Underpressure

[Nuwat]

Budget (including currency):  an absurd amount

Country: Brazil

Games, programs or workloads that it will be used for: honestly, if it turns on it would be a miracle

Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc): 

 

 

So... I had an interesting idea after watching the video about the 6.9Hz processor. I believe it would be possible to modify a vacuum chamber with the contacts for the output of the computer and the charging part so you could maintain a proper seal and not have to worry about that. And have 3 holes in the vacuum chamber, one for pulling the vacuum, one for the intake of nitrogen, and one for releasing the pressure... Under pressure you would be at room temperature and it would be possible to control the temperature inside with the release valve (kinda like a deodorant), there would be no humidity inside so you couldn’t have to worry about that. Warming it back up would just need to close the valve. I believe it also would represent a less wastage of nitrogen too. Absolutely no idea how much this could cost and most definitly not planning on doing so. But here is a good challenge...

[tkitch]

won't work.  The LN2 would kill any vacuum unless you had a suuuuuuper high end pump that could stay ahead of it.

 

Also, the vacuum would cause the LN2 to evaporate faster.

[Nuwat]

13 minutes ago, tkitch said:

won't work.  The LN2 would kill any vacuum unless you had a suuuuuuper high end pump that could stay ahead of it.

 

Also, the vacuum would cause the LN2 to evaporate faster.

No, I mean just for removing the air first so there is no water, thinking now u would just need 2 holes, one for the intake, and one for the exhaust (that u would have to pull a vaccum first.)

[RONOTHAN##]

This is not a super new idea, K|NGP|N has made something like this, believe he called it the "Roboclocker." There was a revision that had everything in a vacuum chamber, though can't find it in my 5 minutes of Googling.

[Nuwat]

10 minutes ago, RONOTHAN## said:

This is not a super new idea, K|NGP|N has made something like this, believe he called it the "Roboclocker." There was a revision that had everything in a vacuum chamber, though can't find it in my 5 minutes of Googling.

No idea about this roboclocker, but I meant more like, no tubes, no vent , no heat sinks, no nothing, and u wouldnt have to worry about the vacuum that would be just for removing the air and water. I meant more like literally placing a computer inside a nitrogen tank, it would cool down with the same effect a deodorant cools down when u open the release for too long...

[RONOTHAN##]

2 minutes ago, Nuwat said:

I meant more like literally placing a computer inside a nitrogen tank, it would cool down with the same effect a deodorant cools down when u open the release for too long...

That won't work. A lot of the parts on a computer don't actually like running below 0C. A lot of RAM doesn't like getting too cold, Alder Lake CPUs won't boot if they're below ~-140C (I forget the exact number, it's somewhere around there), VRMs usually have their voltage regulation go to hell as they get further and further below 0C.

 

There's more reason than just condensation why people use board heaters when doing XOC, the only parts on a system that likes being cold are the CPU, GPU, and sometimes RAM. Everything else wants to run at ~20C, plus all of those have different ideal temperatures (colder doesn't always mean better), so doing that method wouldn't let you have the GPU running at -130C, the CPU at -160c, and the RAM at -100C like what might be ideal for what you're trying to do. 

[Eigenvektor]

20 minutes ago, Nuwat said:

…it would cool down with the same effect a deodorant cools down when u open the release for too long...

That effect only lasts while there's gas in there to expand. Ideally you'd want a closed loop, where gas can expand to cool down, then gets compressed again (heating it up) somewhere else and then recirculate it. That's effectively a heat pump, i.e. how a refrigerator works. Pretty sure you still want a heat sink, to increase the surface area of the CPU, to be more efficient at transferring its heat.

 

29 minutes ago, RONOTHAN## said:

This is not a super new idea, K|NGP|N has made something like this, believe he called it the "Roboclocker." There was a revision that had everything in a vacuum chamber, though can't find it in my 5 minutes of Googling.

 

[Somerandomtechyboi]

The only subzero systems that can daily at really low temp is phase change, single stage usually gets around -50c but you can cascade phase change which is just putting multiple phase change systems in series so chilling multiple times for lower temp, but these are expensive asf, draw crap loads of power, and there are very few builders for these things out of the already few that build phase change systems (linus ac/pc video is a phase change cooling system)

 

Theres also waterchiller which is just take an ac and put the evap into a bucket of antifreeze or windshield washer fluid and chill it to subzero (basically what linus used for chilling threadripper 1 and 2), but usually quite abit weaker than phase change. I plan on going this route in the future by just cannibalizing a used 60-100w refrigerator and using that to chill rams and northbridge

[Nuwat]

20 minutes ago, RONOTHAN## said:

That won't work. A lot of the parts on a computer don't actually like running below 0C. A lot of RAM doesn't like getting too cold, Alder Lake CPUs won't boot if they're below ~-140C (I forget the exact number, it's somewhere around there), VRMs usually have their voltage regulation go to hell as they get further and further below 0C.

 

There's more reason than just condensation why people use board heaters when doing XOC, the only parts on a system that likes being cold are the CPU, GPU, and sometimes RAM. Everything else wants to run at ~20C, plus all of those have different ideal temperatures (colder doesn't always mean better), so doing that method wouldn't let you have the GPU running at -130C, the CPU at -160c, and the RAM at -100C like what might be ideal for what you're trying to do. 

right... but underpressure it would be at room temperature, the computer could boot no problem... the liquid nitrogen being cold is a just cuz it is boiling at room temperature and evaporating. The thing I was wondering is the end valve, cuz as the cpu and gpu heat up they would make the pressure rise faster, so you would need to regulate the exhaust according to what u need.

[Nuwat]

19 minutes ago, Somerandomtechyboi said:

I plan on going this route in the future by just cannibalizing a used 60-100w refrigerator and using that to chill rams and northbridge

I wonder how effective it would be, cuz the freezer only works cuz there is no heater inside the freezer... The thing I was wondering would be super expesive to run and require an infinite nitrogen tank on hand,... it could make sense for a supercomputer perhaps not to have at home.

[Nuwat]

26 minutes ago, Eigenvektor said:

Pretty sure you still want a heat sink, to increase the surface area of the CPU, to be more efficient at transferring its heat.

And u are absolutely right... a little oversight of mine.

[Nuwat]

29 minutes ago, Eigenvektor said:

That effect only lasts while there's gas in there to expand

Also... I am aware, not saying it would be cheap by any means...you could recicle the nitrogen tho, but perhaps that could limit the amount of pressure u can release, limiting the cooling.

[RONOTHAN##]

9 minutes ago, Nuwat said:

right... but underpressure it would be at room temperature, the computer could boot no problem... the liquid nitrogen being cold is a just cuz it is boiling at room temperature and evaporating. The thing I was wondering is the end valve, cuz as the cpu and gpu heat up they would make the pressure rise faster, so you would need to regulate the exhaust according to what u need.

You still run into the issue of cooling things that don't want to be cooled, and you can't cool parts to exactly the degree that you want them to. 6900 XTs, for example, will just stop working if you bring them below ~100C. Samsung B die (high performance DDR4) does very well at certain memory temps that depend on that particular stick, sometimes it's 20C, sometimes it's -50C, sometimes it's lower, yet DDR4 doesn't draw enough power to lower the temps enough to boil liquid. VRMs, on the other hand, output more heat than memory but hate being cold. 

 

This just isn't a concept that would make a ton of sense. In order to have the granular temperature control over each part individually that XOC needs this just won't work, plus it would be insanely expensive. The Roboclocker is about as close to something like this as you can get, but it's actually somewhat practical. 

[Nuwat]

1 minute ago, RONOTHAN## said:

You still run into the issue of cooling things that don't want to be cooled, and you can't cool parts to exactly the degree that you want them to. 6900 XTs, for example, will just stop working if you bring them below ~100C. Samsung B die (high performance DDR4) does very well at certain memory temps that depend on that particular stick, sometimes it's 20C, sometimes it's -50C, sometimes it's lower, yet DDR4 doesn't draw enough power to lower the temps enough to boil liquid. VRMs, on the other hand, output more heat than memory but hate being cold. 

 

This just isn't a concept that would make a ton of sense. In order to have the granular temperature control over each part individually that XOC needs this just won't work, plus it would be insanely expensive. The Roboclocker is about as close to something like this as you can get, but it's actually somewhat practical. 

Yeah... u are probably right... maybe you could try to place a heat pipe from the CPU to the parts that dont like the cold that way they can stay warmer overall, which is very counter intuitive... and maybe require u to run the cpu under a bit of stress to keep stuff warm.  Also... not a concept that makes sense, just wondering if that would even be possible

[RONOTHAN##]

6 minutes ago, Nuwat said:

Yeah... u are probably right... maybe you could try to place a heat pipe from the CPU to the parts that dont like the cold that way they can stay warmer overall, which is very counter intuitive... and maybe require u to run the cpu under a bit of stress to keep stuff warm.  Also... not a concept that makes sense, just wondering if that would even be possible

Basically, the way you're describing it, it would have to be specifically engineered for whatever the exact hardware config you're trying to run. If it's possible, it's barely possible (don't have enough education in thermodynamics to know specifically if it's possible)

[Nuwat]

25 minutes ago, RONOTHAN## said:

Basically, the way you're describing it, it would have to be specifically engineered for whatever the exact hardware config you're trying to run.

the whole thing would have to kinda be engineered lol... doubt anyone would have a computer case can pulled a vaccum laying around... now the multi phase kinda ordeal is making a bit of sense.