Sign in with
Sign up | Sign in
Your question

Cooling LGA 775 CPU using Peltier

Last response: in Overclocking
Share
February 28, 2006 1:37:59 AM

I only found out the name a couple weeks ago, and now I want to do it. My one and only concern is burning out the Peltier itself. I'm using a Zalman all copper 120mm heat sync/fan for the cpu right now, and i dont plan on much overclocking if any, i just want cooler temperatures since the 500 series seemed to be evil for heat.

I have the peltier chip sitting in front of me (one out of a portable cooler, got the core at a local bulk store today :D ) but i havent bothered to install it on my comp yet.

My plans for installation...

1. Lay computer on side, were it will remain during the testing stages.
2. Take bolts out for heat sync, I really think this is possible since my heat sync bolts are very long and will easily give me that extra play i think, if not i have longer bolts from a MAFS off a truck.
3. place peltier chip on processor, add heat sync compound if necessary/
4. put heat sync back on, screw down very lightly until the heat sync is nicely snug.
5. boot up and immediately watch the temperature.

I am using it off the 5 volt (22amp) line of an old AT power supply, since i dont want it to kill itself, though im probably better off with 12 volts(7amps).

I have everything i need to do this already, just need some advice or extra confidence before i go ahead with it. Im sure it wont do any damage to my cpu, and will transfer heat alright when its off, but i do not want to burn out this peltier chip.

Thanks a ton guys :D 


And now for pictures..


The Zalman cooler inside the case, the case is a thermaltake xaser 3


This is the assembly the peltier came in


And finally after about 15 mins i dug the peltier out


Testing it for heat on a celeron slot one, with no fan the heat sync got Voodoo 3 hot but the cpu still felt cool :D 


Ill post more pics as i get further along, as well as a last question... has anyone done this with a socket 775 yet?
February 28, 2006 6:16:42 AM

You gotta worry about condensation with peltiers because they can easily reach temperatures that will make them go below ambient temperature. If you didn't already know that...

I also demand that you apply a thermal paste, not applying will ensure that the entire die is not connected to the peltier.

Also this seems like overkill for just normal everyday computing, but you can get a nice cool setup with it.
February 28, 2006 12:47:42 PM

I figured out the condensation part having it sitting on my desk with the hot side sitting on the giant heat sync. I will be monitoring it very closely to make sure there isnt any problems.

As for thermal paste, i cant beleive i was stupid enough to think otherwise lol, not enough caffene yesterday, i knew id have some on already since i applied a good amount already (socket 775 like silver oxide compound better).

I know its probably overkill :D  perhaps i should consider some overclocking?

Thanks for the help


Also i have a small miniature train transformer (12 volts) thats adjustable, perhaps this would be better than the power supply?

And a last concern.. if during testing i need to turn it off, it will still let heat transfer through alright? It seemed to do it ok when its off when im just playin with it... but on a cpu would be a different story..
Related resources
February 28, 2006 3:26:34 PM

well.. a complete failure this was. I saw temperatures at 80c with it, this was just in the bios, and it decided to keep increasing so i checked the setup and found no flaws, i bumped it to 12 volts with no resolution. I took it back apart and now im back to were i was -a bit of heat sync compound. I think my problem was i wasnt giving it enough voltage. Ill try again later on, and hopefully get better results.
February 28, 2006 4:04:57 PM

make sure you use a non conductive thermal paste, such as Arctic Silver Ceramic or a generic silicon TIM. I say this purely as a precaution as the little that I know about peltier cooling is that it's to do with the flow of ions across the material.

To me, a flow of ions = current = bad to apply to a CPU heat shield.

It's been a long time since I did any physics, so I suggest you have a read...

http://www.madshrimps.be/gotofaqlink.php?linkid=2188

These guys seem to have a cracking howto section for us amateurs!

:p 

Let us know how you get on!
February 28, 2006 4:26:03 PM

If i dont use that silver oxide based compound though ill probably end up with the same temperatures i figure.. Its running fine now without the chip after trying, and i wont be trying for a while i figure until i get a more powerful transformer for it.

Heres some random pictures i took during installation..

Processor cleaned off


All assembled, you can just barely see the peltier between the cpu and heat sync.


It fit surprisingly well without any modification :D , perhaps in a couple weeks i'll stumble across a more powerfull transformer.

Thanks for the input guys! For now my peltier makes a nice beer cooler hehe
February 28, 2006 5:30:19 PM

Just been digging around for some more gumf on Peltier, and found:

http://www.heatsink-guide.com/peltier.htm

which goes through the principles quite nicely. I guess you have already, but this is probably for those who want to know a bit more....

By doing what you've been doing, you're going to need a much larger voltage than 12v, and you're going to need to exhaust all the heat that the peltier method generates as well as the heat it transfers.

I'm still dead curious where I can get me one of those peltier pads to have a play with though..... Looks fun!

Certainly less of a mindfield than putting together a vapochill unit! That was a mare!

Why is it companies never advertise the fact that sometimes their products are incompatible...... Curious that!
February 28, 2006 6:20:28 PM

I got a peltier off ebay about a month ago it was $6 and 127watts. the thing I found out was their worthless for cooling a new cpu. They really hit the limit with the P3s. One of the big cooler makers [can't remember which one] has a water cooler that uses a 250 or 275 watt pelter. It only reaches around 5c not the negatives like vaporchill. My plan was to use it after my radiator. It just doesn't seem worth the work now.
February 28, 2006 6:41:48 PM

suddenly my desire to look into this further has died!

Think I'll add the peltier onto the the dangerous "I'm just in form the pub, I think I'll just have a quick look on ebay" list.....
February 28, 2006 8:31:36 PM

I think i am going to still try it given the proper transformer, if all else fails, ill use it to cool a video card lol
February 28, 2006 9:38:41 PM

Ok ive been up to no good again.. Off to the FX 5600 it goes!

On my simple celeron slot 1, 300mhz (oc to something like 333). No questions about why im using a 5600 in a celeron :D .


And a close up.


Gets hot, but not too hot, it will only tell when i get it fired up into Windows 98, this card happens to have a fan tachometer and temperature sensor.
March 1, 2006 1:07:17 PM

Alright a bit of an update today, last night i improved on the cooling for the card.

Heres some pictures since there the most fun part haha.

1: The card before adding the heat sync, i used hot glue to keep the Peltier in place as well as tape and hot glue to hold the wiring down to the board and make it look pretty hehe.


2: The mounting for the heat sync i used, very unconventional i know, but this isnt permanent so i am not going to use silicone or anything to seal it.


3: And the other side, i put my useless P4 cooler fan to work on this one, does the job just awsome.


Im still trying to get the asus smart doctor working on windows 98se, for some reason the windows 98se version of it is just a fancy 7mb program to tell you its not compatible with windows 98se/me lol, go figure.

While messing around in windows 98se the card did not exceed room temperature, at least to touch, the real tale will be told when i get the smart doc working with the temperature readings which is my plans for today.
March 16, 2006 12:31:18 PM

I finally got XP loaded on the Celeron machine.. wow was that ever slow.. doesnt help i as useing an IBM hard drive.

allthough i did get the temperatures...

GPU was 36c and the memory was 36c. I overclocked the gpu as far as the smart doctor would let me and the temperature went up 1 degree.


And on another bit of an update, i felt like screwing around with the Zalman heat sync a bit, and it turned out a lot better then i thought.



I use silver tape for everything, and my only idea with this was that it would help get the fan to push the air down rather then lose half of the air out the sides, which would be ok but this cpu is hot as hell. My games were happy to let the machine just plain out restart several times, ever since this they havent hit me yet :D !
March 17, 2006 2:03:40 AM

Already tried, didnt work, no damage done luckly, dont want to remind myself of the stupidity :oops:  . Works for the GeForce 5600 just awsome though :D 

And off topic i beleive im going to need a more powerfull power supply. The comp would reset itself once or twice per day (this was before trying the peltier, just so no one thinks right away to point fingers at my failed experiment), i unplugged a fan, and no problem since :roll: . I was looking at thermaltakes web site and noticed a nice 750 watt one as a future product :D  hehe.

Thanks for all the input guys
March 17, 2006 2:58:15 AM

holy fucking crap

did i just see peltier with air cooling?

here is what you are doing wrong

1) the peltier is waay underpowered at 110watts... thats less than what the prescotts produce in heat

2) aircooling is definitly NOT sufficient to cool the peltier, what would happen is the hotside will melt the cold side and you'll have yourself a cpu with a peltier and heatsink attached



peltier in my opinion is just plain stupid, it wastes a shit load of electricity, don't go that low temperature wise and its a pain in the ass to setup not to mention the high risk factor

if you really want to use it, get a 170+watt peltier, a good water cooling setup, and a rock solid powersupply and you're good to go


or go phase
March 17, 2006 3:43:41 AM

Well i experimented and found out for myself, I dont regret it, played it safe (though i feel like a total idiot now). Immediately when starting up I entered the bios to monitor the temperature. I watched it raise from 40 to 70 then shut off the machine with one of them "yeah it was too good to be true" looks on my face lol.

So about 3 weeks later now..... i just wrapped some silver tape around the heat sync which seems to help quite a bit (confused on this). I havent seen it exceed 60C while in games, were before it would reach 65-67. I also tightened down the clamp for the heat sync (bent it a bit to make more pressure) which may of helped too.

All in all i agree peltiers are usless for comps, unless of course someone wanted to fork out the money for watercooling and a high quality one. Im not sure how powerfull mine was suppose to be, since its from one of them portable coolers..


Edit time.. Rather then wasting another post ill just add it here.

Took a screen shot of Asus probe temperatures as of right now, this room is pretty damn cold this morning so thats something to keep in mind, normally it is too warm for most people.


I usually leave a samurize taskbar config on so i can watch the temperatures, but it seems to be a bit higher than Asus probe, something i just noticed while posting this hehe.
March 21, 2006 11:31:22 PM

Well a bit of an update on the peltier, i sandwiched it between a celeron cpu and a massive heat sync and the factory 775 cooling fan (no heat sync on that of course).

Starting up i see around 17-22c, during idle itl reach as low as 13c, full load 22c.

The cpu temp sensor seems to be off 6F, so i was getting 50f on my TPI voltmeter temperature sensor, and 56 on the asus temperature sensor. I trust the TPI version more since the temperature diode is expose more. but either way it is pretty damn cold...

I failed to mention its a celeron 300A running at 450mhz
March 22, 2006 12:32:39 AM

lol, watch out for condensation
March 22, 2006 12:36:51 AM

hehe, condensation? bah im not all too concerned for a slot of celery...

I also put some crazy sticky piece onto the back of it so its probably sealed pretty goot for condensation, as well as i surrounded the peltier in hot glue which might also of sealed it. Dont know, not really too concerned about it :D .

Next step is to get it to run world of warcraft, ill try an get some screen shots of asus probe, i have one on the machine, but of course its off right now.
August 29, 2008 11:40:02 AM

Hi guys, like the article your discussing, just joined actualy, as im also trying this out too.

I have played with these devices for a while at work, as i used to make laser systems and they use a simmilar technology to cool the units.

I have often thought about using the same technology to cool my cpu, as i have overclocked it to 3.8Ghz, but it seems to run ok with the standard heatsink. between 60 - 70 'c with no problems.

However, i feel this is warming up my room a lot, and in the summer the PC really is under stress esp if i get home and start running X3 or simulating some stupidly complex circuit.

Anyways Id just like to comment, dont give up using peltiers, as they can be stacked, to achive a colder "cold" side. On our lasers we had to keep the temp down to about 50, so we had to stack 10 of the modules on top of each other. Condestaion was a concern, so we always used a polyurethane potting compound to envelope the entire assembly. that way, any condesation formed on the outside and was allowed to drain off.

The hot side of the peltier was as you can emagine, was very hot by now. So we designed an oil cooling radiator to take the heat from the hot side and dissipate it into the air.

I shall try to get my old duron 1200mhz pc back from my friend and try to cool it in this manner. in my eyes its expendible, so worth it for the test.

You guys ever seen oil immerced computers people placed in fish tanks, lol americans did that, its ok to submerge a power supply with 110V but id get a bit worried trying that on our british 240V system lol. anyways the setup looks cool and on one vid u can see the convection comming off the heatsink and chipset.
September 11, 2008 4:16:38 PM

I have some news, i have completed a prototype that i will spend some more time solving the condensation problem. However the fact its producing temperatures below 0 is good with just one stage..

I used two blocks of aluminum i machined for use as buffers between the cpu and the cooler and the heatsink. This at least adds some thermal mass to the system, and prevents the cpu from going into thermal run-away. Il be trying out a much larger heatsink also, the one pictured is off an old athalon cpu. and i think i used a 130W module for the test.

September 12, 2008 10:41:49 AM

^HOT. in a cool sort of way >.>
a b K Overclocking
September 12, 2008 1:48:52 PM

well if your planning to cool a quad then ure going to need a 400w pelt cos even then they can only transfer 250w of heat at no temp diffrence which means at 250w of heat its not better than if you didnt have the pelt there. also an air cooled pelt is fine but it wont reach sub ambiant temps so condenstation shouldnt be a problem. but tbh its not worth the 400w of juice that its going to suck down :p .
a b K Overclocking
September 12, 2008 1:51:31 PM

you cant stack pelts cos the first one generates and trasnfers heat from the source then the second one has to transfer its own heat plus the heat off the first pelt and also the heat of the source and so on :p .
September 13, 2008 12:27:42 AM

Actauly I know differently,

Assuming you get the peltier into a state whereby you dont actualy create a hot side, (as its well heatsinked) you then create a condition where the othr modules run much more efficently.

As you correcly said, the devices pump the heat from one module to the other. The result will be of course the need for a very big heatsink and possibly a radiator of sorts on the hotside, to remove the heat, but the cooling power of the combined units will be astronimoical. I have made one in the past using copper blocks between modules, and had the hot side mounted to a 1" thickk 500mmx500mm slab of aluminum as a heatsink. It generated a lot of ice, hovewer i have no pictures of it in action, however il be making a new one very soon when my 10x 138W modules come in the post.

I think if you have tried and failed the above technique, you probaly should consider machining your own heatsinks, unless you already have... the goal in any peltier system, get the hot side as cold as you can, and the cold side will get bloddy cold.

hope this helps.
a b K Overclocking
September 13, 2008 3:06:54 PM

yea well i bet if you try useing 10 pelts all you will do is burn out the last few cos of the heat loads. you cant run a pelt without there being a hot side its just how they run. as for the ammount of heat needed to pass through the last pelt in your 10x138w is near 1380w plus what ever your trying to cool. as you will know a 138w pelt cannot transmit that ammount of heat and it will burn out. the only way of doing it is to stack ever more powerful pelts that will give your desired effect. eg 10w then 25x then 60w then 150w then 400w etc etc
!