liquid coldpalte heat pipe thingy idea

[athlonnerd]

i had an idea to use heat pipe fluid to convect heat from the cpu to the cold side of a pelt, water is pumped over the hot side. i didn draw he barbs. what to yall think?

[athlonnerd]

nother view.

[athlonnerd]

yet another

[Lt. Max]

i think it seems like an awesome idea to me, does water transfer heat better then copper tho? i think so right? one problem is that if the bottom water leaks, and also u would prolly havta put antifreeze in it mab

[DManeKid]

it looks like you are planning to put the liquid directly on the die...are you?

[athlonnerd]

yeah, the heat pipe fluid is on the die, it owuld be epoxied to no leaking, id have to drill holes for the pelt leads. the pelt would cool the heat pipe liguid, adn woujld be directly cooled by water. i dont know what the ideal heat pipe fluid would be though.

[Diggrr]

Well the drawings look good, but I gather the substrate (large flat ceramic) of the chip won't hold water, vacuum, or pressure. Kinda kills a heatpipe altogether. Coatings won't hold up either.

A heatpipe uses distilled water as the working fluid (many do anyway) and it's contained in the pipe under hard vacuum to change it's boiling point to a more desirable range.
Some direct die experiments are failing due to the ceramic leaking water into the socket with newer processors, unless care is taken to avoid direct ceramic contact.

Too bad Scotty's transparent aluminum weren't going to be invented for a few hundred years. <Star Trek movie, remember?> Waterblocks would rock using that.

[athlonnerd]

well, i thought for one thaat the new amd "ceramic" a new organic material, but then what do i know. it would be easy so seal the substrate, just cover it with super glue. and woulnt isopropyl alchohol be a better heat pipe fluid, i know that mmethanol, amonia, adbn acatone are used.

[Diggrr]

All four of the fluids you mention would dissolve super glue and epoxy, and most paints/sealers/resins. Especially if left to soak in it.

And yes, it's called organic....ceramic was just a descriptive word. Who really knows exactly what it is anyway, except that it leaks.

[athlonnerd]

lol, well im shure some good combination could be found, im shure those chemicals dont attack all possible adhesive thingies. i know silicon is supposed to react with very few things, maybe we could use silicon sealant.

[sillyunclemark]

Hmmm. A fluid that can boil around room temperature and not eat away at seals or the processor. I would suggest propane or butane but I don't think the processor or ceramic base would stand up to the vacume or pressure needed to get them to vaporize at the right temperature.

Using the Idea for a thermal spreader with a nonconductive fluid like machine oil might work. I have no Ideas about the long term affects of soaking the cpu in oil :( I know some High dollar stuff has been mentioned for direct die cooling and dunking the whole motherboard. 1/2 a cup of the expensive stuff for direct die cooling results with a standard water block. :)

IMHO the fluid would need to be stirred (magnetic stirring ?) to maximize heat transfer to the water block.
Also I fear the added thermal interfaces will negate any gains.
Somebody please prove me wrong. :D

Somebody try it or we will never know.:beer:

[athlonnerd]

if i get enough input, adn we agree on a material selection that would work well than i would liek to try it.

it was supposed to be a heat pipe, heat pipes use convecion, and thus dont need stiring, however, you brign up an interesting point, one oucld use plain water(any body know of the affects of water on cpu die), and rely upon stirring the wateraround. if that was done then it would be like a direct die + peltier setup.

also, i dont know if yall realized it, but the blue/red soloids are the cold/hot side of the peltier, so your actually reducing the amount of thermal interfaces. now we have 4, cpu to water, water to pelt, pelt to water, water to radiator. in a conventional pelt set up, wed have cpu to tim(thermal interface material aka as3), tim to coldplate, coldplate to second tim, second tim to peltier, pelt to third tim, third tim to wb, wb to water, water to radiator. also a heater pipe/ aggitated heat transfer fluid setup will signifigantly ou efficientize a coldplate, even if a new type of coldplate is made sich as silver's idea(which seems really good).

i really like the idea of the aggitated heat transfer fluid, ill see if i can come up with a design.

[DManeKid]

Originally posted by sillyunclemark
Hmmm. A fluid that can boil around room temperature and not eat away at seals or the processor.

i think the point in to make it not boil...am i wrong? if i am why would you want liquid to boil in a compressed area?

[athlonnerd]

its all to do with how heat pipes work, if it boils on the die, then it turns to gass, and subsequently rizes faster, thus taking heat away from the die quicker. it doesnt have to boil, but if it dont, then your relying purely on convection, which i belive qould not be sufficient.

btw, nobody said that the housing absolutely had to be lexan, it could be a metal, i mainy drew it as being transparent so youd have a better idea of what i was talking about.

[DManeKid]

Originally posted by athlonnerd
its all to do with how heat pipes work, if it boils on the die, then it turns to gass, and subsequently rizes faster, thus taking heat away from the die quicker. it doesnt have to boil, but if it dont, then your relying purely on convection, which i belive qould not be sufficient.oh ok that us understandable

Originally posted by athlonnerd
btw, nobody said that the housing absolutely had to be lexan, it could be a metal, i mainy drew it as being transparent so youd have a better idea of what i was talking about. yes an i must say that it looks very cool and i think you did a good job with it and you avatar...may i ask what program you did it with?

[yoink]

i was just going to ask the same thing. Looks very professional indeed :)

[athlonnerd]

AutoCAD, best program ever.

heres a design for agitated cooling. you would have to obvioulsy turn the white thing in the middle wiht a motor, ive never used one of those things that lets a shaft spin, but also seals a joint, but you do get them. other wise it could just be like the desing pumps use. the white thing is an impeller, and sucks liquid over the core, adn then throuws it out, a duct/shield would be needed to focus the water over the pelt.'

btw, the previos desing would need to be horizontal.

[Mizzery]

If your going to be gluing it directly to the cpu, then I'd just rip the rubber spacers off. I'm pretty sure that water would not work as well as a copper plate in this instance. Yes water holds more heat, but I don't think it transfers it as well. Heat pipes work good cause the fluid evaporates. You won't have a vaccume and will be relying on convection to transfer the heat, unless someone thinks up a magic fluid with a low enough boiling point. Conduction always beats convection.

[athlonnerd]

i know, if i were to use water, i would use an agitator setup. if i were using a heat pipe device, i owuld use something else. adn pulling the rubber pads off would just be too easy.

[SkiFletch]

I for one am still confused as to how you guys know that the "organic" surface on the new XP's will leak water... I would think that if you used some say 5 ton 24hour epoxy you should be more than able to seal it for a standard water application.

but, I question the whole idea behind this. how is it better than a standard direct peltier contact type design? it would make sense to my moderately engineeringly challenged mind that introducing a medium such as water would add another material to transfer heat through, rather than the standard pelt-core direct contact design. maybe I just dont understand "heatpipes" well enough.

[DManeKid]

ahhh! its one of those pizza things! (i know you know what im talking about)

[athlonnerd]

itsn called an impeller, works like a centrifugal fan.

and the problem with normal pelt set ups, is that the copper copld palte isnt as effective as it could be, some studies were done, and they found that the outer parts of the pelt dont even cool the die. this set up would theoretically perform much better.

[drunkmonkey]

That is a great idea, I think that would be the only on-die cooling that would do extremely well since the liquid boiling of of the proc would very rapidly take away heat. As long as the whole chamber doesn't fill with gas of coarse:D, that would be very bad. I hope you can do it cause I really wanna see ow it turns out. I wanna get autocad so bad, I LOVE that program.

[Since87]

An integral part of a heat pipe is a wick to carry condensed fluid from the cold side to the hot side. You could probably let gravity do the job if you could make it work with a large enough quantity of low surface tension fluid.

Another thing to keep in mind is that the manufacturer's suggestions for using pelts means compressing them at around 200 PSI. To do this without warping your cold plate (and therefore losing thermal contact with the pelt, the section of the cold plate which is in contact with the pelt needs to be pretty thick. (1/8") Assuming the heat pipe works very efficiently, the thickness of the copper at the pelt junction is likely to be the determing factor for the overall system thermal conductivity. I'm not sure if the heat pipe is really gonna buy much over a 1/8" coldplate directly between processor and pelt.

[DManeKid]

Originally posted by Since87
An integral part of a heat pipe is a wick to carry condensed fluid from the cold side to the hot side. You could probably let gravity do the job if you could make it work with a large enough quantity of low surface tension fluid.

Another thing to keep in mind is that the manufacturer's suggestions for using pelts means compressing them at around 200 PSI. To do this without warping your cold plate (and therefore losing thermal contact with the pelt, the section of the cold plate which is in contact with the pelt needs to be pretty thick. (1/8") Assuming the heat pipe works very efficiently, the thickness of the copper at the pelt junction is likely to be the determing factor for the overall system thermal conductivity. I'm not sure if the heat pipe is really gonna buy much over a 1/8" coldplate directly between processor and pelt. your worried about the metal?? what about the cpu? i really dont think that it would hold under that pressure

[Mizzery]

I for one am still confused as to how you guys know that the "organic" surface on the new XP's will leak water... I would think that if you used some say 5 ton 24hour epoxy you should be more than able to seal it for a standard water application.


The organic material is not water proof and I believe water seeps through it. If you do a search on direct die, you will come across this. Many people have had cpu's fall to direct die water dammage.

[MajinSSJVegetto]

Originally posted by DManeKid
ahhh! its one of those pizza things! (i know you know what im talking about)

Lol... the things that keep the pizza box from touching the pizza... you are weird.

But anyways...

Couldn't you just put a VERY thin layer of copper or silver between the liquid and the CPU?
Just enough to keep liquid and pressure contained?

[Since87]

Originally posted by DManeKid
your worried about the metal?? what about the cpu? i really dont think that it would hold under that pressure

You don't apply 200 psi to the processor, you apply it to the pelt.

Copper is pretty soft, so you need fairly thick copper to prevent it from deforming and giving you a bad coldplate/pelt junction.

[athlonnerd]

well, you need 200 psi when using a coldpalte, so as to ensure good thermal contace betweer the copper and pelt, 1/4" wasnt enough for me, (in my currnet project), in fact it warped so much that the pelt cracked, im using 3/8" no, but you can still see form the theraml grase imprint that the middle int applying as much pressure. however, im not using copper in this idea, i was just gonna have the water touch the pelt it self.

Couldn't you just put a VERY thin layer of copper or silver between the liquid and the CPU?

one copuld make a heat spreader desing. but that just takes away from performance. i was thinking of putting alot of super glue or ehat ever aroung the die, to the heighth of the die.

you could always coat the area aroundd and on the die wil a very thin layer of arctic epoxy......

if you want auto cad go to kazaa..... get auto cad 2000, not 2000i or 2002, they both suck donkey genetalia.

[drunkmonkey]

Originally posted by athlonnerd

if you want auto cad go to kazaa..... get auto cad 2000, not 2000i or 2002, they both suck donkey genetalia.

good idea. I gotta take advantage of my grandparents satillite internet while i'm still here. I'm downloading redhat7.3, .net rc1, and tribes2(not illegally, mom broke my cd:().

[athlonnerd]

i might try this idea if amd prices go down. actually, someone with a p4 could easily test this, due to the heat spreader.