Weird idea, requesting input

[HAL-9000]

About a year ago I had an idea for a case that would use R707 (CO2)refrigerant. The idea was that you would just have the coolant physically evaporate off of the selected bits for focused cooling via tubes exiting directly on exposed silicon. Rather than have a return feed like a Vapochill back to the compressor, the evaporated refrigerant would just blow around in the sealed case and the compressor inlet would be strategically placed in the enclosure to draw the evaporated coolant (which would still be pretty chilly) across all the system components, thus cooling them while heading for the compressor inlet. The advantages would be the case would not interact with the outside atmosphere, be humidity free, and not needing any openings you could seal it up so it would be dead quiet in operation.

You could use less extreme coolants than R707 in such a setup, but its so chilly (-77C) and its heat capacity is so good. Plus, with R707 you could match the internal pressure to the local ambient by either purging the system of some coolant, or "priming" it to higher pressures using common CO2 cartridges for BB guns with an identical valve used on the gun.

The tricky part that kept me from ever trying to specifically design/build such a contraption was that sealing the backplane part of the computer where all the cards and I/O goes without making the install permanant would be a sunbitch. Plus, it would take some serious kilowattage for a CO2 compressor that could deal with ALL the heat of a modern PC. Oh yeah, I was also relatively dirt poor when I thought of it. Lately though, I've got the cool chips bug again and am considering possibley building something like this. Anyone have any ideas on sealing up the backplane?

[bigben2k]

Yeah, you have to replace the connectors, with pressure proof ones, hence the sunbitch...

[jaydee]

Could you use some Neoprene insulation and cut it to fit around the backplane? If you had the right thickness you could probably make a decent pressure seal when you mount the mobo.

[killernoodle]

Dont use the backplane: use a few extension cords and find a way to seal around them (I'm thinking a lot of hot glue.)

[maxSaleen]

hmm.... I'm not a fan of the whole kilowattage thing...

Killernoodle was on target with his idea. I've considered building an air-conditioning computer before where it would be sealed from the outside world and that was the idea I came to. You take extension cords from the backplane to some sort of home made panel that you can reasonably seal.

For example:

I'd take a sheet of aluminum and coat the back of it in rubber first and then neoporene. From there I'd screw in what cables I could (VGA, DVI, Serial, etc.) USB and audio cables will be a little more difficult. Use lots of silicone based sealants and some creativity. It can be done.

Your Problem:

The case would get WAY WAY WAY WAY too cold. You say "why is this a problem? Hell, its my goal." Your mofsets and caps won't like the cold. Nor will your RAM. Nor will your PSU should it go in there. I was surprised to. I threw an idea up on the ExtremeSystem forums about a year and a half back for a system suspended in flourient (a non conductive, non toxic fluid made by 3M, used on CRAY computers, very expensive) and then chilled be a phase change system. The general consensus was that the only parts of the system that would take well to the cold would be the CPU, GPU, and NB. From that respect it would be much more effective to cool using a liquid chiller. It would produce better temps and be more cost effective (not to mention more energy efficient).

CRAP!!!! Sorry for the rambling. You guys will get used to it in time.

I hope this information will be useful to you.

[bigben2k]

Originally Posted by killernoodle

Dont use the backplane: use a few extension cords and find a way to seal around them (I'm thinking a lot of hot glue.)

It'll leak through the wire insulation.

Originally Posted by jaydee116

Could you use some Neoprene insulation and cut it to fit around the backplane? If you had the right thickness you could probably make a decent pressure seal when you mount the mobo.

There's no way it'll hold any heavy pressures.



I didn't even think about what maxSaleen mentionned, and he's right. The other chips aren't designed to run under the freezing point (typically); wether or not they work will be a surprise. The caps could be replaced, but swapping the MOSFETs with mil grade ones might be a sunbitch...

[jaydee]

Heavy pressures? If the pressure is that heavy it will cause other problems long before the neoprene goes out. Mobos and the parts on it are not going to handle to much pressure.

[MadHacker]

I read somewhere a cople years ago about a guy that tried to submerge his MB in fluorinert and use liquid nitrogen to cool the fluorinert.
he seamed to get it working but then the fluorinert got to cold and became a gel. it would not pump through any more so then the system died? don't remeber exacly what happened in the end.

[andy497]

Originally Posted by HAL-9000

About a year ago I had an idea for a case that would use R707 (CO2)refrigerant. The idea was that you would just have the coolant physically evaporate off of the selected bits for focused cooling via tubes exiting directly on exposed silicon. Rather than have a return feed like a Vapochill back to the compressor, the evaporated refrigerant would just blow around in the sealed case and the compressor inlet would be strategically placed in the enclosure to draw the evaporated coolant (which would still be pretty chilly) across all the system components, thus cooling them while heading for the compressor inlet. The advantages would be the case would not interact with the outside atmosphere, be humidity free, and not needing any openings you could seal it up so it would be dead quiet in operation.

You could use less extreme coolants than R707 in such a setup, but its so chilly (-77C) and its heat capacity is so good. Plus, with R707 you could match the internal pressure to the local ambient by either purging the system of some coolant, or "priming" it to higher pressures using common CO2 cartridges for BB guns with an identical valve used on the gun.

The tricky part that kept me from ever trying to specifically design/build such a contraption was that sealing the backplane part of the computer where all the cards and I/O goes without making the install permanant would be a sunbitch. Plus, it would take some serious kilowattage for a CO2 compressor that could deal with ALL the heat of a modern PC. Oh yeah, I was also relatively dirt poor when I thought of it. Lately though, I've got the cool chips bug again and am considering possibley building something like this. Anyone have any ideas on sealing up the backplane?

I'm not sure what low pressure side working pressure would be, but I seem to recall that keeping C02 liquid around room temperature requires something like 500-700 psi at the very minimum. Those 12 gram bb gun cartridges are 1800 psi. I would guess that the strongest case would be quickly destroyed by 1 psi or so (even that would be hundreds of pounds on a side panel), so this might be a little further than you'd think.

[HAL-9000]

Originally Posted by andy497

I'm not sure what low pressure side working pressure would be, but I seem to recall that keeping C02 liquid around room temperature requires something like 500-700 psi at the very minimum. Those 12 gram bb gun cartridges are 1800 psi. I would guess that the strongest case would be quickly destroyed by 1 psi or so (even that would be hundreds of pounds on a side panel), so this might be a little further than you'd think.

The CO2 would never be liquid in the case, as it would have been vaporized into a gas by deposition on the heat source (CPU core).

I do not think the vaporized gas would be warmer than 5C though, which would be a problem for certian components as alluded to by maxSaleen. Another issue would be the copper pipes necessary to carry the Co2 liquid to the heat source. It would be hard to re-align them depending on the motherboard and processor.

What I was thinking with those cartridges was that since the case would be hermetically sealed basically, it doesn't naturally adjust to the ambient pressure. You could use the cartridges modify the case's internal pressure to match the local ambient, that's all.

[MadHacker]

just an idea but what if there would be a double layer instead of just one layer shell...
perhaps the preasure would be easier to control?
the diffrence of the presure between layer would be less...
just a thought..

[Emperor Que]

With the project that I did I used a Nitrogen atmosphere in a semi sealed container, and had the evap inside.

Good Idea though but I think it might be to hard to maintain a sealed container, with the right amount of insulation.

hmmm.... continues to think......


Later GL

[redleader]

Lot of problems with this idea:

1) CO2 is R744 not R707

2) You can't condense CO2 in a single phase system, so you'll need to use some sort of cascade. Take a look at the condesing pressure of CO2 to get an idea what I mean.

3) CO2 is a pain to work with since it becomes dry ice rather easily.

4) You can't possibly expect a standard case (let alone one using hot glue!) to withstand that kind of pressure. Yes your evaporating pressure would only be about 1 ATM. But were you ever planning on turning it off and letting the low side be pulled to high side? Would the entire system detonating as soon as theres a power failure and the compressor switches off be a problem for you? If so . . .

5) You're going to be a LOT colder then 0C. More like -70 to -100 C depending on what your evaporating pressure is. You'll probably kill the board with those temps.

[HAL-9000]

Originally Posted by redleader

Lot of problems with this idea:

1) CO2 is R744 not R707

2) You can't condense CO2 in a single phase system, so you'll need to use some sort of cascade. Take a look at the condesing pressure of CO2 to get an idea what I mean.

3) CO2 is a pain to work with since it becomes dry ice rather easily.

4) You can't possibly expect a standard case (let alone one using hot glue!) to withstand that kind of pressure. Yes your evaporating pressure would only be about 1 ATM. But were you ever planning on turning it off and letting the low side be pulled to high side? Would the entire system detonating as soon as theres a power failure and the compressor switches off be a problem for you? If so . . .

5) You're going to be a LOT colder then 0C. More like -70 to -100 C depending on what your evaporating pressure is. You'll probably kill the board with those temps.

I think you can condense CO2 in a single phase system. I am aware the pressure is 700psi, but you're seeing this refrigerant (R744 apparently) being implemented in cars for the express purpose of getting rid of other refrigierants due to the relative greenhouse gas/toxic issues of other refrigerants such as R134a. European regulations are set to ban these other refrigerants from being legal in autos there, so there is some considerable work being done on compact, portable implementations of this technology from the compressor side. I am not sure how intense of a power draw will be going on with such a system, but at least 1 HP I would guestimate, probably more.

Secondly, it should seem pretty obvious any implementation of this would not use a standard case of any kind, with hot glue to seal it no less. I have no idea of what left you with that impression from this discussion of that being a valid concept to begin with.

[maxSaleen]

Could you get away with a single stage system.... if they are doing it in autos, I would assume so

Could you deal with a cascade system.... yes, though it would be a pain.

Could you deal with sealing the case... yes, though it would be major pain. (I have more ideas for sealing the case if you want them, BTW)

Could you find a way to prevent dry ice from forming... probably

Could you find a way to deal with the caps, mofsets, and memory freezing to the point where they are not functional......no. Don't see any (reasonable) way to do it. From that standpoint this idea is dead.

Maybe you could use C02 in a cascade chiller? At least we know you could somehow make that work. Yes, it would be cold. Yes, you would be using a rediculous amount of electricity. Yes, it would be EXTREME.

[HAL-9000]

Originally Posted by maxSaleen

Could you get away with a single stage system.... if they are doing it in autos, I would assume so

Could you deal with a cascade system.... yes, though it would be a pain.

Could you deal with sealing the case... yes, though it would be major pain. (I have more ideas for sealing the case if you want them, BTW)

Could you find a way to prevent dry ice from forming... probably

Could you find a way to deal with the caps, mofsets, and memory freezing to the point where they are not functional......no. Don't see any (reasonable) way to do it. From that standpoint this idea is dead.

Maybe you could use C02 in a cascade chiller? At least we know you could somehow make that work. Yes, it would be cold. Yes, you would be using a rediculous amount of electricity. Yes, it would be EXTREME.

You know, the idea of using such a system is more of a thought experiment than anything else really. While I was thinking of this idea today writing my previous response, one other problematic issue occurred to me.

Assuming all these other issues are taken care of, one issue remains: namely, how do you turn the thing on?

If you try and turn the processor on after the thing has reached "operating temperature" you will thermally shock that processor, and probably crack the die in the process. Obviously, if you turn the processor on first, it will be dead before the coolant system will matter. Maybe a speedstepping proc would make a difference if its supported, but the processor won't last long getting cycled up and down like that anyways. This would even matter between running at idle and max is the gradient were steep enough. The third option is a gradual rise in either the cooling system capacity, or the processor speed. While that idea could be made to work with a ton of money and work, it doesn't matter.

No matter what you do you end up with a rig like what I see posted all the time at xtremesystems.org. Basically a top fuel dragster that can't do anything more useful than post 3D Mark scores. Not what I'm after for sure.

[jaydee]

Originally Posted by HAL-9000

No matter what you do you end up with a rig like what I see posted all the time at xtremesystems.org. Basically a top fuel dragster that can't do anything more useful than post 3D Mark scores. Not what I'm after for sure.

Whaaa? There are other uses for computers? I thought surfing the net and 3D Mark were it.

[redleader]

Originally Posted by HAL-9000

I think you can condense CO2 in a single phase system.

If you live at the south pole, and its winter, sure. Otherwise, given taht its critical temperature is a few degrees above room temp, you have a problem.

I am aware the pressure is 700psi

I'd say more like 1000 psi.

, but you're seeing this refrigerant (R744 apparently) being implemented in cars for the express purpose of getting rid of other refrigierants due to the relative greenhouse gas/toxic issues of other refrigerants such as R134a. European regulations are set to ban these other refrigerants from being legal in autos there, so there is some considerable work being done on compact, portable implementations of this technology from the compressor side.

I'm not saying its impossible to condense CO2, I'm saying you won't be able to get a compressor that will do it. Simple as that. And I really really doubt Europe is using straight 744 in cars. Given that its critical temperature around room temp, that would make it pretty useless (who wants an AC that only works during the winter and costs a few grand?). They'll probably develope some blend with 10 or 20 percent CO2, and some other gases making up that rest.

I am not sure how intense of a power draw will be going on with such a system, but at least 1 HP I would guestimate, probably more.

Power has nothing to do with this, and using a larger compressor will not address the condensing pressure issue.

Could you get away with a single stage system.... if they are doing it in autos, I would assume so

Could you deal with a cascade system.... yes, though it would be a pain.

Search on google. People have been building lots of cascaded systems with CO2. Thats not special. The difference is, they condense it at -40C because its not a first stage gas.

Assuming all these other issues are taken care of, one issue remains: namely, how do you turn the thing on?

If you try and turn the processor on after the thing has reached "operating temperature" you will thermally shock that processor, and probably crack the die in the process. Obviously, if you turn the processor on first, it will be dead before the coolant system will matter. Maybe a speedstepping proc would make a difference if its supported, but the processor won't last long getting cycled up and down like that anyways. This would even matter between running at idle and max is the gradient were steep enough. The third option is a gradual rise in either the cooling system capacity, or the processor speed. While that idea could be made to work with a ton of money and work, it doesn't matter.

No matter what you do you end up with a rig like what I see posted all the time at xtremesystems.org. Basically a top fuel dragster that can't do anything more useful than post 3D Mark scores. Not what I'm after for sure.

You're not going to thermally shock a processor by slowly cooling it and then powering it on. The limitation here is how well your system works and for how long it can keep running.

[Dukemurmur]

Originally Posted by MadHacker

I read somewhere a cople years ago about a guy that tried to submerge his MB in fluorinert and use liquid nitrogen to cool the fluorinert.
he seamed to get it working but then the fluorinert got to cold and became a gel. it would not pump through any more so then the system died? don't remeber exacly what happened in the end.

Not that the fluorinert gelled it wasnt cold enough for him he dumped the LN2 right into it...then it got something so cold that it was all corrupted...but hey why not!! ...i would like to try it though just for kicks...

[Fimbrius]

These drastic temperature changes will stress the die and package considerabley. Although not "shocking" the material properties of the package are not very forgiving to below zero temperatures coupled with rapid rises in core temperatures. This will create a significant stress gradient that will significantly reduce the life of the package. I think it would be safer to use a heater to bring the die, unpowered, to a certain temperature (say 30C) while the system powers up. This way the package and the core could be cooled simultaneously. Just my 2c.

[redleader]

Originally Posted by Fimbrius

These drastic temperature changes will stress the die and package considerabley. Although not "shocking" the material properties of the package are not very forgiving to below zero temperatures coupled with rapid rises in core temperatures. This will create a significant stress gradient that will significantly reduce the life of the package.

What drastic temperature changes? Its going to take a little while for a 2 stage system to pull down far enough that you could even turn on the system without burning out the chip. The chip itself is undergoing much sharper temperature gradients under normal use.

Hit google. This is not an issue.

I think it would be safer to use a heater to bring the die, unpowered, to a certain temperature (say 30C) while the system powers up. This way the package and the core could be cooled simultaneously. Just my 2c.

That doesn't make sense. If you use a heater to keep the die at a temperature above your evaporator, you have just defeated the point of having the evaporator. And no you couldn't just have the heater on during power up because the system most likely will not POST above subambient temperatures. Remember, the point is overclocking, not to have a computer at stock speed that is uselessly cooled to subambient temperatures.

[Fimbrius]

Two points of clarification. Whether you slowing decrease or increase temperatures, the delta T's between steady state on vs off can be quite large. Package degradation mechanisms increase significantly with lower temperatures. The second point is that by using a heater, during startup only, you can reduce the die to package delta T's. Once the die is powered, then power down the heater.

Just trying to point out the disadvantages for turning this thing off, not trying to be hostile. Most of these packages are not really evaluated for less than -25 usage/cycling.

[redleader]

The second point is that by using a heater, during startup only, you can reduce the die to package delta T's. Once the die is powered, then power down the heater.

The entire package is at the same temp as the core (since they're in direct physical contact), so I don't see what the heater is supposed to accomplish besides preventing you from POSTing while overclocked.

Just trying to point out the disadvantages for turning this thing off, not trying to be hostile. Most of these packages are not really evaluated for less than -25 usage/cycling.

I don't think anyone who is OCing really cares what the rated tolerances are. If they did, they wouldn't be doing this in the first place. Experimentally this has been shown to not be an issue.

[Fimbrius]

It has been verified quite a bit on the effects of low temperatures coupled with large delta T's on package reliability performance. Take a look at the JEDEC 22 and Milspec 883 standards. Normal Temperature cycle B is an accelerated stress test than can go as low as -55C to 125C and is used to evaluate a usage condition delta T of 40C from Room Temp (Intel's website has some interesting papers on package reliability performance). Below -25C temperatures coupled with 50C cycling shifts have been experimentally verified to have a negative impact on product life. I would completely agree that most OC's don't care about this as whats the differance between 10yr rated reliability performance vs 3yr or 2yr (it will be old tech by then). Whether we care or not it does occur and my point was to minimize the number of times it is cycled to room temp from such a low operating condition (meaining -40 or less) in order not to place yourself it that risk catergory.

As for the heaters, you have a valid point about the shock affect not cause instantaneous problems. Once again though, its the temperature gradient and time to temp that affect the reliability performance. The heater would be used to control the gradient via a PID that could power down once a certain POST code is reached.