Overclock.net › Forums › Cooling › Cooling Experiments › Subambient full submersion phase change cooled pc
New Posts  All Forums:Forum Nav:

Subambient full submersion phase change cooled pc - Page 4

post #31 of 333
Thread Starter 
Hi again all and sorry for the silence on this....I considered it time to shut up and get on with things.....so I've been busy doing the list of domestic chores my wife had written that must be done before this project, I've finished those and I'm just putting the final touches to the shed /outhouse I've been building to house this project.

So in all I'm finally ready to start building this project.

After much indecision and basically sound boarding various approaches you will be I'm sure more than pleased that the final design is now decided.

The chamber is going to be very much like a large vapor chamber, the mobo will be fully submerged in the working fluid of liquified refrigerant gas contained in a pressed steel basin at the base of the chamber. The air con evaporator will sit above the steel basin and fans will circulate the vapor phase which will re-condense on the ac unit evap and drip back into the working fluid basin.

The chamber itself will be completely sealed and so as the vapors condense to liquid due to the cooling effect of the ac unit then the pressure in the chamber will decrease.

The lower pressure within the chamber will lower the boiling point of the refrigerant gas.

So in effect the liquid and vapor phases will be in a state of thermal/pressure equilibrium. This has the advantage that regardless of the refrigerants boiling point under standard temperatures and pressures, in this system the refrigerant will constantly be right at its boiling point so that any heat input will cause phase change to gas.

As a consequence of the low pressure operation which will be at it's lowest -10psi compared to atmospheric then the chamber and seals have to be of robust construction able to withstand considerable compressive forces.

The system will not however be required to contain the liquified gas between uses. The pressures involved to do this would be too high to safely work with. So it is my intention to drain the fluid into a pressure bottle between uses, placing the pressure bottle in a freezer will cause low pressures within it which will suck the working fluid out of the chamber and into the bottle whilst still in the liquid phase where it can be contained safely.

So that's the outline of the final design and I'm ready to start building.

I intend to make the chamber out of heavy gauge aluminium either 1cm or 1.5cm thick, the plates of the chamber will interlock in such a manner as that they will resist compressive forces.

I do however need to include some way of viewing what is happening inside the chamber so that I can monitor the working fluid levels and fan operation.

I was considering mounting a usb camera inside the chamber but not sure if it would operate in the low pressure low temp environment.

Alternatively I was considering cutting out a section of the chamber wall and overlaying it with thick plate glass which could withstand the pressure.
***************
Not sure which would be the best means of doing this and would appreciate opinions.
***************

Another area of indecision I'm having is regards component heat sinks.....please bare with me on this as it is a little involved.

To start with as this system uses phase change cooling the heat sink has to have a surface which encourages the conversion of liquid to gas correctly called nucleation.

From information I've gleaned from Novec 7000 builds which also involve submersion phase change cooling then a surface coated with sub 10 micron particles forms the best nucleation surface.

To achieve this I'm considering coating the heatsink surface with Gallium (melting point 29c) and embedding diamond dust in this grit size 2k - 3k.

************
again any info or opinion on this is welcomed
************

An addition problem with submersion phase change is the heat density of the components. Again referring to info from Novec builds the maximum heat density that can be coped with is 6 - 10 Watts/cm^2.

Beyond this point the layer of gas generated at the component surface starts to insulate the component.

So projecting this to the real world the area of my 2700K heat spreader is 3.7cmX3.7cm= 13.7cm^2

So this gives maximum power coolable at the quoted max heat densities of 82watts to 137watts which is too low for an overclocked 2700k.

So I have to increase the area of the heatsink to enable a greater cooling capacity.

Here in lies a problem as I also want to keep the cooling as direct as possible and keep the additional layers to a minimum.

I was originally considering using a layer of panasonic pyrolitic graphite to spread the heat laterally but have some concerns about this.

Although this graphite sheet has a high lateral thermal conductivity, 1700 w/k/m as compared to 400 for copper, the sheet is very thin (25 microns). So considering the total amount of heat that could be transferred laterally through the very thin cross section of this sheet it would in effect be no more than a 0.5mm sheet of copper placed directly on top of the IHS.

I doubt such a thin layer of graphite or copper would actually be able to transfer a sufficient amount of heat energy in a lateral direction to cope with the heat generated.

So I've moved away from the graphite idea.

I'm now thinking of using an Akasa all copper 1U server air cooler heatsink.
I would cut off all the fins as they would only serve to trap bubbles and so insulate the component. So it would leave me with a flat plate of approx 8cmX8cm so even using the lowest heat density figures that would be sufficient to cool 384watts.

Additionally to try to remove as many layers as possible I'm considering removing the center of the heatsink so that it only makes contact with the IHS around it's outer edge...perhaps for a width of 1cm.
So that the IHS will be in direct contact with the working fluid directly over the cores and yet the heat sink will still be able to remove additional heat through the edges to the larger surface area of the heatsink to allow for greater heat dissipation required by the oc'ed component.

I'd also be considering using gallium or gallium alloy as a tim that would be solid under usage conditions as I'm not sure if a normal tim would be washed away by the working fluid.

****************
Again would value views and opinions on this.
****************
post #32 of 333
Thread Starter 
Okay, well I bought myself a cheap webcam to test it out, and the one I have works fine in my current chamber at -24c, so that's a good start, I'd rather include a webcam in the chamber than have to cut a great big viewing portal in the side.
I've no mean of testing this at low pressure in a saturated hydrocarbon vapor at this time but will just have to trust to luck on that.
post #33 of 333
Awesome I am keeping an eye on this as it is intriguing me thumb.gif
kamikaze
(15 items)
 
  
CPUMotherboardGraphicsRAM
3770k @ 4.8 asus maximus v extreme Asus GTX 770 DCUii SLI GSkill 2400Mhz 
Hard DriveOptical DriveCoolingOS
Sandisk ultra pro liteon bluray xspc 240 raystorm Windows 8.1 Pro 
MonitorKeyboardPowerCase
x3 aoc 23' led nvidia suround G19 enermax maxrevo 1500w nzxt 820 Phantom 
MouseAudio
saitik rat 9 Asus Xonar DX 
  hide details  
Reply
kamikaze
(15 items)
 
  
CPUMotherboardGraphicsRAM
3770k @ 4.8 asus maximus v extreme Asus GTX 770 DCUii SLI GSkill 2400Mhz 
Hard DriveOptical DriveCoolingOS
Sandisk ultra pro liteon bluray xspc 240 raystorm Windows 8.1 Pro 
MonitorKeyboardPowerCase
x3 aoc 23' led nvidia suround G19 enermax maxrevo 1500w nzxt 820 Phantom 
MouseAudio
saitik rat 9 Asus Xonar DX 
  hide details  
Reply
post #34 of 333
Thread Starter 
Thanks Feznz, I think this truly does qualify as a "cooling experiment"...I've certainly never seen any reference to anyone doing anything similar to this, novec 7000 submersion is about the closest but of course that is ambiently cooled and uses liquid not liquified gas....being a first makes things harder as its all going to be trial and error...probably more of the latter....cheers and keep following...if people are interested I'll do a full build log if not I'll just post up results later.
post #35 of 333
Subbed...just curious, haha
 
Play Server
(7 items)
 
NAS
(10 items)
 
CPUMotherboardGraphicsRAM
Intel i7 3930k @4.8-5.2 Asus Rampage IV Black Edition 3x R9-290 Corsair Dominator Platinum 16GB @1866C9 
Hard DriveHard DriveHard DriveCooling
Intel 750 400GB 2x Samsung 840 EVO 500GB (1TB) SanDisk Ultra II 960GB EK Rampage IV BE Water Block 
CoolingCoolingCoolingCooling
2x Swiftech D5 w/ EK Dual Pump Top XSPC EX480, EX360 & 2x Phobya 1080 EK Supremacy EVO CPU Block EK Advanced 250mL Res 
CoolingCoolingCoolingOS
Copper Tubing w/ push fittings 3x EK-FC R9-290X Lamptron FC-5V2 Windows 10 Pro 
MonitorKeyboardPowerCase
Samsung U28D590D 4K Leopold Otaku Corsair AX1200i Silverstone TJ07, heavily modified 
MouseMouse PadAudioAudio
Corsair M65 Corsair MM600 Sennheiser HD595 / HD428 / HD280 Corsair SP2500 
CPUCPUMotherboardRAM
Intel Xeon E5620 Intel Xeon E5620 Dell R710 Mobo Hynix 144GB DDR3 ECC PC3L-10600R 
Hard DriveCaseOther
Intel X25-M G2 80GB Dell 710 2.5" - Google Search Appliance Dell Perc H700 Raid Controller 
CPUCPUMotherboardRAM
Intel Xeon E5-2670 v1 Intel Xeon E5-2670 v1 SuperMicro X9DRL-iF 64GB Kingston ECC (kvr16r11d4/16HA) 
Hard DriveCoolingOSPower
12x 4TB HGST 5K4000 Scythe AP-29 FreeNAS Corsair AX750 
CaseOther
Norco RPC-4224 3x LSI 9211-8i (HBA) 
  hide details  
Reply
 
Play Server
(7 items)
 
NAS
(10 items)
 
CPUMotherboardGraphicsRAM
Intel i7 3930k @4.8-5.2 Asus Rampage IV Black Edition 3x R9-290 Corsair Dominator Platinum 16GB @1866C9 
Hard DriveHard DriveHard DriveCooling
Intel 750 400GB 2x Samsung 840 EVO 500GB (1TB) SanDisk Ultra II 960GB EK Rampage IV BE Water Block 
CoolingCoolingCoolingCooling
2x Swiftech D5 w/ EK Dual Pump Top XSPC EX480, EX360 & 2x Phobya 1080 EK Supremacy EVO CPU Block EK Advanced 250mL Res 
CoolingCoolingCoolingOS
Copper Tubing w/ push fittings 3x EK-FC R9-290X Lamptron FC-5V2 Windows 10 Pro 
MonitorKeyboardPowerCase
Samsung U28D590D 4K Leopold Otaku Corsair AX1200i Silverstone TJ07, heavily modified 
MouseMouse PadAudioAudio
Corsair M65 Corsair MM600 Sennheiser HD595 / HD428 / HD280 Corsair SP2500 
CPUCPUMotherboardRAM
Intel Xeon E5620 Intel Xeon E5620 Dell R710 Mobo Hynix 144GB DDR3 ECC PC3L-10600R 
Hard DriveCaseOther
Intel X25-M G2 80GB Dell 710 2.5" - Google Search Appliance Dell Perc H700 Raid Controller 
CPUCPUMotherboardRAM
Intel Xeon E5-2670 v1 Intel Xeon E5-2670 v1 SuperMicro X9DRL-iF 64GB Kingston ECC (kvr16r11d4/16HA) 
Hard DriveCoolingOSPower
12x 4TB HGST 5K4000 Scythe AP-29 FreeNAS Corsair AX750 
CaseOther
Norco RPC-4224 3x LSI 9211-8i (HBA) 
  hide details  
Reply
post #36 of 333
Thread Starter 
And we all know what curiosity did to the cat.........unless of course you're Schrodinger's cat...lol
post #37 of 333
Thread Starter 
Okay guys bare with me on this one, obviously this is going to be a fairly time consuming build as I have to make just about everything myself.
But I have made a decisive start and there is no going back now.

I've stripped the air con unit out of my previous project, man I made that chill box strong so it had to be a fairly destructive tear down and sadly irreversible step. frown.gif

Still as I now don't own a usable gaming pc it must be full steam ahead.

I've wrestled the ac unit downstairs and into the outhouse/shed I've built to house this project and you can see it below with the cold evaporator rad balanced on top of the pressed steel basin, I've done that so I can measure up the height of chamber I'll need to construct, thats how it's going to be set up with the evap just above the basin both of which are sealed in the chamber.

The chamber is going to be made of 1.5cm alu plate...so now I've got the dimensions it's time to order. wheee.gif



Uploaded with ImageShack.us



Uploaded with ImageShack.us

Like I said early days...
Edited by technogiant - 7/11/13 at 1:50pm
post #38 of 333
Thread Starter 
Just ordered my aluminium plate for the chamber.......£380....ouch that hurt....this darn project had better work out now or I'll be spitting feathers.
post #39 of 333
Quote:
Originally Posted by technogiant View Post

Just ordered my aluminium plate for the chamber.......£380....ouch that hurt....this darn project had better work out now or I'll be spitting feathers.

I hope it does too wink.gif work not hurt I mean
I am a little miffed at how the vapour chamber will be at -10psi, I wait eagerly for some updates
kamikaze
(15 items)
 
  
CPUMotherboardGraphicsRAM
3770k @ 4.8 asus maximus v extreme Asus GTX 770 DCUii SLI GSkill 2400Mhz 
Hard DriveOptical DriveCoolingOS
Sandisk ultra pro liteon bluray xspc 240 raystorm Windows 8.1 Pro 
MonitorKeyboardPowerCase
x3 aoc 23' led nvidia suround G19 enermax maxrevo 1500w nzxt 820 Phantom 
MouseAudio
saitik rat 9 Asus Xonar DX 
  hide details  
Reply
kamikaze
(15 items)
 
  
CPUMotherboardGraphicsRAM
3770k @ 4.8 asus maximus v extreme Asus GTX 770 DCUii SLI GSkill 2400Mhz 
Hard DriveOptical DriveCoolingOS
Sandisk ultra pro liteon bluray xspc 240 raystorm Windows 8.1 Pro 
MonitorKeyboardPowerCase
x3 aoc 23' led nvidia suround G19 enermax maxrevo 1500w nzxt 820 Phantom 
MouseAudio
saitik rat 9 Asus Xonar DX 
  hide details  
Reply
post #40 of 333
Thread Starter 
Quote:
Originally Posted by feznz View Post

I am a little miffed at how the vapour chamber will be at -10psi, I wait eagerly for some updates

The chamber will have two phases of refrigerant, the liquid phase contained in the steel basin and the vapor phase above which will be 100% refrigerant vapor.

At standard temperatures and pressures the refrigerant has a boiling point of -1c.

So as the air conditioner evap drops the temperature, it can go down to -30c, then the vapor will condense on it, that conversion of vapor to liquid causes a pressure drop as the chamber is sealed (1ml of fluid has a gaseous volume of 240ml). As you lower the pressure then the boiling point of the liquid phase also lowers and would so tend to boil off more readily and so increase pressure.......the whole thing basically is in liquid<>vapor and temperature<>pressure equilibrium.

So the lower the units temperature then the lower the internal pressure and the lower the boiling point of the liquid phase.

You can actually can work out what the chamber pressure will be from a vapor pressure / temp chart as this plots the pressure (boiling off pressure) exerted by a liquid at a given temp.......the vapor pressure exerted by this refrigerant at -30c is about 5psi. Normal atmospheric pressure is about 15psi so the unit will be at -10psi when compared to atmospheric.

That's why I've gone with a substantial chamber build using 15mm alu plate, the chamber has to be large enough to contain the basin and evap so the panels are quite large.....if you work out the force exerted by the atmosphere on those panels at 10psi then there is considerable load present.
Edited by technogiant - 7/12/13 at 12:34pm
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Cooling Experiments
Overclock.net › Forums › Cooling › Cooling Experiments › Subambient full submersion phase change cooled pc