peltier

[r0ckstarbob]

hmmm

thought i knew about peltiers but maybe i don't. here's the question.

i'm told you'll see a -20/-30 degree temp difference or so at max using one of these (at least from ambient that is)

i guess the question is this.

do you get increased performance from your peltier if you cool it more? does the peltier perform better with better cooling on the hot side? is this progression linear???

example... instead of watercooling with room temperature water to give me say 20C, i'm going to watercool with water that is say, -2C... could i, in theory see temps down at -15C? is the rule with pelts "the more you cool it, the better it works?"

and yes, i'm very familiar with the condensation issues involved...

[KILLorBE]

Originally posted by r0ckstarbob
is the rule with pelts "the more you cool it, the better it works?"

Yes, lets say the max. temp difference is 60C and you can keep the hot side @ 30C the cold side will be -30C (roughly) that's w/o load of course.
Also you need pretty much pressure.
Quote from Kryotherm (http://www.kryotherm.spb.ru/Cooling3.HTM): For the assembly installation it is required to use the following torque:
Modules of FROST line - 25-36 kg x cm2
Modules of SNOWBALL line - 13-28 kg x cm2

[r0ckstarbob]

you da man. thats great news!

so curiously, is 60 degrees a realistic number to expect (depending obviously on the strength of the pelt) or were you just pulling that number from the air to use as an example??? i'm being quite serious. i have no experience with peltiers.

[r0ckstarbob]

and is a cold plate required or optional. i'm pretty sure i can figure this one out but thought i'd ask...

[KILLorBE]

Originally posted by r0ckstarbob
you da man. thats great news!

so curiously, is 60 degrees a realistic number to expect (depending obviously on the strength of the pelt) or were you just pulling that number from the air to use as an example??? i'm being quite serious. i have no experience with peltiers.

Its sort of an example but it is possible (in theory at least)and max. temp difference of most pelts is around 60C, Lets say you cool the hot side to about 30C the cold side will be -30C, but only w/o load (IE no CPU or anything else), now there's already a little load and that is ambient/room temp, but I'm sure if you wait a while you'll begin to see ice on the cold side.
It also depends on the strenght (watts) of the pelt and how much heat you need to get rid off.

I would say a cold plate is a must not really required but it improves performance a lot, as a pelt doesn't cool at a constant rate, and also your CPU doesn't put out the same amount of heat all the time, the cold plate will keep your CPU temp at a more constant level.

If your cooling your pelt with a water block and your using water that's -2C I'm pretty sure you'll see sub-zero temps, but keep in mind when your using water that cold condensation will appear all over your setup (IE hoses, reservoir).

Kryotherm (http://www.kryotherm.spb.ru/Cooling3.HTM) has some good info (I think), have a look.

[r0ckstarbob]

thanks. i think i will.

am curious because actually it's not -2C... it's actually going to be closer to -27 C so we'll see....

alrighty, off to the link now!...

RSB

[It_The_Cow]

Post more parts of how and what you're going to use and I'll try to get some more specific numbers for you

[r0ckstarbob]

http://216.254.0.2/~language/interface/computer.html

[It_The_Cow]

All right... At 1.60GHz and 1.85v, you'll be putting off about 95 watts of heat. With a 172 watt peltier at 24 volts and with a DTMax of 69, cooling that will bring it down to -27° C, you should expect load temperatures around 3°C. Basic equation is add 30 to your coolant temperature

[Colin]

-27C will not happen. You need to take into account the losses in the thermal interface between the pelt/waterblock, pelt/coldplate and the coldplate/CPU. The software from Kryotherm is great for modeling but does not address these factors.

Some good reading on the application of pelts:

Peltiers in Practice (http://www.dansdata.com/peltprac.htm)

Applying Peltiers (http://www.heatsink-guide.com/applying-peltiers/)

[r0ckstarbob]

thanks for that.

actually those are the concept sketches for this thing, and theres alot that i glossed over just to put something together.

things that didn't get detailed in there.

the freezer unit will actually reach -40F/C guaranteed. i can get the liquid coolant to -27F (-32C) no sweat, guaranteed. that means -27F for the CPU, the AGP, and the NB, SB, Mosfets, and Clock Multiplier. with its current mounting situation being flush against the reservoir, i'm effectively watercooling the entire back of the board as well incidentally.

i've got some voltage mods going on too.

the Vcore will no longer be limited to 1.85V, it will now go as high as 2.7V

The VIO will no longer top off at 3.7v but go as high as 4.9V

i'm still looking for a VIO1 mod for the memory.

i guess the question i'm asking now is will adding a peltier or two be more efficient or less efficient then to straight liquid cool this thing?

[Random Nonsense]

well pelts will make it colder if thats what you mean!!! efficiency wont be as good though as you are having to dump more heat overall..... be very careful, as with colant at -27 AND peltiers you will definately need a way of getting the chip warmed up a bit before slowly cranking up the cooling..... if you just flick the power button the die will probably shatter as it is cooled to stupidly low temps very quickly......

[r0ckstarbob]

well, apparently the improvements work on something of a bell curve so i guess thats what i'm lookin at. eventually the pelt puts out too much heat for the water to cool off. while that shouldn't be a problem with my system running 4 gallons of water/methanol through a routed copper reservoir (see the link i posted earlier) that will effectively make that liquid travel 6 or 7 feet before it reaches the pumps again... still, it's something to consider)

even with three pumps going its going to take an entire 60 seconds for all the liquid in the system to completely cycle through... that means all the liquid in the reservoir has a full 60 seconds to run through 7 feet of -40F copper before being pumped back into the CPU.

actually i've been in contact with AMD about this problem and several others... it seems this project of mine is generating some interest abroad. we'll see how it turns out.

at any rate

the cracking of the chip was a problem i'd considered for awhile and after talking with several thermodynamic engineers and the guys over at AMD, we've all kinda come to the same conclusion... that it's pretty much a misgnomer. remember, heat is a byproduct of current, not the other way around. by supercooling it first and then turning it on, the current starts to flow, and then the heat slowly accumulates. when we apply that with heat transfer theory, we understand that that heat is actually being drawn through the core and into the cold plate in order to cool down. just the process of heat transfer itself is warming the chip up.

according to AMD though the likelyhood of the chip cracking under temps like that anyhow are pretty unlikely i'm told. theres little or no expansion or contraction in the silicone chip due to the chemical makeup they tell me.

reference several systems where guys plunged their stuff into liquid nitrogen (N2) and fluroinert, cranked it up and started their benchmarks as soon as they had a screen to do it on.

reference the vapochill which supercools it to -14 and then lets you power your board up.

lots of other stuff too. but i appreciate the heads up! :)