2 Phase Immersion cooling for the home

[jimmothy]

Anyone know if this is practical?

Edit: https://www.youtube.com/watch?v=ivVoANqFBuY (nice video about 2 phase)

[1714160729]

1714160729

[1714160729]

1714160729

[1714160729]

1714160729

[1714160729]

1714160729

[Dexter770221]

If you want to heat your house then not so practical. To be efective hot water that enters the heaters must have high temperature like 65C and more. And with immersion cooling you want to cool chips with liquids with boiling point below 40C. To warm up water for bathing then it may work with liquids with boiling point at ~50C. My 2 cents.

[jimmothy]

If you want to heat your house then not so practical. To be efective hot water that enters the heaters must have high temperature like 65C and more. And with immersion cooling you want to cool chips with liquids with boiling point below 40C. To warm up water for bathing then it may work with liquids with boiling point at ~50C. My 2 cents.

This would be for 1kw+ systems so that would probably be too much for a heater. I am wondering if a scaled down version of allied controls system would work. (Water cooled condenser and outdoor radiator)

[aerobatic]

i spoke with alex at allied control about doing a scaled down version for the home and he thought there would be lots of problems and that its best kept at an industrial scale where it can be monitored by professionals and that the potential for escaping vapour wont present a problem.

[jimmothy]

Anyone know if this is practical?

You have check out the other thread on the small system for 2 phase right?

https://bitcointalk.org/index.php?topic=255613.40

http://www.youtube.com/watch?v=_pNSUKp8ov4

Saw it but it looks somewhat inefficient/inpractical. It looks like that massive fishtank is only cooling something like 100W worth of bitfuries.

And would novec 7000 be the best fluid? I would think it would not be optimal with such a low boiling point. Wouldn't it require much colder water to condense the vapor? Not sure how well it would work on hot summer days.

[jimmothy]

Im confused how that system works.

Is it 2-phase? If so where is the condenser?

[aerobatic]

2 phase inside the board side and a water block on the other.

The novec is boiled and condensed changing phase on one side of the sealed unit. The others side removes heat with water rad system. If you have a mining board that conforms to a the Iceotope form factor it could be ready now. You would still need the server modular units though check this out: http://www.youtube.com/watch?v=lHHHYcdPFnQ

iceotope isnt 2-phase, theyre using the novec without boiling it (may as well be mineral oil).  theyve got pumps and radiators inside each blade.  makes no sense.  ignoring the best bit of novec tech.

[jimmothy]

2 phase inside the board side and a water block on the other.

The novec is boiled and condensed changing phase on one side of the sealed unit. The others side removes heat with water rad system. If you have a mining board that conforms to a the Iceotope form factor it could be ready now. You would still need the server modular units though check this out: http://www.youtube.com/watch?v=lHHHYcdPFnQ

iceotope isnt 2-phase, theyre using the novec without boiling it (may as well be mineral oil).  theyve got pumps and radiators inside each blade.  makes no sense.  ignoring the best bit of novec tech.

Even if its not 2 phase I am interested in how they can unattach the server without leaking water.

What is the name of magical this part??

[aerobatic]

2 phase inside the board side and a water block on the other.

The novec is boiled and condensed changing phase on one side of the sealed unit. The others side removes heat with water rad system. If you have a mining board that conforms to a the Iceotope form factor it could be ready now. You would still need the server modular units though check this out: http://www.youtube.com/watch?v=lHHHYcdPFnQ

iceotope isnt 2-phase, theyre using the novec without boiling it (may as well be mineral oil).  theyve got pumps and radiators inside each blade.  makes no sense.  ignoring the best bit of novec tech.

Even if its not 2 phase I am interested in how they can unattach the server without leaking water.

What is the name of magical this part??

you can buy drip free 'quick connects' from lots of places ...

cheap ones from koolance.. but there are much better ones available

http://www.xtremerigs.net/2013/07/02/2013-quick-disconnect-roundup/

[seriouscoin]

2 phase inside the board side and a water block on the other.

The novec is boiled and condensed changing phase on one side of the sealed unit. The others side removes heat with water rad system. If you have a mining board that conforms to a the Iceotope form factor it could be ready now. You would still need the server modular units though check this out: http://www.youtube.com/watch?v=lHHHYcdPFnQ

iceotope isnt 2-phase, theyre using the novec without boiling it (may as well be mineral oil).  theyve got pumps and radiators inside each blade.  makes no sense.  ignoring the best bit of novec tech.

Are you an idiot?

Iceotope is 2 phase. Do you even know what 2 phase means?

Its liquid-gas vapour chamber design. Like heatpipe, it IS 2 phase. Dumbass.

[seriouscoin]

2 phase inside the board side and a water block on the other.

The novec is boiled and condensed changing phase on one side of the sealed unit. The others side removes heat with water rad system. If you have a mining board that conforms to a the Iceotope form factor it could be ready now. You would still need the server modular units though check this out: http://www.youtube.com/watch?v=lHHHYcdPFnQ

iceotope isnt 2-phase, theyre using the novec without boiling it (may as well be mineral oil).  theyve got pumps and radiators inside each blade.  makes no sense.  ignoring the best bit of novec tech.

Even if its not 2 phase I am interested in how they can unattach the server without leaking water.

What is the name of magical this part??

Think of ink cartridge. Very similar design, the vapour chamber ,in which the server is completely submerged with the Novec liquid, is sealed with the cold plate (a gigantic waterblock).

The coolant inside the waterblock is cheap and i'm sure the external heatexchange has resorvoir built in so even if some tiny amount of coolant lost (air got into the block), air bubble will be bleed out as soon as the pump runs.

 

[jimmothy]

How would I deal with the vapor pressure?

Basically how can I maintain a safe psi or ratio of gas to liquid?

I am afraid that the gas might not condense quick enough and build up pressure/explode with a closed system.

And with an open system I am afraid too much vapor will escape (shits expensive)

[seriouscoin]

How would I deal with the vapor pressure?

Basically how can I maintain a safe psi or ratio of gas to liquid?

I am afraid that the gas might not condense quick enough and build up pressure/explode with a closed system.

And with an open system I am afraid too much vapor will escape (shits expensive)

I can imagine there is a safety mechanism like a valve with PSI threshold setting or a monitoring controller that shutoff the server if the coldplate failed.

As long as your cold plate can keep the Nove liquid to stay at 50c or whatever the boiling point is.

This is where R&D come in.

I suggest you should look at open immersion cooling instead. Alot safer and cheaper.


Btw, did you find a good source to buy Novec yet? I keep hearing $400/gallon..... thats alot higher than i thought (i thought it was like $200/gallon)

[seriouscoin]

How would I deal with the vapor pressure?

How can I maintain a safe psi or ratio of gas to liquid?

I am afraid that the gas might not condense quick enough and build up pressure/explode with a closed system.

And with an open system I am afraid too much vapor will escape (shits expensive)

What pressure? They are not pressurized.

There are a lot of design issues to do this.

1. Do you have a board that is designed to be housed in the system?
2. Do you have system for the board? You would at this stage have the Engineer do the calculation so you are not building a BOMB.
3. Do you have a cooling system designed?
4. Do you have a racking system designed?

It is pressurized if your cold plate failed.

[jimmothy]

How would I deal with the vapor pressure?

How can I maintain a safe psi or ratio of gas to liquid?

I am afraid that the gas might not condense quick enough and build up pressure/explode with a closed system.

And with an open system I am afraid too much vapor will escape (shits expensive)

What pressure? They are not pressurized.

There are a lot of design issues to do this.

1. Do you have a board that is designed to be housed in the system?
2. Do you have system for the board? You would at this stage have the Engineer do the calculation so you are not building a BOMB.
3. Do you have a cooling system designed?
4. Do you have a racking system designed?

The pressure from evaporating liquid?

1. I am under the impression that any board would work with novec or something similar.
2. What do you mean exactly? I have not build anything I am just conceptualizing.
3. My idea is similar to this: http://www.allied-control.com/publications/Immersion_Cooling_Concept_Xeon_Phi_Poster_WEB.pdf

4. No, but how hard can it be?

[seriouscoin]

How would I deal with the vapor pressure?

How can I maintain a safe psi or ratio of gas to liquid?

I am afraid that the gas might not condense quick enough and build up pressure/explode with a closed system.

And with an open system I am afraid too much vapor will escape (shits expensive)

What pressure? They are not pressurized.

There are a lot of design issues to do this.

1. Do you have a board that is designed to be housed in the system?
2. Do you have system for the board? You would at this stage have the Engineer do the calculation so you are not building a BOMB.
3. Do you have a cooling system designed?
4. Do you have a racking system designed?

The pressure from evaporating liquid?

1. I am under the impression that any board would work with novec or something similar.
2. What do you mean exactly? I have not build anything I am just conceptualizing.
3. My idea is similar to this: http://www.allied-control.com/publications/Immersion_Cooling_Concept_Xeon_Phi_Poster_WEB.pdf

4. No, but how hard can it be?

1. he meant custom pcb size
2. do you have R&D to build a safety measurement system
3. Its a open bath system, as i said earlier... its alot safer, not as elegant and datacentre-friendly but who cares.
4. Agree with you, rack is not part of the system. Icesotope just want their system to be adapted by current datacentre

[seriouscoin]

It is pressurized if your cold plate failed.  

Explain how the cold plate would fail? Doubtful if you have it properly engineered.

Yes and I would agree.

Use the Asicminer design as starting templating.

Closed Immersion Tank (Not pressurized)
Condensing Coil

uh.... because the heatexchanger failed? or the tube is leaked? corrosion?.....you want me to list it?

properly engineering mean they take all safety measurement for a failure prevention system.... not " it cant fail"

[seriouscoin]

It is pressurized if your cold plate failed.  

Explain how the cold plate would fail? Doubtful if you have it properly engineered.

Yes and I would agree.

Use the Asicminer design as starting templating.

Closed Immersion Tank (Not pressurized)
Condensing Coil

uh.... because the heatexchanger failed? or the tube is leaked? corrosion?.....you want me to list it?

properly engineering mean they take all safety measurement for a failure prevention system.... not " it cant fail"

That would not be under pressure.

LOL yeah right.

Dont just repeating the same crap, either elaborate or dont bother replying.

Also,,, the design the OP wanted is called open bath even tho its closed tank. Its not Closed Immersion cooling.
 
http://www.hpcadvisorycouncil.com/pdf/vendor_content/3M_DatacenterDynamicsFocusTuma.pdf

[seriouscoin]

It is pressurized if your cold plate failed.  

Explain how the cold plate would fail? Doubtful if you have it properly engineered.

Yes and I would agree.

Use the Asicminer design as starting templating.

Closed Immersion Tank (Not pressurized)
Condensing Coil

uh.... because the heatexchanger failed? or the tube is leaked? corrosion?.....you want me to list it?

properly engineering mean they take all safety measurement for a failure prevention system.... not " it cant fail"

That would not be under great pressure. There are systems out there now that have off the shelf components. You are going to have issues if you use crap parts. The distances and pressures for a single unit are not that great. Just get yourself a GOOD plumber. I think what you want is something that you can play around with have newer miners that you can drop in and test. Not like this is going to make money cause we all know the numbers.

Pressure is from vapourizing , nothing to do with plumber.... 

[jimmothy]

2. do you have R&D to build a safety measurement system

I have plenty of time to r&d but not much funds.

Would a pressure releif valve and temp monitoring be good enough?

[seriouscoin]

Also OP, you should be careful with certain computer board. The boiling from the Novec can be strong enough to blow small components on the board. I heard someone tested with a cheap Foxcom board and it happened the boiling killed the board.

[seriouscoin]

2. do you have R&D to build a safety measurement system

I have plenty of time to r&d but not much funds.

Would a pressure releif valve and temp monitoring be good enough?

I would say temp monitoring and water pressure of the coolant loop monitoring are very doable.

As your design would be open, you will just lose Novec liquid if it build up pressure.

[jimmothy]

2. do you have R&D to build a safety measurement system

I have plenty of time to r&d but not much funds.

Would a pressure releif valve and temp monitoring be good enough?

I would say temp monitoring and water pressure of the coolant loop monitoring are very doable.

As your design would be open, you will just lose Novec liquid if it build up pressure.

My design would be closed. Basically allied controls design with a lid. Why would you need it open?

[seriouscoin]

2. do you have R&D to build a safety measurement system

I have plenty of time to r&d but not much funds.

Would a pressure releif valve and temp monitoring be good enough?

I would say temp monitoring and water pressure of the coolant loop monitoring are very doable.

As your design would be open, you will just lose Novec liquid if it build up pressure.

My design would be closed. Basically allied controls design with a lid. Why would you need it open?

I meant the tank is not fully closed (sealed). The tank lid ofcourse close for condensing to work but not sealed. 3M called it open bath. Icesotope is a fully closed system.


Btw, i think the greatest challenge would be how to power your boards. Traditional PSU wasnt built for submerged and hence it might not be compatible (you have to check the liquid to confirm). I believe you have to remove the heatsink, clean off any grease and hope nothing else that can pollute your liquid.

[seriouscoin]

2. do you have R&D to build a safety measurement system

I have plenty of time to r&d but not much funds.

Would a pressure releif valve and temp monitoring be good enough?

You want to come up with your concept and PM this member: https://bitcointalk.org/index.php?action=profile;u=124509

Here was a thread I had done on immersion. https://bitcointalk.org/index.php?topic=248184.0

Now i understand where you got "no pressure" from....  

Yours isnt a 2 phase system. And you simply using oil as medium of exchange.

I hope you understand its completely different.... hence waste of time for OP to even follow it.

[seriouscoin]

2. do you have R&D to build a safety measurement system

I have plenty of time to r&d but not much funds.

Would a pressure releif valve and temp monitoring be good enough?

You want to come up with your concept and PM this member: https://bitcointalk.org/index.php?action=profile;u=124509

Here was a thread I had done on immersion. https://bitcointalk.org/index.php?topic=248184.0

Now i understand where you got "no pressure" from....  

Yours isnt a 2 phase system. And you simply using oil as medium of exchange.

I hope you understand its completely different.... hence waste of time for OP to even follow it.

Dude...

I know what the difference is.

Take a breath and read what I have posted on this thread. Stop trolling.

Then explain the "no pressure" claim.... oh wait you deleted that post.

Atleast admit  that you didnt have a clue b4 posting some project thats no help to the topic.

[seriouscoin]

2. do you have R&D to build a safety measurement system

I have plenty of time to r&d but not much funds.

Would a pressure releif valve and temp monitoring be good enough?

You want to come up with your concept and PM this member: https://bitcointalk.org/index.php?action=profile;u=124509

Here was a thread I had done on immersion. https://bitcointalk.org/index.php?topic=248184.0

Now i understand where you got "no pressure" from....  

Yours isnt a 2 phase system. And you simply using oil as medium of exchange.

I hope you understand its completely different.... hence waste of time for OP to even follow it.

Dude...

I know what the difference is.

Take a breath and read what I have posted on this thread. Stop trolling.

----------


Here is what you need to do  jimmothy.

Get your idea done and your parts list then send off a PM to that engineer. I am sure he will steer you right because he has built systems and knows what is possible. He has looked into and designed a variety of different cooling systems.

I'm sure Jim would listen to you..... great portrait of your expertise....

Go boil some water in a sealed pot... please.

[seriouscoin]

Here is what you need to do  jimmothy.

Get your idea done and your parts list then send off a PM to that engineer. I am sure he will steer you right because he has built systems and knows what is possible. He has looked into and designed a variety of different cooling systems specifically for mining and super computing.

2. do you have R&D to build a safety measurement system

I have plenty of time to r&d but not much funds.

Would a pressure releif valve and temp monitoring be good enough?

PM this member: https://bitcointalk.org/index.php?action=profile;u=124509

Here was a thread I had done on immersion. https://bitcointalk.org/index.php?topic=248184.0

If he doesn't reply let me know jimmothy and I will PM you his email.

PM me if you want to continue that tangent.

Holy cow.... Jim i would completely ignore this useless crap. He would waste your time and money to even follow his suggestion. Let me explain:

His project was an oil submerged setup. Only use oil as medium of exchange.

Oil viscosity means high pressure cooling loop as he need to cool the oil.

Thus you will need proper [ump, tubing and fitting setup to move oil from the tank to a heatexchanger/radiator. Its the dumbest idea hence the oil submerged cooling can never be scalable. Thats why no datacentre touches it.


For your setup, you're simply need a water coolant to remove heat from the vapour (condensing). A low pressure high volume can do the job easily. The hard part will be the monitoring system and integrated into your miner controller. If you're not looking to move enormous amount of water, pump from aquarium (pond pump) and normal fittings are fine.

[aerobatic]

2 phase inside the board side and a water block on the other.

The novec is boiled and condensed changing phase on one side of the sealed unit. The others side removes heat with water rad system. If you have a mining board that conforms to a the Iceotope form factor it could be ready now. You would still need the server modular units though check this out: http://www.youtube.com/watch?v=lHHHYcdPFnQ

iceotope isnt 2-phase, theyre using the novec without boiling it (may as well be mineral oil).  theyve got pumps and radiators inside each blade.  makes no sense.  ignoring the best bit of novec tech.

Are you an idiot?

Iceotope is 2 phase. Do you even know what 2 phase means?

Its liquid-gas vapour chamber design. Like heatpipe, it IS 2 phase. Dumbass.

take a look at the design and tell me again if you still think its 2 phase.

there is no phase change going on in this design.  it uses water to cool the novec for the entire tank of novec.  there's nowhere in their design for the novec to boil.  it has nowhere for the vapour to go, and it has no condenser above it. It has no expansion vessel.  In short, this design *ISNT 2-phase*, it keeps the novec in a liquid state, ALL THE TIME, and cools it with direct contact with the cooling plate thats cooled by Water.

http://www.iceotope.com/assets/files/pdfs/module-a4-icm-t1-si2-spec.pdf

in 3m's own PR, it states that Allied Control's solution is 2-phase, and that Iceotope's is convection cooling.

http://prec.pr/symposium/2014/pdfs/feb20-pm/Kevin-Cabrera.pdf

and i expect an apology!   don't go calling people dumbasses and idiots.  especially when you don't know what you're talking about.

[seriouscoin]

2 phase inside the board side and a water block on the other.

The novec is boiled and condensed changing phase on one side of the sealed unit. The others side removes heat with water rad system. If you have a mining board that conforms to a the Iceotope form factor it could be ready now. You would still need the server modular units though check this out: http://www.youtube.com/watch?v=lHHHYcdPFnQ

iceotope isnt 2-phase, theyre using the novec without boiling it (may as well be mineral oil).  theyve got pumps and radiators inside each blade.  makes no sense.  ignoring the best bit of novec tech.

Are you an idiot?

Iceotope is 2 phase. Do you even know what 2 phase means?

Its liquid-gas vapour chamber design. Like heatpipe, it IS 2 phase. Dumbass.

take a look at the design and tell me again if you still think its 2 phase.

there is no phase change going on in this design.  it uses water to cool the novec for the entire tank of novec.  there's nowhere in their design for the novec to boil.  it has nowhere for the vapour to go, and it has no condenser above it. It has no expansion vessel.  In short, this design *ISNT 2-phase*, it keeps the novec in a liquid state, ALL THE TIME, and cools it with direct contact with the cooling plate thats cooled by Water.

http://www.iceotope.com/assets/files/pdfs/module-a4-icm-t1-si2-spec.pdf

in 3m's own PR, it states that Allied Control's solution is 2-phase, and that Iceotope's is convection cooling.

http://prec.pr/symposium/2014/pdfs/feb20-pm/Kevin-Cabrera.pdf

and i expect an apology!   don't go calling people dumbasses and idiots.  especially when you don't know what you're talking about.

The Novec liquid they use boils at 50c. Vaporizing MUST happen to cool the chips and components.

Its essentially a heatpipe (which is also 2phase btw), but your component is completely submerged.

[aerobatic]

2 phase inside the board side and a water block on the other.

The novec is boiled and condensed changing phase on one side of the sealed unit. The others side removes heat with water rad system. If you have a mining board that conforms to a the Iceotope form factor it could be ready now. You would still need the server modular units though check this out: http://www.youtube.com/watch?v=lHHHYcdPFnQ

iceotope isnt 2-phase, theyre using the novec without boiling it (may as well be mineral oil).  theyve got pumps and radiators inside each blade.  makes no sense.  ignoring the best bit of novec tech.

Are you an idiot?

Iceotope is 2 phase. Do you even know what 2 phase means?

Its liquid-gas vapour chamber design. Like heatpipe, it IS 2 phase. Dumbass.

take a look at the design and tell me again if you still think its 2 phase.

there is no phase change going on in this design.  it uses water to cool the novec for the entire tank of novec.  there's nowhere in their design for the novec to boil.  it has nowhere for the vapour to go, and it has no condenser above it. It has no expansion vessel.  In short, this design *ISNT 2-phase*, it keeps the novec in a liquid state, ALL THE TIME, and cools it with direct contact with the cooling plate thats cooled by Water.

http://www.iceotope.com/assets/files/pdfs/module-a4-icm-t1-si2-spec.pdf

in 3m's own PR, it states that Allied Control's solution is 2-phase, and that Iceotope's is convection cooling.

http://prec.pr/symposium/2014/pdfs/feb20-pm/Kevin-Cabrera.pdf

and i expect an apology!   don't go calling people dumbasses and idiots.  especially when you don't know what you're talking about.

The Novec liquid they use boils at 50c. Vaporizing MUST happen to cool the chips and components.

Its essentially a heatpipe (which is also 2phase btw), but your component is completely submerged.

wrong again!

novec is available in different versions at several different boiling points.  there's novec 7000 which boils at 49, and novec 7100 which boils at 61 degrees (there are also other novecs that boil at higher temps too).

you can see clearly that they are using water that flows from top to bottom (or vice versa) through a heat exchanger at the other end, that cools the entire vertical height of the blade's contact plate.   the novec is alongside is is cooled via the contact plate.  the novec has nowhere to boil and nowhere for the vapour to rise (you need a large cavity ABOVE of the liquid for the vapour to collect), and there is also no condenser above the novec, so this is *definitly not* a phase change implementation.  The NOVEC does not change from liquid to gas in the iceotope design.   They are simply using the novec as an inert liquid, and they're keeping it under the boiling point, which is very easy to do and they can achieve this with whichever novec version they choose to use, to ensure it doesn't boil.  presumably its novec 7100 but they could easily use a different one like 7200 since they will never achieve boiling and in this design, they can't cope with it boiling.  as i said, there's nowhere for the gas to go, and no condenser above the novec to turn it back into liquid, so you are dead wrong, as wrong as you could possibly be, that this is a 2-phase design.

Vapourising does NOT happen in the Iceotope design.  They keep the entire board immersed in novec - and it stays in liquid form the entire time - and they cool the novec via the contact plate with the water that runs throughout the entire rack, and regular convection moves the novec around its tank.  This is *NOT* a 2-phase design - the novec never changes from liquid to gas (and back again) - and you are wrong, and extremely rude and misguided in your insults.

[edit] Since writing this, i have heard direct from the horses mouth - 3M - and they confirmed that Iceotope uses Novec 7300, which boils at 98 degrees, and they do not allow the novec to boil in their specific application.  Confirming, there is NO PHASE CHANGE in the iceotope implementation.  It is singe phase.   Have confirmed without a shadow of doubt that you were wrong.  And are actually a nasty piece of work, for putting others down so easily, especially when being ignorant of the facts.

Just in case you don't believe 3M... (or me, for that matter)...   here's Iceotope explaining it themselves in their own blog, stating that they use convection (not phase change) for the cooling.. and confirming that they use Novec 7300 (boiling point 98 degrees), and thus do not ever boil the novec.  It NEVER changes into a gas in the iceotope design.

http://blog.iceotope.com/2012/03/its-cooler-to-soak-your-servers.html

[Bicknellski]

+1 thanks for that detail notes and research.

Now that means you would have a sealed section that is not under substantial pressure correct or designed to take pressure and not explode even if the water is shut off and can't flow on the otherside? How to do they fill the space with Novec and I am guessing there will not be 100% fluid in the space depending on how the fill it the have to account for expansion of the liquid at temperature.

I'd like to see how the boards are loaded inside and how they keep the unit sealed.

http://blog.iceotope.com/2012/03/its-cooler-to-soak-your-servers.html

Each module is a server is encased in a leak-proof box filled with 3M's Novec 7300. On heating, this inert liquid expands 10 times more than water and so generates convection currents that carry the heat away, aided by a chimney effect created by the internal shape of the box. It convects very fast – up to 6 cm/sec according to Hopton -- and is less viscous with a lower surface tension than water, which enables it to reach the parts that water might not.

And this is obviously way off what Jimmothy is looking to do as he said he is looking at 2-phase closed tank system as a possible DIY for his miners. Maybe we need to get back on that track.

My design would be closed. Basically allied controls design with a lid. Why would you need it open?

[jimmothy]

+1 thanks for that detail notes and research.

Now that means you would have sealed section that is not substantial pressure correct or designed to take pressure and not explode even if the water is shut off and can't flow?

I'd like to see how the boards are loaded inside and how they keep the unit sealed.

Each module is a server is encased in a leak-proof box filled with 3M's Novec 7300. On heating, this inert liquid expands 10 times more than water and so generates convection currents that carry the heat away, aided by a chimney effect created by the internal shape of the box. It convects very fast – up to 6 cm/sec according to Hopton -- and is less viscous with a lower surface tension than water, which enables it to reach the parts that water might not.

This is what I asked AC about. I am wondering if a pressure releif valve and temp monitoring would be good enough.

I think the novec 7300 single phase solution is dangerous because it could easily explode if water is not flowing/cooling fast enough to prevent boiling.

Novec 7500 might be a better choice for single phase in a sealed container at least.

[Bicknellski]

They have to evacuate the air somehow no matter what unless what you build is a pressure vessel I guess or as you say a relief valve that allows gas out but not liquid. I'd like to send them a few boards so they can seal them up and mine on them.

But you don't need that in your tank, condenser version at home. There is plenty of space and everything is under normal atmospheric pressure in that system.

[jimmothy]

They have to evacuate the air somehow no matter what unless what you build is a pressure vessel I guess or as you say a relief valve that allows gas out but not liquid.
But you don't need that in your tank, condenser version at home. There is plenty of space and everything is under normal atmospheric pressure in that system.

I think there would still be a chance of explosion of there is no condensing and all the liquid is boiled.

[Bicknellski]

They have to evacuate the air somehow no matter what unless what you build is a pressure vessel I guess or as you say a relief valve that allows gas out but not liquid.
But you don't need that in your tank, condenser version at home. There is plenty of space and everything is under normal atmospheric pressure in that system.

I think there would still be a chance of explosion of there is no condensing and all the liquid is boiled.

Really?

Doubt that. The design would be rated so that wouldn't be possible meaning they would allow for a complete failure of the cooling and the unit will simply get hot and never boil the fluid. Besides they would also set the pressure to something like sea-level or more so that boiling would not be possible. I think they have already engineered it very well as the units are fully functional and this is 8-9 year design build in the making design right? You wouldn't never do this as DIY at home.

I don't see any valves looks to be a simple sealed unit.



Iceotope Unit $6300 Euros.

[jimmothy]

They have to evacuate the air somehow no matter what unless what you build is a pressure vessel I guess or as you say a relief valve that allows gas out but not liquid.
But you don't need that in your tank, condenser version at home. There is plenty of space and everything is under normal atmospheric pressure in that system.

I think there would still be a chance of explosion of there is no condensing and all the liquid is boiled.

Really?

Doubt that. The design would be rated so that wouldn't be possible meaning they would allow for a complete failure of the cooling and the unit will simply get hot and never boil the fluid. Besides they would also set the pressure to something like sea-level or more so that boiling would not be possible. I think they have already engineered it very well as the units are fully functional and this is 8-9 year design build in the making design right? You wouldn't never do this as DIY at home.

I don't see any valves looks to be a simple sealed unit.



Iceotope Unit $6300 Euros.

How can the fluid never boil? The boiling temp for that 1 phase system is only 100 degrees. Are you telling me that not even a small amount can boil?

And my design would be 2 phase so without condensation wouldn't it explode?

[Bicknellski]

Like I said in PM get an engineer to explain it after you have given them your thoughts.

In terms of safety if you are unsure don't even begin a project like this.

Given that the Iceotopes are set in a multimillion dollar server facility and functioning without any reservations from the engineering team that gives me confidence to say they have resolved the issues you are talking about or they are not an issue as they have designed that issue out of the system.

In terms of what you want to do you should try and get a hold of the details on the Allied Control system. Obviously it is operating at standard pressures those boxes are not sealed with extreme pressures in mind and ratcheted down and by the looks of them you can get the lids off fairly easily with a socket wrench just a little more than finger tight I bet. The tanks are there to seal out the dust etc and keep the vapor in not to put the system under pressure. The tanks have enough volume to allow for the gas to condense and not build up pressure and explode. They talk about the tanks as only 30% capacity. Again get an engineer to do the calculations. Also look at the other immersion thread someone already built a DIY unit and I believe this was also discussed. https://bitcointalk.org/index.php?topic=255613.0



http://www.allied-control.com/blog/immersion-2-tech-details-cooling-hundreds-of-kilowatt-with-1500-watt

Only 25% of Condenser Capacity in Use

The condensers are very efficient. The current hardware is only scratching the surface of single condenser capacity at water temperatures of around 35°C. Total capacity of a condenser is ~25kW at that temperature and we can install three more per tank. Our systems can also modulate water flow rates or increase water temperatures (hot water cooling) to get more out of them. We could easily move up to different fluids with higher boiling points to save more energy. Or we could automatically switch on or off our dry cooler modules. All this will have a large impact on performance and efficiency (we are not just talking a few percent here). And all these changes, except adding more condensers, don’t even require a screw driver. And having modular condensers also means we could always upgrade for something with more capacity.

Only 30% of Space Used in Tanks

Only the lower 30% of the physical space in the tanks is full with hardware and fluid at this moment. The condensers are modular and there is a rail on top. Filling our enclosures with more hardware or arrange it differently (ie. two layers), is very easy if the client decides to move to next gen hardware, and so is packing things a little denser (given that they can get rid of the LAN wiring and integrate communications to a backplane).

http://www.slideshare.net/petehopton/iceotope-presentation-dec2011

An older exploding view of the Iceotope. Obviously some tweaks in the design.



Notice the TOP loading board in the newer version?

Why did they do that?

Easier to load and maintain the unit?

Pressure issues?

Who knows why.

[jimmothy]

Turns out pressure is an issue and AC/3M prefer to keep their immersion tubs at atmospheric pressures so a vent (breathing system) is required.

Heres a nice video with tons of details about immersion cooling:

https://www.youtube.com/watch?v=ivVoANqFBuY

[Bicknellski]

Cool I think I saw that video before... just note that I am in contact with an Engineer at Allied Control... they are working on an open bath system as well Jimmothy.

Safety relief valve / purge valves I guess?



https://www.swagelok.com/downloads/WebCatalogs/EN/MS-01-62.pdf

[antirack]

What 3M and Allied Control promote is 2-phase open bath immersion cooling. Open means there is no pressure build up. Pressure is bad. Things can go wrong. Leaks are inevitable.

Iceotope is single phase and completely different. They could be using mineral oil instead, but don't because for mineral oil they would need to pump it around. They might have some pressure build up, given that they say the Novec expands 6 times more than water. Notice that Iceotope (and oil for that matter) still need heatsinks and stuff. And these heatinks actually can't be using thermal grease, so they have to be removed, cleaned and put there with indium foil or similar things, it would otherwise contaminate the oil or Novec.

http://tv.pionier.net.pl/Default.aspx?id=2786&type=Vod

Might take a while to load, but shows a lot of details.

[Bicknellski]

Ya definitely.

The fact you have to use Indium shims or carbon based heatsink paste / pads is an issue in terms of prep.

I haven't seen that link nice one you get to see the board pulled out. The fluid sure isn't filled to the top is it.