daishi's drunken ramblings

[DFI Daishi]

so, i recently spoke to an individual who said that he had the ultimate master plan for direct die cooling.

he said that he was going to strip the heat spreader off of a prescott, and then rig up a phase change system such that the die itself would be the evaporator, and then the refrigerant would be drawn out of the custom enclosure and re-cycled through the loop.

i said that it was a pretty silly idea, becase the die does not have enough surface area to effectively act as an evaporator surface, and that the close contact between hot surface and cold coolant wouldn't mean a thing because the coolant would hit evaporation temps before it actually contacted the die, thus reducing overall cooling efficiency.

he said that the close contact between the coolant and the die would be the ultimate, and that i am overestimating how easily the refrigerant will evaporate. he said that he would score the surface of the die with a razor blade, and that it would then have plenty of surface area.

i really think that this is a very stupid idea, but could you please give arguments for and against, please.

 

[gclg2000]

I think i understand what your trying to say, but not sure. However, if he is going to try to enclose the area and focus the freon onto the die it would have to be literally brazed shut or it will leak. Also, the evaporator will take the shortest path back to the suction. So he unless somehow he makes the suction really close to the die of teh chip, it won't even get close .I think the way its done now w/ normal direct die is the only option.

I still don't quite understand what he's trying to do. but, the biggest arguement i would make is that the copper dispenses the temperature of the evap head very, very well. He would maybe gain 5 degree's C if doing this way, that is a well designed evap.

Besides, cascade is colder than one single stage even if he could achieve this, the cascaded compressor's would be colder and hold more load anyways.

 

[fore1337]

I agree with the cascade being a better aproach.
The greater the temperature difference (delta T), the greater the heating process (or heat transfer Q).

It sounds like a great way to fry a chip though.

 

[DFI Daishi]

so far you guys seem to be in agreement with me. this was just a conversation at the campus bar with a friend of a classmate, and the only diagrams were what buddy drew on a napkin.

now just a few points:
glcg2000: by "normal" direct die, i assume that you mean circulating a liquid coolant directly agains the die of the processor. is the correct? in this proposed systsm, the closer the freeon nozzle(or tube) gets to the die, the less area it can cool, so he will probably hit a wall with that, right?

fore1337 and glcg2000: my understanding of the way that a cascade works is that one compressor's cold coil cools a second compressor's hot coil, and the cold coil of the second compressor is used for cooling. if that is correct, then there is no reason that he couldn't do this project with a cascade. he did not say that he was going to, i'm just saying that particular argument might not be the best point against.

i don't phase change cool myself, so my statements this round of conversation weren't all that certain. i'm just trying to get my stuff down if we ever end up chatting again.

it's damn hard to find another reasonably hardcore overclocker at the campus...........

 

[Peach]

is it just me or that if the person was to score the die with a razor blade it would kill the CPU?

 

[Jonsey]

First, you arn't getting the heat spreader off a prescott. Second, there is no way to seal an enclosure that has as its bottom a cpu. If you could get the heat spreader off the cpu without taking half the core with it (unlikely); and you managed to somehow seal the cpu into a evaporator enclosure that could handle 100 PSI of pressure (impossible IMO), then you could focus the capillary tube outlet on the die, which would allow liquid refrigerant to spray on to the cpu. However, you would have very little capacity (most direct-die evaporators have lots of surface area for the refrigerant to fully evaporate, you would have almost none in your friend's brainstorm). Also, the mass of the copper in a direct die evaporator protects the cpu from wild temperature fluctations as it goes from load to idle. A regular copper evaporator also allows the cpu to cool (relatively) slowly. How do you think a cpu core would respond to a blast of liquid refrigerant hitting it directly on the die? I would guess the lifespan of the cpu would be reduced.

If he still doesn't like your arguments, let him build it. Then you can make fun of his ruined Prescott and wasted money. If somehow it does work out, take some of the credit when he patents it and makes a billion selling it to Intel, and maybe he'll give you a new system every once in while.

 

[gclg2000]

glcg2000: by "normal" direct die, i assume that you mean circulating a liquid coolant directly agains the die of the processor. is the correct? in this proposed systsm, the closer the freeon nozzle(or tube) gets to the die, the less area it can cool, so he will probably hit a wall with that, right?

Direct die is evaporating the liquid inside the copper block, which resides against the CPU. You idealy want the point of flashing (evaporating) to be right over the CPU, for that is the "coldest" point. That gas is then normally circulated around the evap to the suction and then it returns to the compressor.

You've got the right idea about cascading. It's like what you described but with a whole lot of junk in the middle. It's main job is combain pressure and cold temps to condense a gas. Once you can condense a gas (liquify it), you can meter it and then flash it to an evaporator.

This is just a general idea of how it works.

Oh, and tell this guy it won't work out as well as he hopes. It's much safer to do it the normal way. Not only to him, but to the environment and his roomates and people that live around him.

What if he uses a flammable gas? What happen's when it leaks out and a ball of fire set's of teh carpet in his room and then start's burning the building down. What happens if he somehow get's ahold of a controlled freon and just let's it spew into the ozone? These are things that have laws for a reason.

It's not long before the phase change thing REALLY becomes a giant fad. And everyone start's to try to build these things. I hate to say it, but we will be reading someday about the noob that blew himself up.

 

[DFI Daishi]

i guess this is your direct die phase change, vs my direct die liquid cooling background happening here.

i was thinking circulating water-antifreeze mix against the proc when i heard you say direct die cooling.

the first thing that i told this guy was that he was just asking to poison himself, with this plan, and he was rather blasee about it.

 

[BrainEater]

It sounds like a great way to fry a chip though.

Yep...haha

especially when the score marks from the razor become thermal fracture zones.

 

[wtiger]

Is it just me that thinks this, but if he somehow managed to get it to work wouldn't you risk cracking the core spraying a supercooled refrigerant on it? The thing would go from room temp or better to well below zero in a fraction of a second. The few times I've played around with dry ice I can't help but forget how it wildly destorted anything you touched it too; because of the wild temp change. you would think doing something like that to something with a crystaline structure (core) would cause it to shatter.

 

[toymachineman19]

Is it just me that thinks this, but if he somehow managed to get it to work wouldn't you risk cracking the core spraying a supercooled refrigerant on it? The thing would go from room temp or better to well below zero in a fraction of a second. The few times I've played around with dry ice I can't help but forget how it wildly destorted anything you touched it too; because of the wild temp change. you would think doing something like that to something with a crystaline structure (core) would cause it to shatter.

Yeah, with dry ice, I was playing with it, and I put it in a beaker, and it went from room temp to really cold so fast the glass was shaking, and didn't stop 'til the glass got much colder, it almost broke the beaker too. It does very strange things to metal too..

 

[Megadeth_Guy01]

nvm

 

[Lite_User]

He'll never get the heat-spreader of the chip,if he does, it will crack at startup, due to fast and big temp changes.
That's why the nice people who make those CPU's, seat a heatercore for us.
We go from there.
Nice dream

Grtz