Overclocking Crazy submersion phase-change system

High guys....just thought I'd link you to my latest project.
It's an evolution of my previous chill box build...but this time....well where do I start.

The chamber is basically like a huge actively cooled vapor chamber with the motherboard submerged in the bottom of the chamber in the working fluid which is liquified gas.

...perked your interest?

Here's a youtube link to the web cam footage from inside the chamber.

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

So the chamber is completely sealed and made of 1.5cm alu plate as it has to withstand negative pressures of -10psi.

The evaporator of my 3.6kw air con unit is sealed inside the chamber and the chamber is filled with the gaseous refrigerant.

When you turn the ac unit on the temp drops, liquifies the vapor causing a partial vacuum which is used to draw liquified refrigerant from a pressure bottle into the chamber sump where the pc mobo becomes submerged.

So the mobo is submerged in liquified refrigerant....the refrigerants boiling point at normal temp and pressure is -1c but because the cold in the chamber is causing a pressure drop the liquid boiling point also drops.....in fact the liquid is in a state of equilibrium balanced right on the edge of a variable boiling point the value of which depends on the pressure/temperature of the chamber.......you can see this in the video...the liquid is in a constant bubbling state even with no load...like a witches cauldron. The lowest temp the chamber produces is about -30c.

The components are cooled by the liquid phase changing...this is where it's good to have the liquid on the edge of liquid/vapor equilibrium as it's always ready to boil off given a heat input.

I have to contain the liquid in a pressure cylinder between uses as the chamber was not designed to withstand expansive pressure...just compressive pressure.

So I basically reverse the filling process.....the cylinder is place in a chest freezer so causing a low pressure inside....the chamber cooling is turned off and as it warms the liquid is transferred across through a linking pipe because of the pressure differential.

You like.....I think this is a first.....

 

I'd be inclined to agree that it is a first. Well done.

What sort of temps are the system monitors giving when pushing things?

 

Not bad temps but not as good as I'd hoped.



Uploaded with ImageShack.us



Uploaded with ImageShack.us



Uploaded with ImageShack.us

I was hoping to be completely sub zero......but that obviously hasn't happend.....I used a home brew TIM to try and prevent it being washed out by the submersion liquid....think next time I will use Indigo extreme...always room for improvement.

 

Awesome!

I know next to nothing about overclocking with more exotic means. Do you encounter the so called 'cold bug' and also can you turn the system on straight away or do you have to keep restarting the system as temps come down?

 

With my previous chill box I couldn't boot beyond 5.5Ghz....Haven't tried with this new system but think it will be the same.

No you don't have to do that....but you do have to get the chamber temps down before you let the liquid refrigerant in otherwise it would boil off quickly causing high pressure that may damage the chamber.....so it's not an instant on system....it's still early days and I'm experimenting with what I can do with it.

 

I like! Have some rep for even attempting it

 

This is hardcore!

 

I wonder if this is capable of overcoming Ivy's heat transfer problems, because you're cooling the chip through the bottom too, and the traces are good heat conductors.

 

Nice can we have some pics of the computer or are you keeping them on lock down?

 

It's bound to help...heat removal is heat removal....the liquid is much less viscous than water and must get into the socket and around the pins...although the small spaces may gas up and effectively exclude the liquid.....but of course its also on the other side of the mobo making the traces cold.

There is a limitation with this kind of "passive phase change" cooling though..at heat densities greater than 6 to 10watts/cm^2 the bubbles start to insulate the component.....that's why I have a copper heat sink not just a bare IHS.....so a de-lidded IvyB would have too large a heat density for direct die/liquid cooling and you would have to utilize a heat spreader of some description.

 

It's not a secret....it's just an ugly monstrosity....its a mess of pipes cables and insulation board....not a look to be proud of...but it was quite a feat to work out how to do this and build something capable of doing it, no hi-tech plasma cutters or big business investment here, just an idea and a man working in his shed.....so looks went out the window.

 

Who cares about the looks I want to see the build. It's not always about looks but function and making your self happy. I'm just curious to see how you did it.

 

okay don't say I didn't warn you just uploading a little youtube clip and will give a more detailed explanation.

 

so there you go...huumm....

http://www.youtube.com/watch?v=ug8D0B0Q13g&feature=youtu.be

It's hard to see whats happening there, but at the back is a stripped down 3.6kw window air con unit......the evaporator is built into an a heavy duty alu chamber 1.5cm thick plate used to make it as the negative pressures are considerable, that is for obvious reasons encased in insulation board.

In the bottom of the chamber I've got a pressed steel basin so there is absolutely no chance of catastrophic leak and 20 liters of liquid refrigerant dumping out ( although the alu chamber is of course sealed as well, just thought a seamless pressed steel basin could not fail)

The evaporator is place above the basin and shrouded with three scythe gentle typhoon fans built into the shrouding to pull vapor through the evap fins.
The mobo sits in the basin and the cables are routed/sealed through the side wall where they enter secondary chamber (that white pipe)...the reason for this is because the chamber operates under negative pressure ...approx -10psi.....air can be drawn through the interior of the cables....so I have cable extenders joined in that white pipe and the joins are submerged in oil....so oil is pulled into the cables instead of air which solidifies as it enters the cables in the cold chamber effectively sealing them.

At the bottom of the basin is a flexible stainless steel hose which leads out of the chamber to the pressure cyclinder in the freezer.....I went for expensive flexi stainless steel hose as it carries the liquid refrigerant to and from the chamber.....again don't want that fracturing.

The bag and other gas bottle are just for purging the systems vapor space.

I don't know if that is enough detail for you to see whats happening but I can go through the process if you like.

 

Would be interesting to see what temperature your heat sink is reaching (May help work out why your temps are bad -I.e "hot" sink suggests poor transfer of heat to your coolant)

Have others had good success with this method of cooling? The reason i ask is that rapid boiling of the liquid would seem a less efficient method of heat transfer than say the board being submerged in a liquid (I.e not a liquid boiling off) at -36.

 

Well this is a different way of doing phase change cooling

My only experience is my single stage at home which I used for benching, and I went to a workshop a couple of years ago where we built a couple but I don't remember anything

 

Yes a probe in the heatsink would help with trouble shooting.....I don't think anyone has done it like this before....the nearest comparison is 3m Novec 7000 builds which use submersion passive phase change...but they are at ambient temps.

They use a special coating that enhances boil off of the liquid...apparently increasing efficiency by 15x. But I wasn't sure I could copy that coating effectively and just hoped the larger surface area of the heatsink I'm using would compensate (they use just the coated IHS)...but it may be that.

Boiling off should be much more effective than just submerging in liquid....when the liquid boils off it absorbs 220 joules of heat per cm^3 of liquid boiled...in comparison to water just absorbing heat which only absorbs 4.2 joules per cm^3 per deg c

It may be the TIM I used....I've used a home made one as all silicone based commercial TIM's would dissolve and wash away in the fluid....It may be the TIM is not as good as I thought or I may have just put too much on there...silly mistake.

I've noted that my gpus which have solid heat sinks which appear glued on with thermal epoxy has better temp drops when compared to my previous chill box build..... going from +50c in furmark to around +10c....they have no special coating so it may be just the TIM.

 

Yep always thinking outside the box......or perhaps inside the box would be a better description.

I just got this bee in my bonnet thought this cooling method out and had to see it through.


It does have many advantages over "normal" phase cooling....no condensation concerns...no need for a dedicated cooling head for cpu and multiple gpu's also the ram,chipsets and vrm's all benefit from the cooling.

Yep my build is a monstrosity.....I'm using it more as a proof of concept and test bed for several ideas I've got that could lead this to become significantly smaller and neater.....maybe even to the point where it could become a commercial product....although I'm not really planning on doing that...but we will see where it leads.

 

Thanks for the Video I appreciate it. It doesnt matter that it's a mess what matters is the effort/engineering/thought process. Trying to make things run better/longer and at higher frequencies is what it's all about. Keep up posted if you make a new/smaller chamber so we can see the build process. I'm going to assume you have a background in HVAC?

 

Thanks Mech....I have no background in HVAC...just thinking all this out as I go.....now I've got proof of concept in that the temperature driven pressure drop in the chamber causes the liquid to be on the point of liquid<>vapor equilibrium held right on it's boiling point I see no reason that I couldn't use a low pressure refrigerant....one with a boiling point in the vicinity of ambient temperatures (at standard pressure)....again if that was in a hermetically sealed chamber in equilibrium with a 100% vapor phase then its actual boiling point would vary and be dependent on the chamber pressure so would go as low as you could drive the pressure.

With a low pressure refrigerant there would be no need to recover that to a pressure bottle or need for a freezer to drive that recovery.....so that is half the build cut out to start with, and it is something I could test by adapting my current build.

Another improvement would be removing the air con unit evaporator, having that in the chamber means the chamber has to be large......but there is really no need for it....it is just using temperature to drive the pressure drop...but the compressor suction line can do that directly......so basically just have the compressors suction line opening into the chambers vapor space and the capillary tubing returning fluid to the sump.

Such a system would have to have an oil separator in there to allow for lubricant return to the compressor......and it wouldn't be hard to trap any small amount of lubricant that escaped the separator and made it to the chamber.

The hardest part and most hit and miss is being able to pass all the cables into the hermetically sealed chamber.....but if I ever did go commercial with this I'd have to get a custom bulkhead made.

So I will have to get some HVAC skills as I'd like to evolve the project to include those ideas.....I'd even like to design it like a split ac system so the pc could be indoors and the noisy old compressor outside...lol.