited in space, and have no room for an aux. PSU. Both of those negatives really tell me that I should stick with plain old WC, but I'm not one to play it safe. If I could find a workable solution, I just might do it...

So, what would you consider a reasonable power level?

Well what I had in mind was 50% or thereabouts certainly you will have to do this or less for COP 1. Another reason why you have high Qmax TECs. You certainly wont be using 127 couples so your full power voltage will be 24 or even 32v.
You will have to ensure your setup is quality to avoid any further losses when undervolting like this i.e. good TIM, correct clamping pressure etc.
Later on if you find the CPU heat output does rise considerably in line with multithreaded progs you could eaily, at the expense of some COP, raise the power input to the TECs to obtain better cooling, as long as your PSU allows it. Running 2 high power 50's at 75% power should see you easily covering the ouput of a big CPU for a while anyway, aslong as you don't mind a COP of about 0.4 - 0.5. Your main concern at this point would be hotside cooling. This block is probably one's best bet (of those currently available.) for the near future without going all out on an expensive chiller.

[craigbru]

A dedicated TEC psu is much smaller than trying to fit a second ATX psu in your case, and they werent badly priced from memory, maybe thats what you need.

http://www.meanwell.com/

I apologize for the blurriness of the pic, but here was my last mod. I did indeed use a Meanwell 24volt unit. It supplied full power to the Wintsch block with a 437 watt pelt on a Q6600. The hotside was cooled by a pair of 240 GTS radiators, and temps rarely wavered from an icy -30 C, regardless of clock. In fact, I think the highest temps I ever received at 4 Ghz loaded, were around -20 C.



The next project is much smaller in size, but not in complexity... (It's the rig in my avatar)

Well what I had in mind was 50% or thereabouts certainly you will have to do this or less for COP 1. Another reason why you have high Qmax TECs. You certainly wont be using 127 couples so your full power voltage will be 24 or even 32v.
You will have to ensure your setup is quality to avoid any further losses when undervolting like this i.e. good TIM, correct clamping pressure etc.
Later on if you find the CPU heat output does rise considerably in line with multithreaded progs you could eaily, at the expense of some COP, raise the power input to the TECs to obtain better cooling, as long as your PSU allows it. Running 2 high power 50's at 75% power should see you easily covering the ouput of a big CPU for a while anyway, aslong as you don't mind a COP of about 0.4 - 0.5. Your main concern at this point would be hotside cooling. This block is probably one's best bet (of those currently available.) for the near future without going all out on an expensive chiller.

I'll admit that I'm a modder by nature, but I'm not always as good with the mathematics. That's why I've inquired about the power. While it may be theoretically possible to keep the CPU at reasonable temps with 50mm pelts, I've got some finite limitations. I've got a Black Ice 240 GTX, and a 240 GTS w/ San Ace 1011's to cool the hot side, and am limited to a Silverstone DA1000 for the PSU. Not exactly ideal, I realize. Then again, my last rig shouldn't have cooled as well as it did either... I know why dedicated PSU's are a good idea, but assuming I had the wattage headroom, could the DA1000 handle a pair of 50mm pelts at 12 volts? Would 12volts even be enough to cool properly? (I'm sorry if my questions have obvious answers, I'm just trying to grasp the technical aspects.)

[aussie-revhead]

wow craig you have given me new hope , -30 ? wow

I would like to see how the same setup would perform with a newer cpu/higher clocks

[craigbru]

Yes, I wish I still had my old data. The system was running quite well for a few months until I had a self induced catastrophic hard drive failure. Shortly after that, my crappy MSI motherboard decided to stop functioning. I didn't have the heart to put more money in to a S775 system, so I saved up a little and went i7. I can tell you this though, I'm not sure how cold the peltier actually got. I couldn't find any monitoring software that would read much, if any, below -30 C or so. Water temps stayed in the mid 30's, although that's really irrelevant as long as the CPU was cold.

So, if the setup I had could keep an OC'd Q6600 that cold, how much hotter does the i7 really run? I'd have better hotside cooling... I really want to find out, but I have to be committed to something that would likely work. I hate moisture proofing motherboards, and don't want to do it if I don't have to. That's the only reason I haven't tried my Wintsch block. Well, that, and the fact that I don't have room for the Meanwell in the next mod.

[shemer77]

are tec good for 24/7

[Ultrasonic2]

are tec good for 24/7

yes

[[XC] Oj101]

-30'c from 437w?

[aussie-revhead]

Who sells the true 437W Peltiers ? There are some cheap ones on ebay with high wattage ratings but are they really that high ? Is there a way you can tell ? Can I use a Heatkiller to cool a 437W pelt ? Is this the highest wattage pelt you can get :

http://cgi.ebay.com.au/Gigantic-62mm...item48357d2b88

It says there:

Operates from 0-16 volts DC and 0-32 amps

So it really needs 16v/32A to get the most out of it ? Meanwell have some 600W 15v 0-40A units that should do the job.

.

[zipdogso]

Who sells the true 437W Peltiers ? There are some cheap ones on ebay with high wattage ratings but are they really that high ? Is there a way you can tell ? Can I use a Heatkiller to cool a 437W pelt ? Is this the highest wattage pelt you can get :

http://cgi.ebay.com.au/Gigantic-62mm...item48357d2b88

It says there:

Operates from 0-16 volts DC and 0-32 amps

So it really needs 16v/32A to get the most out of it ? Meanwell have some 600W 15v 0-40A units that should do the job.

.

Unfortunately ther aren't any true 437w TECs they are considered to be the ones from .....frozen CPU I think. But they aren't strictly 437w either....the true figure escapes me at the moment and I can't remeber where to find it.

That gigantic TEC has a cooling power of just 257w (Qmax.) at a hotside temp. of 25ºC. 545w is just the maximum power consumption from the wall socket and means very little as far as cooling goes.

257w is at maximum power with a Dt (difference betwen temp of hot and cold sides.) of 0ºC. Of course if your hot and cold sides are the same temp. then there is no cooling going on. As Dt rises and cooling starts to take effect the Q (that figure of 257w) starts to come down. Dt is affected by input power the higher the power the greater the Dt Usually a max of around 70ºC....but you will NEVER achieve this figure.) and the load on the coldside, the higher the load relative to the max. cooling the lower the Dt. It is also affected by the cooling on the hotside. The Dt figure is subtracted from the hotside temp to arrive at the temp of coldside, so in a country like yours low Dt's won't count for much.

There is a problem with the specs because the true maximum voltage is 15.6V and the maximum current is 30A (at a hotside of 27ºC.) and though 0.4v might seem irrelevent it should be stressed that you should not exceed either maximum figure if you want your TEC to have a long life.

there is very little to be gained by powering a TEC more than 75% input power as at this point it is already operating at 80-85% of it's maximum cooling power. It is common to power 15.6v TECs at 12v since it is close to 75% input power. This explains why 127 couple TECs are so popular, 127 couple TECs are all 15.6v parts.

If you scroll down to the specs of your gigantic TEC you will see it is a 12730 - 127 is the number of couples and the 30 is the maximum current, usually at 27ºC hotside. Standard specs for a 12730 show it has a Qmax somewhere in the 260w range it can vary a bit depending who makes/tests them. Ebay sellers don't want you to see this, so they don't quote it. But ironically it is one of the most important specs.

I don't see any reason why your heatkiller would not work but you would need it to be connect to a suitably sized rad/rad bank to dissipate the heat as you will have roughly twice the cooling wattage coming of the hotside. So if you have an amd phenom II running at Stock speed with a TDP of 140w you would need a TEC with a Q of 140*2.50 (to allow for 75% power and a theoretical Dt of 30-40ºC,)= 350w (Qmax) minimum and running the CPU at full throttle you will have a maximum of 280w coming off the hotside. 'Course if you overclock you have to allow for that also and the heat output of the CPU rises fast with overclocking.

Given that very few people ever run 4 cores (more likely only 2.) on a CPU, unless benching or they frequently use multithreaded programs of which there are few at the moment it is perfectly possible a smaller TEC will suffice as the calculations are done at full power but in the future when you started using more multithreaded programs would you remember to "up" your TEC size or just watch your TEC and CPU die ?

[aussie-revhead]

OK so then its no use trying that TEC on a w3520 ?

Frozen cpu is out of stock on the 437W TEC, does anyone else sell them ?


.

[Ultrasonic2]

OK so then its no use trying that TEC on a w3520 ?

Frozen cpu is out of stock on the 437W TEC, does anyone else sell them ?


.

Frozen cpu don't sell a real 437w TEC the cheat be telling you it power at a higher hot side temp so it seems more powerful than a TEC rated at the normal 25c hot side temp

[Duh]

yeah TEC's are still alive and kicking .. actually i think they are making a come back

heres a 62mm water block



and results below

How well does it work ?


I have a Q6600 overclocked to 3600@ 1.41Vcore
Please note the Core temps doesn't go below 11,4
Ambient 18.6
Run at 5 volts off the PC's PSU
CPU at idle
Please be aware the below numbers were tested on one of my older prototypes, so the newer blocks will perform better.



Full load 3600mhz 40 minutes (yes i know it says 2.4 but that's a win7 thing)
Mean well PSU @17.2v

is that fastfreddy ´s?? very sexy indeed

@Ultrasonic2: who makes that monster?

[Ultrasonic2]

is that fastfreddy ´s?? very sexy indeed

@Ultrasonic2: who makes that monster?

I make that block

but it's not a monster

This is a MONSTER two 50x50

[aussie-revhead]

Ultrasonic - what is that 62mm and the dual 50 setup worth ? Do you ship overseas ?

[zipdogso]

Ultrasonic - what is that 62mm and the dual 50 setup worth ?

They are worth a good deal....possibly a revival in TEC usage especially the 2 x 50....he has spent ages designing these and there is no reason why they can't handle the higher TDP CPU's with the correct TEC selection....these are worth whatever he charges and since he charges little/nil for his time and they do take hours to mill never mind the hours he spent designing them, they are a bargain.

Do you ship overseas ?

Well he ships to the UK so since your only in his backwater....

[Ultrasonic2]

thanks zip

i find your original question very interesting “are TEC dead and buried or are they still viable?”

TEC’s have and will be viable for years to come
In my opinion TEC’s died out because people didn’t understand them and therefore they were implemented incorrectly which has lead to failures
There has always been 3 camps of miss understanding

Camp 1 .. i have a 1 watt tec and i want to use it to cool my 500watt CPU

Camp 2 TEC’s are to inefficient to work

Camp 3 i want to make a TEC unit as cheap as possible

All the camps are of corse wrong. And poor ZIP, Flak and ME have been trying to tell people how it really is. Unfortunately people simply don’t believe what we say .. To be honest not believing has become so prevalent that people’s reactions to our comments have become predictable and rather amusing. No matter what we say people wont believe us

Interestingly enough people still don’t believe me. The few block i’ve sold i’ve tried my best to get them to buy a high voltage PSU but they wont ..they all want to run it off their PSUs which means they wont work as well as they could. And of cores when the blocks don’t perform they won’t blame themselves for not listening they will throw their hands in the air and ether blame me or they’ll say TEC’s dont work

After many attempts of trying to convince people to make a proper direct die or chiller unit and failing i decided id have to do it my self

OS i borrowed the money and bought a mill and since then i’ve been pumping $1000 a month into this project and getting virtually nothing back.

I do believe 2010 will be the year of the TEC revival after the duel 2x50 block ( still need the help of a AMD guy) im goanna concentrate on making a practical water chiller . i’ve made chillers before and seen others make them and so far they’ve all be to impractical to use for more than a few weeks. So i want to concentrate on making them useable as possible for the long turm

[craigbru]

-30'c from 437w?

Correct! I'll poke around and see if I can find a screenshot. Although I lost most of my data, I still might have one on a imageshack account.


Here is one at stock clocks. I'm still looking for an OC'd pic, but the temps didn't rise. This leads me to believe I was having a software limitation with reading the temps, and the peltier had headroom yet...