Has anybody tried this?

[Neuromancer]

Soemthing that might be simpler....

Check out the world record overclock videos for AMD. They use LHe.. since there is no way to keep it a liquid once it hits air... They blast the fog into the pot...

Now you wont hit hte -170C with it,, but perhaps a machine that either slow drips or fogs dedicated pots, air tight the case, and use a very low speed speed exhaust fan for chanellign the excess nitrogen outside (or if you have an accumulater a seperate system to recoolect and compress it.

LN2 generators are not cheap though so blowing it off might be better.

With some tweaking and experimentation you could get the drip working so that you can maintain a slightly sub-ambient temp under load. Which will drastically improve your LN2 consumption. Airtighting the box with only a very low (or even passive) exhaust system might help chill the entire box without worrying about condensation. Still probably blow through a 230L in a month.. but would be fun

Plus you could always dial it up for those 6Ghz benching runs

 

[B-Hizzle]

Soemthing that might be simpler....

Check out the world record overclock videos for AMD. They use LHe.. since there is no way to keep it a liquid once it hits air... They blast the fog into the pot...

Now you wont hit hte -170C with it,, but perhaps a machine that either slow drips or fogs dedicated pots, air tight the case, and use a very low speed speed exhaust fan for chanellign the excess nitrogen outside (or if you have an accumulater a seperate system to recoolect and compress it.

LN2 generators are not cheap though so blowing it off might be better.

With some tweaking and experimentation you could get the drip working so that you can maintain a slightly sub-ambient temp under load. Which will drastically improve your LN2 consumption. Airtighting the box with only a very low (or even passive) exhaust system might help chill the entire box without worrying about condensation. Still probably blow through a 230L in a month.. but would be fun

Plus you could always dial it up for those 6Ghz benching runs

Really? -270°C... found my new supercoolant. Hell is about to freeze over.
I don't know if I will really get to do that but LN2 may be perfect and I have more room to expand. I don't want a superfluid or 0K.
Hmm. Found a way to prevent cold bugs. Add a 3rd pipe going to all parts that won't go through LN2. Maybe a 4th for liquid Fluorinert. And a heatpad on all heatsinks for a coldbug failsafe.

 

[B-Hizzle]

Start liquid flow for low cool. Once ready switch to dry ice cooling. Once stable you can start solid flow run 1. Turn on run 1 and your DIY heatpads and turn off liquid flow. Turn off DIY heatpads as soon as the solid flow is in every water block. Once flowing good start liquid flow and quickly close solid flow to 2 config. Also start your DIY heatpads until liquid flow is in all water blocks. At the same time quickly change N2(L) reservoir 1. As soon as you are finished turn on solid flow run 2 turn off your liquid flow and turn on your DIY heatpads until solid flow is in all water blocks. That's about how it should be done.
DIY heatpads- An option to heat up the material stuck inside water blocks to come out when a new material passes through.
red on pic are pipes
blue arrows are flow and run #

 

[B-Hizzle]

| this is no triple post...
\/
ITS QUADRUPLE! Stupid internet.

 

[B-Hizzle]

Woah triple post

 

[B-Hizzle]

Woah triple post!

 

[Neuromancer]

I admit, I am lost. Im drinking though.

Still "adding heat" to a cooling loop, seems counter productive. Your not only expending enrgy to add a heting element you are wasting your cooling.

You can't cool anything technically. You can only remove heat. So remove less heat instead of adding heat. A bypass instead of a heatpad makes more sense. Add a shunt for certain temps if you need to control that to maintain a semblance or fluidity or even a specific temperature.

I still think the entire thing is nonsense. There are subzero devices in the world, super computers even. The cost to run them is ridiculous and the cooling is only part of it. I like your style but pragmatically I have to tell you it is not reasonable.

FYI Bobnova mentioned a -200C cascade.. i joked earlier, it DOES NOT EXIST. LN2 is not even -200 still I think a phase change cooler is what you are after for your setup.

And hey.. no one said the phase change cooler needs to be a direct die Make it a subzero air conditioner that keeps the whole machine at 0C.

 

[B-Hizzle]

I admit, I am lost. Im drinking though.

Still "adding heat" to a cooling loop, seems counter productive. Your not only expending enrgy to add a heting element you are wasting your cooling.

You can't cool anything technically. You can only remove heat. So remove less heat instead of adding heat. A bypass instead of a heatpad makes more sense. Add a shunt for certain temps if you need to control that to maintain a semblance or fluidity or even a specific temperature.

I still think the entire thing is nonsense. There are subzero devices in the world, super computers even. The cost to run them is ridiculous and the cooling is only part of it. I like your style but pragmatically I have to tell you it is not reasonable.

FYI Bobnova mentioned a -200C cascade.. i joked earlier, it DOES NOT EXIST. LN2 is not even -200 still I think a phase change cooler is what you are after for your setup.

And hey.. no one said the phase change cooler needs to be a direct die Make it a subzero air conditioner that keeps the whole machine at 0C.

The heaters I don't like but that cold everything is going to stick together. How will you flush out the water block after run 1. Since you have to switch out the used N2(L) reservoir (it's going to be pressurized more than the second) you have a good 30secs of downtime that must be cooled or you are going to fly passed 100C. With the heat pads and liquid Fluorinert running you may get 30C-70C because its coming out of sub 0. You will need the liquid to keep cool or you are going from sub 0 to 100C in like 20secs, I've gone from 40C to 100C in 3secs and my computer luckily had a temperature failsafe, and you need the heat pads or your water blocks are going to explode and that, 1 clog after 2secs is going to burst 4-5 pressurized tanks. And I made this now to be able to step up and step down your cooling in case of a coldbug or something. The heatpads will also recover you from coldbugs. This is also very efficient, 24/7 sub 0 temperatures (because you can't shut it off or you waste nitrogen and you have to pulverize the solid Fluorinert) I just need to find a good pump. Try making a super high power siphon starting it with a strong vacuum. 0 energy used with the siphon, it only costs money in the beginning. After you get it all, all you need to pay for is the N2(L) and Fluorinert refills. That should take a long time, and the N2(L) refills are small, only to release pressure from heated nitrogen, and it may cool back down to something very usable. I'm not even saying you don't have to use them multiple times per refill, that's just a safety concern.

Any ideas on how to pulverize the solid Fluorinert while still very cool. Maybe smash it inside a vat of N2(L) or in a cooler full of DIce after freezing?

 

[Neuromancer]

No idea how to do your crazy stuff, but before you power up and get to 100C is easy, powerup the loop before you power up the system.

 

[Bobnova]

I've seen a video with a cascade that was built to condense LN2 and be attachable to a CPU. Thing looked to be ~8' tall and maybe 12' long, don't know how thick.
Made a fair stream of LN2 though.
Most cascades "only" hit -100 to -150.
That's still cold enough to coldbug most CPUs.

I strongly recommend doing some general research into phase change systems, dry ice, ln2 physics, pumps, and thermodynamics.

Here's the short version though:
Cooling anything with LN2 will never be efficient, not if going just below 0c is the goal.
If you want to go a bit below 0c you can build an AC unit based water chiller for $50 or so, or you can buy a (even more efficient) single stage phase change unit used for $100-$350.
That will get you -20c to -50c (depends on the SS) all day every day without wasting huge amounts of power.

 

[hokiealumnus]

I've seen a video with a cascade that was built to condense LN2 and be attachable to a CPU. Thing looked to be ~8' tall and maybe 12' long, don't know how thick.

It also had five MASSIVE compressors and was so heavy it broke the first (already very beefy) wheeled cart it was on. Oh, and it had so much circuitry that it would make even a hardened engineer have nightmares.



It was truly a thing of beauty though!

Here it is actually condensing LN2.

Oh, and it's quite loud.

The build log is here.

 

[Bobnova]

Yeah, that!
I guess I got the dimensions wrong. Thanks for digging it out Hokie!

 

[ATMINSIDE]

I like the thinking that you're giving this, but have you ever taken a course in Thermodynamics or Fluid Mechanics?

 

[B-Hizzle]

I like the thinking that you're giving this, but have you ever taken a course in Thermodynamics or Fluid Mechanics?

Uh, no... lol.

 

[ATMINSIDE]

That's why you think some of these things will work then...

Pumping a powder is all but physically impossible, because of its dynamic viscosity. If you make it into a paste, its viscosity and density combined will just destroy a pump in a very short time.

There's other things that I could go into, but they're very in-depth engineering aspects.

 

[Janus67]

What you are planning/thinking of is definitely interesting, but I can't imagine it being realistic/possible.

as others have said, if you want something that can run 24/7 (or on a longer schedule than just during benchmarking) a SS (single stage) phase change system or a cascade is what you are looking for, and in the grand scheme of things is a lot more simple.

For a retail example
http://www.frozencpu.com/cat/l1/g49/Phase_Change.html?id=gudz9PV2

 

[hokiealumnus]

I say you should build a copy of that five stage cascade and run it 24/7. It would probably cost more in electricity than the entire rest of your house, but at least you could do it feasibly.

 

[ATMINSIDE]

Hokie, that would be epic for 24/7 OCing.

 

[B-Hizzle]

That's why you think some of these things will work then...

Pumping a powder is all but physically impossible, because of its dynamic viscosity. If you make it into a paste, its viscosity and density combined will just destroy a pump in a very short time.

There's other things that I could go into, but they're very in-depth engineering aspects.

Lol I've fixed that. Its kind of like crushed ice. And I'm not using a pump. I'm thinking of a vacuum or a siphon to pull the contents.

I say you should build a copy of that five stage cascade and run it 24/7. It would probably cost more in electricity than the entire rest of your house, but at least you could do it feasibly.

I'd like to but I'm kind of new to cascades and I can't really see how it all works.

 

[Bobnova]

Vacuum/siphon powered by what?