EXTREME Overclocking Forums
Home | Reviews | Forums | Downloads | RealTime Pricing Engine | Folding Stats Contact Us


Go Back   EXTREME Overclocking Forums > General CPU Cooling > Phase Change, Dry Ice, LN2, Other Crazy Stuff

Welcome Guest Visitor! Please Register, It's Free and Fun To Participate!
The EXTREME Overclocking Forums are a place for people to learn how to overclock and tweak their PC's components like the CPU, memory (RAM), or video card in order to gain the maximum performance out of their system. There are lots of discussions about new processors, graphics cards, cooling products, power supplies, cases, and so much more!

You are currently viewing our boards as a "guest" which gives you limited access to view most discussions. You need to register before you can post: click the register link to proceed. Before you register, please read the forum rules. By joining our free community you will have access to post topics, communicate privately with other members (PM), respond to polls, upload your own pictures, and access many other special features. Registration is fast, simple, and absolutely free! To start viewing messages, select the forum that you want to visit from the selection below.

After you have registered and read the forum rules, you can check out the FAQ for more information on using the forum. We hope you enjoy your stay here!

Note To Spammers: We do not allow unsolicited advertising! Spam is usually reported & deleted within minutes of it being posted, so don't waste your time (or ours)!


Please Register to Post a Reply
 
Thread Tools
Old 01-20-2013, 12:09 AM   #21
ORL
Bring on the Liquid
ORL's Avatar
Senior Member
 
Posts: 2,141
Last Seen: 07-11-2017
From: Missouri
iTrader: 0 / 0%
Well, even still I am a bit concerned then for your total designs temperatures. Where as yes you will see a slightly lower temp than what the AC unit flow will produce for your submersed temp, without an active compression recovery of the bath I feel that the temps produced for the rig will just be lost in effort. You might as well operate under the standard phase change systems.

However an active recovery/compression/propellant type system should yield substantially better temperatures due to the vacuum effect on the actual cooling chamber, as I am sure you know, the vacuum will substantially increase the rate of evaporation thus allowing the liquid to cool faster and thermally resulting in lower temps. And if your ballsy enough and have the money to do so doing multi staged with a final phase of LH2 as the bath youcould be king of the low temps.

Although on a serious note, I would love even still to see your results and will follow closely. I will help with any thing I can.


If done properly the multi stage system I mentioned could potentially hit one of the early stages of the early transition points for superconductivity for the elements used in these components. Where as this would be insanely awesome if possible I would also start becoming worried about the different expansion/contraction rates of the components causing damage.


Completely different side note! Its also been proven that Absolute Zero is in fact NOT the coldest achievable temperature. Using charged elemental compounds after reaching near absolute zero causes them to disperse when coming in contact with an opposing magnetic field simulating evaporation and cause an additional temperature drop. Just a little educational tidbit!
United States  Offline
    Register to Reply to This Post
Old 01-20-2013, 01:00 AM   #22
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
Lol....while I love your idea of the prospect of sub absolute zero temps I don't think my diy and technical skills would extend to such extremes....lol

Yes I agree using a pressurized system would I'm sure ultimately deliver better temps, but once you get into the area of using pressure and vacuum then the engineering requirements and costs increase exponentially.

I'm pursuing a lower tech lower budget approach but one that still enables the use of direct phase change cooling of the components as this is the most effective form of cooling, on average each cubic centimeter of refrigerant that evaporates removes 200 joules of energy, by comparison to water cooling 1cc of water would have to raise in temperature by 48 Deg C to absorb the same amount of energy.

The submersion aspect of the project is just to provide a simple and effective means of delivering the phase change cooling to the components.

It's essentially just the same method as the 3M Novec 7000 submersion cooling method you can see here:-

http://www.youtube.com/watch?v=EN3m1bJvS4Q

But where as they are using a fluid with a boiling point of +34 Deg C and an actively cooled condenser just to recycle the gas produced by phase change, I'll be using a liquid with a boiling point of -16 Deg C and my ac unit will be used to both produce the temps required to form the liquid by liquifying the gas in the first place and to recycle the gas produced by phase change.

The plus side of my method is that I should have temps which better theirs by some -50 Deg C due to the lower boiling point of the liquid.
So using their figures I hope to be looking at fully loaded temps of -10c on the cpu and around 0c on the gpu's (probably less as they were using gtx480's)

Last edited by technogiant; 01-20-2013 at 01:32 AM.
United Kingdom  Offline
    Register to Reply to This Post
Old 01-20-2013, 02:36 AM   #23
ORL
Bring on the Liquid
ORL's Avatar
Senior Member
 
Posts: 2,141
Last Seen: 07-11-2017
From: Missouri
iTrader: 0 / 0%
Oh I know you wouldnt be able to hit it lol. Was just a funny thought. I havent a clue what the transition points are even for something like a CPUs material makeup. Maybe some day my friend... maybe some day...
United States  Offline
    Register to Reply to This Post
Old 01-20-2013, 05:34 AM   #24
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
Okay I've just done 3 sketches which explains in more detail how this system is proposed to operate, again sorry for the sketch quality.



Uploaded with ImageShack.us

The first sketch above illustrates the chamber filling procedure prior to using the computer.

The freezer is off so that the gas bottle is at ambient temps and the contained gas at high pressure.

Line 2 is opened intermittently to allow gas into the chamber which is condensed out by the ac unit to form the liquid gas pool. Once sufficient gas is pooled line 2 is closed and the pc is ready to use.



Uploaded with ImageShack.us

This sketch represents the situation while the pc is being used, both lines to the gas bottle are turned off, there is no interchange between the chamber and gas bottle/freezer.

Cooling is via evaporation of the liquid at the pc components and re condensation of the vapor produced at the ac unit evap....I may additionally pump liquid over the evap to further lower the liquid temp or perhaps spray it into the chilled gas above...sort of "bong cooling" fashion, as this would allow for greater sub boiling point cooling of the liquid.

Also importantly during this phase the freezer is turned on and both lines/valves are closed. This lowers the temp of the gas bottle to below the boiling point of the gas contained in the bottle and so lowers the sealed bottles pressure below atmospheric in readiness to recover the liquid from the chamber.



Uploaded with ImageShack.us

This last sketch shows the turning off process, the pc is turned off, the ac unit re condenses all the refrigerant gas to liquid and then line/valve 1 is opened to the gas bottle.

The chamber is still at atmospheric pressure due to the expansion sac and the fact that the chamber contains sufficient non volatile gas to fill it (from bottled nitrogen).

The gas bottle is however at lower than atmospheric pressure as it is now at a temp lower than the boiling point of the contained refrigerant, so the gas bottle will actively suck a certain amount of the fluid back into the bottle because of the pressure difference.

Hopefully it will suck back all the liquid but if back pressure builds before that occurs I'm quite confident from what limited experimentation I've done that the remainder will drain back via gravity.

I can maximize the suck back by using a larger bottle as a low pressure in a larger bottle will suck back more volume before pressure builds sufficiently to stop the flow.

Also, again from what I've learned through experimentation agitating the bottle and pipe will cause the fluid to flow back by gravity.....It may be that opening line/valve 2 may also assist the pressure in escaping even though they share the same route as they enter the bottle.

Once the liquid is returned to the bottle the valves are closed and the bottle and freezer are allowed to return to ambient ready for the next run.
United Kingdom  Offline
    Register to Reply to This Post
Old 01-31-2013, 08:04 PM   #25
Orthello77
Overclocker
Senior Member
 
Posts: 410
Last Seen: 12-19-2015
iTrader: 0 / 0%
TG, was just catching up on your thread. Exciting stuff.

From those temperature projections you have , all theoretical of course but if it works out like that you will be totally on par with my load GPU temps (0c or less) but totally blow my cpu load temps out of the water .. which could show up how underbuilt the current waterblocks are for dissipating the typcial heatload of an overclocked cpu. I'm going to push ahead and try and make a larger /efficient cpu water block , as i think that is the problem. Eg the GPUs have huge waterblocks in essence . Easily 7-8x larger surface area and thats why the temps are so much better on the GPUs .. even when they pull more wattage.

Hows the progress coming anyhow ?
New Zealand  Offline
    Register to Reply to This Post
Old 02-01-2013, 02:03 AM   #26
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
Yeah my main concern regards those temp projections are that I won't be able to exactly copy their special coating they use on the cpu and the finned boiler plate they put on the gpu's.....they quote huge factors of heat reduction because of those.

My plan is to de-lid and use that liquid pro metal tim you found to fix a sheet of panasonic pgs graphite sheet to the top of the chip....it will be solid below 8.5c. I'm still waiting on a response from panasonic as regards chemical compatibility of the pgs with the liquid refrigerant.

On the upper surface of the pgs to provide a roughened surface which helps nucleation I'm planning on again spreading a layer of liquid pro on which I'll sprinkle diamond power/dust (which isn't as expensive as it may sound). The pgs sheet will be larger in area than the de-lidded chip top and will spread the heat more efficiently than the manufacturers chip lid.

As regards progress....I haven't started yet...still at the thought experiment stage...I'm a little apprehensive/reluctant because of possible dangers and not to mention the costs and still mulling things over. Potentially this could be a very simple direct and efficient system. I'm going to take a steady/slow approach with it and not cannibalize my current system for it...apart from the ac unit so I can go on building it steadily while I still have the use of my current system.

As regards your theories on your cpu cooler block......that sounds spot on.....my cpu is showing about 140w power usage under full load so as you say if you can keep your gpu's at sub zero when they have over double the power usage then there is no reason you can't do likewise with the cpu......looking forward to how that develops.

PS.....your new pumps are looking great.

Last edited by technogiant; 02-01-2013 at 02:41 AM.
United Kingdom  Offline
    Register to Reply to This Post
Old 02-01-2013, 05:26 AM   #27
ORL
Bring on the Liquid
ORL's Avatar
Senior Member
 
Posts: 2,141
Last Seen: 07-11-2017
From: Missouri
iTrader: 0 / 0%
I wonder if Gallium would work as a low temp adhesive TIM for this project. It should be able to transfer heat extremely well and I do not believe this will separate to badly from aluminum/copper when they contract and the gallium adheres to the components.

Something to consider as the gallium is one hell of a heat conductor. I haven't got a reference thought to compare your mentioned product and the latter.
United States  Offline
    Register to Reply to This Post
Old 02-01-2013, 05:59 AM   #28
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
Thanks for that Orl....I think Gallium is one of the components of liquid pro.....I recall in a promo video they showed a range of liquid metal tim's with varying melting point ranging up to pure Gallium with a melting point of 29c.....If I could get hold of that it would be ideal as it would remain solid and bonded at room temp.....I could work it under a heat lamp.....not sure how well pure Gallium or liquid pro would spread and adhere to the pgs sheet though being a non metallic surface?.....In fact it might well be that as both the pgs sheet and diamond are both carbon based being hydrophobic and lipophilic in nature that neither will bond to a liquid metal tim. Although they should both interact well with the surrounding refrigerant liquid which has a similar nature so long as I can find some way of bonding things together both mechanically and thermally......just emailed the Indium Corporation that make a number of Gallium based liquid metal tim's to see if they have any info on wetting properties on non polar surfaces.

Last edited by technogiant; 02-01-2013 at 12:03 PM.
United Kingdom  Offline
    Register to Reply to This Post
Old 02-01-2013, 01:00 PM   #29
ORL
Bring on the Liquid
ORL's Avatar
Senior Member
 
Posts: 2,141
Last Seen: 07-11-2017
From: Missouri
iTrader: 0 / 0%
Applying the Gallium would be easy with a heat lamp. Just place the desired amount on the chip and secure an HS over it. Then place the whole set under the heat lamp and let the pressure do its thing. As it changed states the spreading would occur. You may want to consider a non conductive barrier around the edges of the chip as well to act as a dam per say.

As for the Pyrolytic Graphite, its a good thermal transfer product but you will find its not as desirable once you change its nature by adding other components such as diamond dust or a metal bonding layer. The key thing to remember here is maximum heat exchange over the shortest span possible. Adding any layers will increase thermal insulation.

I am curious about your theories pertaining to the hydrophobic and lipophilic natures of the compounds and how it will effect things. Remember you will need to be very careful about slop and rubbing excess of any insulation compounds etc as the different natures of compounds will have a similar effect to the hydrophobic reaction of say water and oil. This would in turn render some surfaces nearly non-conductive to heat and insulate the components from transfer.

Last edited by ORL; 02-01-2013 at 01:06 PM.
United States  Offline
    Register to Reply to This Post
Old 02-03-2013, 07:44 AM   #30
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
Well my concern about the nature of the materials hydrophobic/hydrophillic non polar/polar is all about the Gallium's ability to spread on the surface or wet it to use the proper term.

Gallium is being a metal polar in nature and will wet other polar surfaces such as metals because of mutual interaction but will I imagine tend to bead on non polar surfaces which pyrolytic graphite and diamond are.

I'm not intending using a heat sink and actually going to be removing layers rather than adding them.

The cpu will be de-lidded so removing the manufactures tim and heat spreader lid from the equation, with the intention of adding liquid metal/gallium as a tim adhesive to directly attach the pgs to the bare chips surface. But that depends on the liquid metals ability to wet onto both those surfaces.

The addition of the diamond dust is very specifically of importance to the phase change means of cooling.

The process where the cooling liquid changes to gas at the hot surface does not occur very well on smooth surfaces. By adding a roughened surface you increase the ability of the liquid to turn to gas, the process is called nucleation. You can see this in action if you add a teaspoon of sugar to carbonated drink.

Have a look at this video of the novec submersion pc

http://www.youtube.com/watch?v=EN3m1bJvS4Q

They use a special baked on granular layer to increase nucleation which increase cooling efficiency compared to a smooth surface by many many fold...actually 15X.

That is what I'm hoping to emulate using Gallium as an adhesive to stick diamond dust onto the surface of the pgs (in addition to sticking the pgs to the bare chip).

They experimented with direct bare chip cooling, which in my case would be the same as using gallium to stick diamond dust directly onto the bare chip but found that the heat density of such a small area was too great to cool effectively, that's why I have to have the pgs layer, it will disperse the heat laterally to give a larger cooling area, and having greater than x4 the conductivity of copper is probably the best material I could use for that.

So I'm not adding any more layers than is absolutely necessary and have in fact removed two layers by de-lidding.

Gallium, diamond and pgs are all fantastic conductors, probably the best I could use, and having them directly attached to the bare chip together with the increased nucleation effect I hope to achieve should provide very direct and efficient cooling.....but it all depends on the gallium being able to wet to the pgs and diamond.

The other feature I made mention of in regard to polar non-polar properties was that the refrigerant liquid is non-polar and so similar to diamond and pgs in that respect so should wet to the surface well and so enhance heat exchange.

If gallium can't wet to the surfaces as I require then panasonic do a range of pgs which comes with a very thin 30 micron layer of acrylic adhesive on the one side which I may use to adhere it to the chip and diamond dust.....but I have no idea how much thermal resistance this layer would have, although I imagine it to be low as it is so thin at 30 microns, also have no idea if the adhesive is compatible with the submersion fluid.....so that is still a definite plan b or even c at this time.

Additional Comment:

Been looking around for suitable items to make the build and have come across this site.

http://www.gaslow.co.uk/

They have two products of interest to me.

Firstly flexible stainless steel high pressure hose.

http://www.gaslow.co.uk/pages/products/list_2.htm

There's a pdf brochure link about them at the bottom left of the page
Think that would be ideal to link the chamber sump to the gas bottle with no risk of cracking or breakage it would save build complication as I wouldn't have to do a pipe in a pipe to guard against fracture.

Secondly they have a refillable gas cylinder that has two completely separate ports on it

http://www.gaslow.co.uk/pages/products/list_6.htm

I'm hoping its possible to remove the fittings that come as standard and replace them with two separate tap valves...emailed the company to find out.
It would mean there would be no problem returning the liquid to the bottle as the second line could just be opened to prevent pressure build up.

Last edited by technogiant; 02-03-2013 at 07:44 AM. Reason: Automerged Doublepost
United Kingdom  Offline
    Register to Reply to This Post
Old 02-03-2013, 08:59 AM   #31
ORL
Bring on the Liquid
ORL's Avatar
Senior Member
 
Posts: 2,141
Last Seen: 07-11-2017
From: Missouri
iTrader: 0 / 0%
I understand nucleation very well. I was expecting something more like a ramsink placed on the chip after a Delid, a small bit of copper to add surface area for the basis of nucleation. The rules of air cooling apply nearly as well here, more dispersion surface area is better. Even despite the agitated surface of the chip itself I have a very hard time believing this will gain over a small sink attached directly to the chip. Especially if your passively relying on the boiling action of liquid to gas as your fresh coolant drive. Remember when things boil they tend to leave behind bubbles although small these bubbles will add an insulation of sorts.

Although I would not use aluminum or anything with after treatment or coatings one of these is what I envisioned fixated on the chip.
http://www.frozencpu.com/products/32...html?tl=g40c18

Additional Comment:

After reviewing the video you linked pay close attention to the GPUs. They did exactly what I was trying to explain.

Last edited by ORL; 02-03-2013 at 08:59 AM. Reason: Automerged Doublepost
United States  Offline
    Register to Reply to This Post
Old 02-03-2013, 10:54 AM   #32
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
Yes the linked video shows that they use a sort of finned boiler plate which is soldered to the de-nickled gpu heat spreader......but I believe this boiler plate has also been treated with the baked on granular coating to increase nucleation.

They found that with the cpu as it produces less heat simply using the baked on coating on the standard but de-nickled cpu heat spreader was sufficient without an additional finned boiler plate.

I just think that de-liding and attaching pgs sheet directly to the bare chip would be better.....the pgs sheet will not be cut down, it will act as a comparatively huge heat spreader and extend laterally beyond all of the edges of the bare chip for several inches in all directions to provide a larger surface area, also the cooling fluid will be in contact with both the upper and lower surfaces of the pgs doubling the surface area.....certainly much larger than that provided just by the cpu heat spreader surface as was used in the video which appeared to be quite sufficient when they demoed it, regardless of the fact that the bubbles are driving the refresh of the cooling fluid, they have already proven the concept works as is...I'll be working with a much larger more conductive surface area.

Overall it will provide a larger surface area than those finned vram coolers, also as it conducts heat 4X better than copper so heat will distribute through it more quickly than through those heat sinks.

Of course if I can't find a way to bond the pgs then something more traditional like those vram coolers would be a simple alternative, could even gallium them direct to the bare chip....they would still have to have a surface coating to promote nucleation....I'd have some reservation about using something with tightly packed fins though as that would probably tend to prevent the bubbles from leaving so well.

Having said all that about surface area I really think that provided the nucleation coat is efficient then a huge surface area is not that essential, they had their cpu cooled sufficiently just using the heat spreaders area.....I think I'm just stressing about having efficient conduction across a larger surface area in case I can't make the coating nucleate as efficiently as they have....may be of more relevance when considering cooling the gpu's.

Last edited by technogiant; 02-03-2013 at 12:08 PM.
United Kingdom  Offline
    Register to Reply to This Post
Old 02-03-2013, 02:02 PM   #33
ORL
Bring on the Liquid
ORL's Avatar
Senior Member
 
Posts: 2,141
Last Seen: 07-11-2017
From: Missouri
iTrader: 0 / 0%
Well Techno, I say this. Either one of us could be directly right here. This is a very important thing to consider for testing, so I say try both.

For testing to avoid blowing an expensive piece of equipment maybe try both methods on an older cheaper GPU per say in the bath when its time? You can do some testing for proper program loads to simulate CPU temps?
United States  Offline
    Register to Reply to This Post
Old 02-03-2013, 02:07 PM   #34
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
Either way it's interesting to knock ideas around....certainly helps me evolve my thoughts
United Kingdom  Offline
    Register to Reply to This Post
Old 02-04-2013, 05:15 PM   #35
ORL
Bring on the Liquid
ORL's Avatar
Senior Member
 
Posts: 2,141
Last Seen: 07-11-2017
From: Missouri
iTrader: 0 / 0%
Its good for me too bud, I love learning anything I can.

Techno, today when at work I was thinking about this project of yours. Have you considered introducing a circulatory type pump or even a prop to this project? I am wondering if introducing a low speed flow to this would increase your thermal exchance during the boil off.

I mean even something as simple as a lego propeller attached to a low speed motor extending via long shaft to the liquid coolant.



Also another thought, today I also got to thinking about your pressure to boil temp ratio. The more you evaporate the more the pressure will increase in the tank, in turn increasing your boiling point the longer the system runs. Its going to be crucial that your chiller can keep up with the full heat load of the system. You may end up needing to add a second chiller if it cannot keep this managed. I have done no math on this yet as I have no idea of volume yet, please keep us informed!
United States  Offline
    Register to Reply to This Post
Old 02-05-2013, 03:48 AM   #36
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
Hey Orl.....yeah I will be having a circulation method for the fluid, I was considering a submersible fuel transfer type pump......Haven't put a great deal of thought to it but I want to kill two birds with one stone....the pump will cause circulation in the fluid but I'm also going to spray the fluid into the gas phase above (bong style) and over the evap surface. This will allow the fluid to be cooled below its boiling point of -16c.....at idle the ac unit holds about -30c.

One of my concerns was that below the fluid boiling point there would be little or no energy exchange between the ac unit and the liquid as there would be no evaporation/condensation cycle......so the pump will fulfill this.

As regards the boiling point/pressure.....well the system is going to be continually at atmospheric pressure regardless of boil off as I'll have a large 260 liter expansion sac attached to the chamber. So the boiling point will remain constant at -16c. The liquid volume to gas volume ratio for the refrigerants I've seen details on seems to be around 1:240 so that expansion sac could accommodate 1 liter of fluid in vapor form which is far more than I anticipate......in that Novec video the whole system was cooled with less than 1 liter of fluid so I think I will be able to scale down on the expansion sac.

From using the ac unit with my current chill box built I know it can maintain a fully loaded temp of -22c with my current components...... so that's one of the reasons I'm planning on using HFC227ea....with a boiling point of -16c there is a fair bit of leeway between the BP ad max loaded temp of the chamber...will also hopefully be sufficient leeway to add some higher power components when I upgrade my gpu's.

But if that does become an issue I would have to find a different refrigerant with a higher boiling point.

Keep the thought's coming Orl...it's all good stuff.

Ps.....one concern I do have about the expansion sac is chemical compatibility with the refrigerant gas.....I currently use a large space hopper toy for this.....I have no idea what type of rubber/plastic this is but it has a very greasy/oily feel. I'm concerned there may be an issue with the refrigerant extracting the plasticizer over time. This could cause it to loose flexibility and crack, additionally I don't really want plasticizer depositing all over my components. If anyone has any thoughts on a different solution for an expansion/contraction chamber I'd be grateful.

Additional Comment:

one way around this would be to have a large ridged container in line between the chamber and the expansion sac. It would act as a large accumulator. As the refrigerant gas expands from the chamber top into the bottom of the ridged container it would displace air from it through a breather in the top of the container into the expansion sac.
The refrigerant being so much more dense than air would pool at the bottom of the ridged container while mainly air would be pushed into the expansion sac so removing chemical compatibility concerns.

Additional Comment:

Just had a reply from the Indium company about gallium alloy wetting to non metals....says he suspects it will wet similarly as it does to glass and quartz which are of course non-metalic......uuhhmm....I still have my doubts though, although they are non metalic those surfaces are still "polar" in nature and so would interact with metal.....graphite is non polar....so not any further forward on that really...will just have to buy some and see me thinks.....there will be an alternative even if this part doesn't work for me so its no major problem....could even try artic silver thermal adhesive?

http://www.overclockers.co.uk/showpr...04-AC&pup_c=gs

Last edited by technogiant; 02-05-2013 at 03:52 AM. Reason: Automerged Doublepost
United Kingdom  Offline
    Register to Reply to This Post
Old 03-19-2013, 03:53 PM   #37
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
*sigh*........getting bored now.....tech news is all pretty static.....bored bored bored.

Gonna have to spend my hard earned on something......so I've been working out exactly what materials I'm going have to order to make a start on this.

So think I'm going to make a start on this now...just trying to find a cheaper source of the HFC 227ea /FM-200....messaging a Chinese supplier on alibaba.com

Additional Comment:

Okay....well....I'm still knocking ideas around and evolving this design and I've had a change of plan.....I guess a lot of you are thinking "make it or shut up" by now, and tbh I'm feeling that way too.

I've been a little reluctant because of the dangers of having >10 liters of refrigerant sloshing around and think I've finally come up with the solution.

Instead of using a full submersion method I'm going to go with a thermosyphone build.

It won't cool as well as a full submersion build but it will be a lot safer only requiring a fraction of the liquid refrigerant.

It's going to be a two chamber design, the inner thermosyphone (ts) and the outer chamber.

The ts will consist of cpu/gpu waterblocks placed below the air con evap. The evap will be contained in a fully sealed sleeve with an expansion chamber exiting from the top and from the bottom short pipes will connect to each of the ports on the water blocks.

The evap will condense the refrigerant gas which will pool in the waterblocks below it...when under heat load refrigerant will boil off and be forced through the evap fins so re-condensing it. The expansion chamber will allow for volume changes that will occur during the process to prevent pressure build up and would also be large enough to contain all the refrigerant in gaseous form should the system fail as only a few hundred milliliters of refrigerant liquid will be required.

Of course the ts will itself require to be in a sealed/chilled chamber to prevent condensation and to remove heat from the other mobo components not being directly cooled....

I'm undecided about the outer chamber.....my initial thought was to use a freezer.

That would have the advantage of being able to keep all the refrigerant as liquid between uses. I'd make parts of the sleeve containing the evap from metal so the evap would be able to assist in cooling the freezer chamber and so still carry the load of cooling the other components rather than hope the freezer could cope with that.

The other alternative would be to just use a sealed chill box type chamber, again cooled by the conductive sleeve of the evap but in this case the refrigerant would be allowed to return to gas between uses and would have to be contained in the expansion chamber.

Additional Comment:

Huummm.....well....having second thoughts about the thermosyphone design,...I'm not convinced the waterblocks and pipe work will deal with the volume of gas that will be generated, they may just gas up and stop cooling effectively......also the shape of the evap wouldn't lend itself to enclosing in a sleeve easily.

So I've come up with a sort of a hybrid of the thermosyphone and submersion design.
It will still allow for open submersion cooling of the cpu/gpu and still use direct cooling of the cpu/gpu die using my idea for the graphite/diamond heat sink rather than a waterblock. Although it has the advantages of the submersion design it would use a far lower volume of liquid refrigerant, it would be similar to an open thermosyphone that is actively pumped and would so be much safer due to the lower volume of refrigerant liquid, and being open would not have the gasing up problems that may occur with the thermosyphone.

The chamber would consist of two levels one above the other. the lower level would consist of a small sump to pool the liquid and contain a submerged pump.
The upper platform would have the mobo on it, the cpu and gpu would be enclosed in small enclosures that would be open at the upper end, the lower end would also be open but would be sealed around the cpu/gpu. Liquid would be pumped from the sump and pour from an open pipe into the open topped enclosure and so pool at the bottom of the enclosures in direct contact with the cpu/gpu. As these enclosures overfilled they would simply over spill and the liquid run back to the sump below.
Liquid boiled off by the components would be re-condensed by the evap and run back into the sump.

Sorry to continually sound board you like this guys but I'm pretty sure I've now arrived at the best compromise with this design....any thoughts?

Last edited by technogiant; 03-19-2013 at 03:55 PM. Reason: Automerged Doublepost
United Kingdom  Offline
    Register to Reply to This Post
Old 03-20-2013, 01:05 AM   #38
Drewmeister
The Need 2 Freeze
Drewmeister's Avatar
Senior Member
 
Posts: 2,455
Last Seen: Today
Age: 52
From: Phobos Lab
iTrader: 22 / 96%
I like your thermosiphon idea the best so far. Actually it would be a gravity fed heat pipe in this situation. I think if you can keep the inlets to the water blocks/boilers flooded then there won't be any problems with proper vapor return to the condenser.

My suggestion would be to have a self contained AIO(all in one) type system inside the chillbox. It would consist of a thin horizontal condenser above the pc compartment and that will gravity feed the water blocks with liquid refrigerant. Vapor return back up to the top rows of the condenser. The beauty of this system would be that you could use r134a by tuning the pipe for around 9psi/-15c at zero load.

Your latest plan would require a whole rework of the chamber right? I guess the pump would make or break this setup. Any loss of pump at high oc and load could kill your hardware. It would probably have to be a diagram type pump with a float switch.
United States  Offline
    Register to Reply to This Post
Old 03-22-2013, 04:27 AM   #39
technogiant
Overclocker
technogiant's Avatar
Senior Member
 
Posts: 335
Last Seen: 09-15-2014
iTrader: 0 / 0%
Hi Drew, I think all the ideas have their own strengths, weaknesses and problems.

The sealed thermosiphon would have the advantage of containing all the refrigerant ensuring that the minimum quantity would be needed and it would be in "the right place" to operate. Additionally being gravity fed it would be completely reliable. (as you pointed out, pump failure in my last design would be disastrous).

But it poses problems in having the evap completely sealed, it would be problematic in cooling the rest of the system. Also the shape of the evap does not lend itself to making a tight fitting sleeve and so would not keep the evaporant as close to the cooling fins as I would have liked. I don't think I'd be able to make the sleeve resistant to 9psi to use R134a, even if the pipes and blocks could cope with it those pressures quickly add up to large loads over bigger surface areas such as the size of the sleeve required to cover the evap.

Additionally the use of waterblocks would not be as direct a cooling method as I had wished to have with the submersion method and there would I think be some problem at full load when the system would require to vent the approx 700ml/sec of gas along pipes in one direction while simultaneously allowing the returning fluid to travel along the same pipes in the opposite direction.

I think the full submersion method while offering the most effective, simple and reliable method is out of the equation simply because of the volume of refrigerant liquid that would be required in an open system could cause dangers.

But having said that I want to retain the direct cooling of the cpu/gpu that the submersion method provides by using open ended cooling pots instead of using waterblocks, and an open system as this would let it breathe properly.

So I'm looking for a solution similar to my last proposal but with less danger of failure.

Having an open system is difficult as gas/fluid will find its way out of the return system, perhaps condensing on the chamber walls and running to the bottom of the chamber and would no longer be available to the cooling pots.

So I'll have to retain the sump and pump to return "escaped" coolant, that was one of the advantages of the submersion method, having the mobo essentially submerged in the sump eliminated this issue.

Also as the cooling pots would contain quite a volume of liquid they will also have to drain slowly from a small hole at their base into the sump so that the liquid can be recovered from there to a pressure bottle between uses.

So if I'm to retain the direct/open cooling as I wish without the submersion method a pump is going to be an essential part of it, even if it was not part of the cooling circulation it would still be required to return the "escaped" coolant and the coolant that will continually drain from the cooling pot drain holes. If the pump failed, even if the cooling circulation was fed by another means, the coolant would just empty into the sump with in minutes and so the system would fail.

So all things considered to retain the features I want without the submersion method I have to rely on a pump.....I could use my last proposal and have two pumps just as a fail safe.

An alternative would be a hybrid between the cooling pot and the submersion methods. It would essentially be the same as a full submersion build but would have cooling pots for the cpu/gpu. These cooling pots would be open at the top for venting and the sides would have many many small holes which would allow for the cooling fluid to enter from the submersion sump. The remainder of the submersion sump would be filled with something like glass beads to reduce the volmue of the sump. So the cooling pots would provide a an open free space for the liquid to cool the cpu/gpu which would be unrestricted by the glass beads. I don't think the fluid entry speed to the cooling pots would be an issue as only 1-2 cc of fluid per second would be sufficient to cool 200-400 watts

This would give all the advantages of the submersion system but a greatly reduced volume of liquid, I guesstimate the volumes would be 14-15 liters for full submersion, 4-5 liters submersion/cooling pot hybrid, and 1-2 litres cooling pot pumped method.

Additional Comment:

^^^^ edited

Additional Comment:

I've been rethinking out and revising the fill/recovery cycle of the liquid refrigerant, and believe I've now got it right...at least as far as I can tell in theoretical thought experiment terms.

I'd simply use a standard bottle upside-down linked to the chamber by a single pipe bottom filling/emptying from the chamber sump. A freezer would be used to create the pressure differences to move the liquid.

I know your not meant to store pressure cylinders upside-down but as far as I can ascertain that is only because the liquid phase in the bottle would be covering the pressure release valve rather than the gas phase, so if any excess pressure did develop in the bottle then liquid would be ejected from the safety release valve rather than vapor...which obviously once in the open would expand to a much greater volume of gas and be a more significant loss.

Excess pressure is not going to be a problem as the gas bottle is going to be in a freezer and I will ensure it is not overfilled with liquid so there is plenty of vapor space in the bottle.....the majority of the times the bottle will be at negative pressures.

The revised fill and recovery cycle is really quite simple.

From the starting point with the bottle in the freezer obviously first chill the chamber down so you don't get explosive decompression, then remove the bottle from the freezer in and place in an upside-down position allow it to warm so that the bottle pressure starts to build, open the tap valve and the bottle pressure will push the liquid phase which is obviously covering the valve opening along the hose into to chamber sump.
Once filled, close the valve and replace the bottle in the freezer.

The recovery is a little more complicated because of the way the tap valve works, the sealing piston is not directly connected to the valve screw, so I anticipate that when the bottle is under negative pressure that unscrewing the the tap will not release the valve piston as the negative pressure will hold it closed.
So the initial stage of recovery will strangely involve again allowing the bottle to warm to build some pressure and pop the piston seal open, once opened the bottle would again be place upside-down in the freezer this time with the valve open.

In this situation the vapor phase in the bottle would be separated from the vapor in the chamber by the liquid phase in the bottle/hose/chamber sump all of which would be in a continuous "U" shaped bend.

The pressure in the chamber would be at atmospheric due to the expansion sac and the fact that it also contains sufficient volume of non volatile gas to fill the chamber (from bottled nitrogen....well non volatile in the temp range I'm considering....lol) while the pressure in the bottle vapor chamber would be below atmospheric as the temperature would be below the boiling point of the liquid and this vapor phase would only contain volatile vapor (no nitrogen).

So the pressure difference would be equalized by liquid being pushed into the bottle. This of course would then mean that the bottle vapor pressure would become the same as atmospheric, but as the temperature is still below the atmospheric pressures boiling point then vapor would phase change back to liquid so restoring the bottles negative pressure again to below atmospheric. The heat given off by the phase change back to liquid would be absorbed by the freezer and buffering ice packs and this energy would essentially be what is driving the liquid movement.

From an energy consideration standpoint the max volume of vapor that would be needed to re-condense would be 15 liters of vapor which equated to about 63ml of refrigerant liquid, the heat of vaporization per ml of liquid is about 200 joules so a total energy of 12600 joules.....If a freezer can deliver 100w of cooling or 100 joules per sec then it would take just over 2 minutes for this process to occur......in any event the freezer will also contain ice packs that will buffer this energy absorption requirement (and any loss of freezer function at a critical moment which may otherwise prevent liquid recovery).

So as you can see this would be a continuous process that would occur without any back pressure build up until all the liquid is recovered at which point you turn the tap valve off.

The only other problem I can envisage would be if nitrogen were to get into the bottle vapor space, that may affect the bottle pressures....but that would be easily removed just by purging the bottle at the chamber fill stage. Initially rather than having the bottle upside-down just put it the correct way up, the boiling liquid phase would then push the vapor and nitrogen out of the bottle, probably the nitrogen first as it is less dense and would be at the top of the vapor space, once purged like this simply turn upside-down and the liquid will then be pushed out.

TBH I'm thinking that as I'm really confident that this method of recovering the liquid refrigerant will work I'll probably go with the full submersion method. It's by far the simplest and potentially most effective cooling method of those I've been considering and was originally what I wanted to do until I started to have concerns about not being able to recover the liquid refrigerant to a pressure bottle.

Last edited by technogiant; 03-22-2013 at 01:51 PM. Reason: Automerged Doublepost
United Kingdom  Offline
    Register to Reply to This Post
Old 04-02-2013, 09:11 PM   #40
Drewmeister
The Need 2 Freeze
Drewmeister's Avatar
Senior Member
 
Posts: 2,455
Last Seen: Today
Age: 52
From: Phobos Lab
iTrader: 22 / 96%
It looks like you have this all worked out pretty good. Hopefully you can find a cheap dual port recovery bottle with the larger 1/2" ftgs. The recovery bottles have a dip tube on the highside so no need to turn bottle upside down.

The heat pipe system that I mentioned before would be separate and completely sealed from the chamber. It's basically a stand alone Gravity Fed Heat Pipe System. See drawing. The gas condenses in the radiator and drains down to the cpu/gpu boiler chamber where it pools and boils off.. the superheated gas rises up back to radiator where it's cooled/condensed.

The piping is large diameter as to not impeded flow in either direction. The cpu/gpu boiler chamber has large volume for capacity. The base would be a boiler enhancement plate for achieving lower core temps. Charge port is soldered to the radiator, all joints can be soldered as pressure is only 80-100psi tops if using r134a.

Anyways.. just throwing some ideas out there.

Any Updates?
Attached Thumbnails
Click image for larger version

Name:	aio gfhp 1.jpg
Views:	110
Size:	108.5 KB
ID:	166935  
United States  Offline
    Register to Reply to This Post
Sponsored Links:
Please Register to Post a Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -6. The time now is 02:43 PM.

Copyright ©2000 - 2016, Jelsoft Enterprises Ltd.
Powered by vBulletin
Copyright ©2000 - 2016, EXTREME Overclocking