Window AC chiller box/phase change build

[technogiant]

Thanks for that Orthello.....my main problem is going to be space in the chamber.....I didn't design it for this.

I can't mount the 360 rad vertically....but I've found a place I can just about squeeze it in lying horizontally....just in front of the evap between it and the mobo.....it will be quite tight but should be enough room...may have to use 90deg elbows coming of the cpu block.

My gtx 360 has G1/4 fittings...is that the thread size?.....will I be able to get 1/2 ID tube for that?....new to all this.

Also just wondering about the coolant mix your using......why have you got glycol in there....it is quite viscous....40% methanol by volume should be enough to prevent freezing....is there a reason your using that?...just thinking with my smaller pump I'll need to keep things flowing as well as possible.

[Orthello77]

The 90 degree bends out of the cpu will be fine if you need to , they impede flow slightly but you shouldn't have much restriction at this stage if its just CPU / Rad.

G1/4 is the standard thread size. Yes you can get G1/4 thread 1/2 " ID barbs or G1/4 thread 5/8 " ID barbs. 5/8 ID hosing is easier to bend if you are in a tight situation. The primoflex LRT tubing is really good for not kinking even with tight bends , i have used this piping for over 2 years now.

Well tbh i think the glycol is there (Drews recommendation) for its anti-corrosive properties , an i'm not sure if that comes from the 5% in the Zerex mix that's not ethylene glycol. You are right a 40% methanol mix should prevent freezing as methanol's freezing point is -90c from memory. I did have flow problems until the new pumps so with that in mind you could just try a methanol / water mix and that should be fine also. Not that i'm a fan of colour additives but the zerex has got a purple colouring to it which i must say has been usefull for spotting slight leaks.

[Drewmeister]

Great advice there orthello.. I especially agree on trying to reduce 90 bends. If you can't fit a res then a T-line will do.. just takes longer to fill/bleed. The Zerex is needed to prevent galvanic corrosion between the dissimilar metals and to reduce the total methanol volume for safety. It's possible to blend a low viscosity -30c mixture which is non-flammable to open flame.

TG, your loop won't have dissimilar metals and will only have a fraction of the volume of orthello's. Your load temps will most likely be in the low -20's so 55% meth/40% distilled /5% any glycol + small bottle biocide should work out just fine. Given the temps and fluid used, I highly recommend stainless hose clamps. Make sure the tubing has good resistance to alcohols. I've had excellent results with clearflex 60.. going on 6yrs and no problems.

Waterblock: Names_Lucky recently linked me to a nice block.. the XSPC RayStorm. It's low restriction and the top is made from black POM so has excellent temp range.

Hope you can fit your GTX360 in there without too much trouble... to bad it isn't a stealth. It'd be nice if you could fit it behind the evap.. then add two 25mm pushers on the front of the evap to make up for pressure drop through the fins. Anyways.. good luck with that.

Sorry to hear the new heat pipe didn't work out. At least now we all know with 100% certainty that heat pipes don't work well below 0c. Thank you very much for sharing all your results with us.

[technogiant]

Sorry Drew we cross posted...but this was my thoughts before your up date....

I'm thinking I'll go with a 40/60 mix methanol/water together with these two products-

Silkolene pro cca ultra is a corrossion inhibitor without anti freeze.

http://www.poweredadventure.co.uk/pr...prod=600485290

And Redline water wetter to improve cooling.

https://www.europaspares.com/product...___redww_.html

As a biocide I'll use couple of drops of coppersulphate based biocide.

The only reason I can think of for also using the glycol is as a back up to mitigate the effects of the methanol evaporating off from the solution as glycol wont be lost by evaporation so easily...at ambient temps a 30% glycol mix has 5x the viscosity of water while a 40% methanol mix is only 1.5X that of water....thats at ambient and will become a lot worse at lower temps....think I'm going to leave the glycol out and monitor fluid levels in the res and top up with pure methanol to mitigate evaporation.

Additional Comment:

That looks like a nice water block Drew...think I'll go with that....only one concern is the acetal top...is that alright with methanol?....they do a copper topped one but much more expensive?

Yeah its a real tight squeeze with the gtx 360...its the only place I could actually put it....I'd actually like to get two 240 dual fan rads that are slightly thinner....they would mount nicely flush on the back of the evap.....but its all extra cash so I'll try the 360 first...I won't have to do anything irreversible to the chamber to accommodate it....I'm just a little concerned its not going to get all the air flow it requires.

[Drewmeister]

That blend looks fine... you can always add more methanol if needed. I only recommended the 5% glycol as a cheap inhibitor.. chances are you have some at home already. I've heard that the Redline acts as a biocide too.. never hurts to add extra measures though.

Yeah.. I really like the design and specs of the XSPC RayStorm. Acetal has excellent resistance to alcohols and is rated down to -40c. That doesn't mean it falls apart at -40c.. it just starts to loose some of it's toughness.. like impact resistance, fatigue stress, etc. Of course a metal top is ideal.. I don't think you need to worry at -25c range though. The weakest link on these blocks are how well the two halves seal. The acetal top looks very solid and should provide a good long lasting seal for you. I recommend 4min jb weld on the threads of the barbs.

I was thinking the same thing on the rads... two 240 stealths would be perfect for cpu + sli/cf. Maybe even go with one Black Ice GT Stealth 240 for now for cpu.. then add another later. Of course all the better if you can fit the GTX360.. may just need two 120x25mm booster fans to get the airflow. Those can be placed right on the back of the evap... maybe with a small spacer as they would be pullers then.

[technogiant]

Drew, carrying on from your update in Orthello's thread I thought R134a had a boiling point of -26c at atmospheric pressure?....that would be too cold for my uses, it would have to be warmer than the loaded temp of my chamber which is about -22c...and I'd want to allow for higher chamber heat loads in future so I think about -15c would be the lowest boiling point that would be of use to me....isobutane or R600a would be good at -11c but of course it is flammable ....I wasn't really considering making the cooling loop under pressure or vacuum but was hoping to use the cold chamber temp rather than pressure to liquify the refrigerant at normal atmospheric pressure and an external expansion chamber to absorb pressure changes due to refrigerant vaporization during use and to contain the gas at ambient temp / atmospheric pressure between uses.

So I want to use standard water cooling tubes, connections blocks, rad and water pump.....rather than more complex and expensive pressure fittings and equipment.

[Drewmeister]

Flare and compression ftgs are cheap.. just need a pump with a good seal. Pressurize the system to 9psi for a BP of -15c with r134a.

Other than that.. maybe look into making your own heat pipe. You can cool the hot end with chilled liquid... maybe just some copper coils. The working fluid could be r134a. I think it would be fairly easy to construct.. maybe just solder some end caps onto a copper tube with an access port for charging. You could stuff some copper wool in there for increased heat transfer to the shell. There's some videos on youtube about this stuff.

Anyways.. just some food for though.

[technogiant]

Additional Comment:

Some interesting ideas there Drew.

But I'm starting to think making a pressurized loop within the chamber is going to have many complications, firstly to allow for the volume changes that would occur due to the phase change I'd require a secondary smaller expansion chamber inside the cooling chamber. The outside expansion chamber could not be used for this as if it was required to be at 9psi when the chamber was cold then it would not be able to accommodate the gas volume increase at ambient temp. Also a gas compressor/injector/pump of some description would be required in line with the external expansion chamber to pressurize the loop and a cut off valve isolating the external chamber once pressure was reached......its all getting more complex than I want.

I like the more simple idea of heat pipes filled with R134a or other refrigerant...but again if I was constructing this from scratch it would be difficult to make an adequate heat exchanger to cool the heat pipes.....may well be better off just changing the cooling agent in a commercially produced heat sink.

I'm after as simple a solution as possible.....I'm wondering how much positive pressure my chamber itself is able to cope with rather than using a cooling loop.....if R134a requires a pressure of only 9psi to liquify at -15c what pressure would it require to be liquified at the loaded temp of my chamber of -22c?...my chamber itself may be able to contain that pressure and produce a pool of liquid refrigerant at the bottom of the chamber.

I think being able to produce a pool of liquid refrigerant in the bottom of the chamber would open up various cooling options.

The simplest would be just pumping liquid refrigerant direct to cpu/gpu water blocks which are open on the exit side evacuating directly back into the chamber.

But it would require enough depth of liquid refrigerant to use a small self priming submersible pump. Obviously the amount of liquid refrigerant I'd be able to use would be restricted by how much volume that would occupy at ambient temp and atmospheric pressure as I'd have to contain it in an expansion chamber between uses unless I had some means of actively pumping it back into the gas cylinder.

If it's not possible to get enough depth of refrigerant for a pump then a water loop could be used with the radiator laid flat in the liquid refrigerant instead of air/gas cooling it.....so it would be phase change cooled......I'm going to have to find out what pressure is required to liquify R134a at -22c.....it's only 4c above its boiling point at atmospheric pressure.

Additional Comment:

Just found some details on that.....at -23.3c pressure to liquify is 1.9 psi and at -20.6c its 4.0 psi........huummm working that out that would give over 500lb to 1000lb of pressure on the area of the lid.......2-4psi doesn't sound much but when you work out the total force over a given area it sure adds up.....don't think I'd like to try that......I'd have to use a mixture of R134a and another refrigerant with a higher boiling point like isobutane boils at -11c at normal atmospheric temps.....if mixed with R134a at least it would not ignite....there wouldn't be any oxygen present.

[Drewmeister]

The condensing pressure under load would be too high for the chamber.. it would have to be closed loop like I mentioned. I think you could make it happen with a well sealed pump. Loop order, small res, pump, blocks, condenser.. gravity fed back to small res. You may not even need a pump.. consider it a heat pipe loop.

Recharging that small heat pipe with r134a won't be easy and would probably give worse temps than your solid hs. If you had the resources, time and money then a custom heat pipe might be worth attempting.

Have you considered something like a Corsair H80 or so. Slappa had his down to -20c with no problems. I'm sure you could change the fluid if necessary.

I think you should at least try the r134a gas.. just charge to 1-2psi at room temp. Fill very slowly from the bottom and bleed out top to displace the air.

[technogiant]

Yeah I also thought of and like the idea of a closed loop without a pump....as you said sort of like a heat pipe, the problem with that with my build is that the graphics cards are towards the top of chamber and the fluid return would be via gravity, I'd have noway of placing the rad above the cards.

Quite interested in the Corsair H80 idea.....did he reach -20c without changing the fluid?......I see that one mrf is also making a self contained hybrid water/air cooler for graphics cards....may be of interest.

I do want to try and change the chamber air for a different cooling gas.....but I'm not sure if I've thought it out correctly.....I was thinking that a gas with a great density than air and specific heat capacity would be better.....but I've been scouring the web for info and come across this post about hydrogen cooled generators.

http://www.control.com/thread/1267097548

Apparently hydrogen has a 10x fold greater cooling capacity than air.....not that I would contemplate using it for a minute...far too dangerous and difficult to contain...but reading the thread it struck me that there were other properties of hydrogen that I hadn't considered that makes it more efficient such as its low viscosity, low density, thermal conductivity and its ability to transfer heat......so I'm not sure my thought process is correct, I would have thought low density would have been counter productive.....I've posted up a question there hoping that someone will have some further info.

Actually this talk of using R134a instead of air has given me an idea......why even have the ac unit evaporator?.....cut it off and have the capillary tube and suction tube directly entering the chamber and use the chamber as a large combined evap/accumulator......there's actually no need for any intermediary cooling medium. Between uses the refrigerant would be allowed to expand into the expansion chamber at normal atmospheric pressure. Could even route the incoming refrigerant from the capillary tube directly to cpu and gpu cooling blocks which then open straight into the chillbox for direct phase change cooling of both cpu/gpu.

just uploaded some pics of my bios settings:-



Uploaded with ImageShack.us



Uploaded with ImageShack.us



Uploaded with ImageShack.us



Uploaded with ImageShack.us

[Drewmeister]

All your bios settings look fine as far as I can tell. I don't see any settings for cpc or command rate.. that's probably set automatically with the strap. Still.. even at 1T those settings are a walk in the park for your ram. I'd test that ram with Memtest86+ just to make sure. Maybe try auto settings for the cpu:ram divider and ram timings. Update bios? Also, your mobo may not be capable of the 500+ fsb like in the reviews, drop the multi and check max stable fsb.

The vrm's and chipset have a heat pipe.. wonder how those temps are? Maybe it's frozen causing the nb to run hot? If that's the case then maybe just add a fan blowing on the nb hs.

You may have just hit that temp vs speed wall.. need colder!

I'm almost postive that Slappa's H50 was stock. I've also heard they are filled with a 60/40 mix.. not sure if that's antifreeze/water or the other way around. Either way.. that's plenty for -20c range

As far as the chillbox evap chamber idea. I think you'd run into oil return problems... even with an oil separator. Direct die evaps would produce lower core temps anyways.

[technogiant]

Yeah.....wasn't thinking about oil return .....Having said that,I'm not sure how oil return is working with my evap as it is....both incoming and outgoing pipes exit from the top of the evap, so any oil carried to this point would pool in the bottom of the evap?

Thinking further not sure that the compressor would produce enough pressure to liquify the refrigerant working from a base line pressure of atmospheric.
Normally the systems are under high pressure even in "off" state and the compressor produces a higher pressure on top of that to liquify the refrigerant.

I don't know how much pressure the compressor would create on its own.....may work with a refrigerant with a higher boiling point......but as you said it may be a problem with the oil return......in that case I'll keep my hack saw away from my evap pipes .

I too had wondered about the heat pipe on the mobo heatsinks....I'm running quite a high northbridge and fsb voltage....heat may be a problem there.

Good idea to use memtest...hadn't thought of that, may help in further isolating the problem.....and I'll try auto ram settings...something strange is going on with them..they keep resetting themselves whenever I enter the bios.

Additional Comment:

Found the problem with my ram......it was a bios bug that resets the ram timings every time you enter the bios.

I have to put the ram timings into auto before I exit and save.....if I want to slacken them I have to put them into auto then back into manual alter them then exit/save.....what a mess around...little wonder I was getting confused.

Still now I'm able to put my ram on a higher strap 1:2.5 and getting it to greater than 1000mhz which is where it should be.

I've settled for a more conservative overclock of 3.7GHz with ram at 1025MHz now which is stable across the board.....54% overclock above stock and stable can't be a bad thing. :-)

Additional Comment:

Although I'm stable for all the benchmarks I've been using I'm still getting quite frequent random crashes in BF3 multiplayer......really really annoying.

So I've updated the motherboard bios to the latest available....seems they have addressed the ram setting issue......I also see that the system is by default putting 2.6v vidimm through the memory? it's rated at 2.2v and the original bios was putting about 1.5 though it at default?...its a combo ddr2 /ddr3 board...think perhaps it was using the ddr3 settings before for my ddr2 modules.

Anyhow things seem a little more stable I was able to run through a couple of less demanding maps at my higher settings of 4.1Ghz....but it wasn't stable and almost instantly crashes on more demanding player filled maps.....dropped it down to 3.80GHz quite stable but still getting the occasional annoying crash.......think I'm going to have to go through my overclock settings again with the new bios......like you said Drew my board goes to 500fsb....I'm topping out at 450-460 so I need to get that straight first....perhaps I'll be able to now that bios bug isn't confusing things.

[Drewmeister]

The oil is miscible with the refrigerant and there's plenty of velcoity to carry it through the evap. Sub -40c is when changes would have to be made as that's very near to or exceeding the pour point temp of the oil. An off the shelf a/c would need mods to achieve that temp anyways.

On the subject of pressure inside the chillbox evap chamber idea.. operating pressures would be no greater than static. The capillary tube creates a huge pressure drop so all you'd be seeing is the resulting suction pressure... r22 =28psi @ -15c. r134a = 27psi@ 0c. But yeah.. pressure still too high and definitely a no go with the oil situation.

Glad to hear you got the ram issue straightened out and are able to run it close to factory. I take it that's at 5-5-5-15-22-2T? Yep.. sounds like both bios messed up the correct SPD settings. The original was defaulting to ddr3 at 1.5v and the latest is defaulting to early ddr2 at 2.6v. I think you only need 2.3-2.4v max for any oc'n beyond 1066... btw, read they can do 1160 @ 2.3v.

I've had my share of crashes in BF3. Turns out 90% of mine were due to random packet loss I was experiencing over a three week period. The other 10% mainly from finding max gpu. You're probably crashing from the cpu's oc/heat threshold. Obviously you'll need every bit of that cpu for the sli. You can see how long it takes the cpu and gpu to render frames using the Render.PerfOverlayVisible 1 command. Lower is better.. higher is bottlenecking. Ideally the two lines will be in sync.

[technogiant]

Also sorry for the delay....having been torturing this old hard ware to the limit and given my chill box a thorough testing out I've decided to finally upgrade.......just bought myself a 2600k....would have liked a 2700k but there was an offer on amazon on the 2600k which I couldn't refuse....the 100MHz on the 2700k doesn't guarantee anything at higher clocks anyway from what I've read. Also gone for a Gigabyte Z77X D3H mobo.

Should be getting the parts by the end of the week.....I'm going to try my Zalman 8900 heat sink for size just to see if it fits.....if it does I go for a solid copper 7500....it should fit okay the gigabyte board doesn't look too cluttered.

I'm gona do some preliminary tests with air cooling on the new set up and then exchange the air for R134a....should be interesting....If I'm not happy with the improvement I'll think about putting in a cpu water loop to start with.....I'll holdback on the gpu's till I upgrade them...but that wont be till a while later next year.

[Orthello77]

Gee all happening aye TG , great to hear from you !!

Can see the 5ghz+ mark coming for you ... nice one :-)

R134a , looking forward to seeing what that does. I'm wondering if it works as anticipated then it could possibly work for my setup also , although the gains for me would be just mainboard / ram temps. It all helps.

I've been thinking about upgrading also, to the asus 4re / sandy and doing the vga hotwire trick as that mobo supports that you don't have to muck around with variable resistors then to hard mod the cards. Still i have to remove a resister and solder 2 cables in for each card connecting to mobo. to enable the hotwire. Not so happy about removing the resister. Still i'm thinking if i did all of that i might get to 1100 mhz or higher with these 580s.

[Drewmeister]

That's great news TG!! The 2600k is a great choice.. it beats the 2700k in every bench on the HWBOT. It looks like there's plenty of room for a larger hs on that Gigabyte Z77X D3H... worst case move the ram to the outer slots.

Yep.. 5ghz+ soon.. my guess 5.5.

[technogiant]

Yeah I'm looking forward to getting this up and running, I ended up getting a 2700k in the end ( before I saw your last comment).....still I'm sure it will be fine...they are exactly the same chip just different bins.

This time, because of the problems I had with my previous memory I've gone for some that has been specifically tested on my mobo.

http://www.technextday.co.uk/product...prodid=1000222

It hasn't arrived yet as I changed my order whilst they were processing it...should be with me within the next week.

I've also ordered my zalman 7500 cu cpu cooler.....I'm sure it will fit the memory is fairly low profile at 40mm.....if not I'll trim the heat sink fins rather than move the ram to the next slot.

I'm thinking I will also reverse the air flow in the chamber, the zalman blows air downward towards the mobo and my chamber flow is going in the opposite direction....so its probably best to reverse it.

I think just for interest I'll test out the new zalman heatsink on my old setup first to get a comparison and see how much improvement its given.

Additional Comment:

Just installed my zalman 7500cu on my old Q6600.....max core temp on the intel burn test dropped from 54c to 39c......nice 15c temp improvement just through getting a better heat sink.....nice....next to see if reversing the chamber air flow helps....atm it's going the opposite way to the cpu air flow.

Additional Comment:

Reversed the air flow.....wasn't expecting too much difference but it dropped the max temp another 9c to 30c......so altogether I've knocked 24c off my max temp just with a little tinkering

....next up is the 2700k upgrade.....then when I happy it's running well.....exchanging chamber air for R134a!!!

[Orthello77]

Jeez ... TG .. way to go !! 24c from just heatsink and fan reversal ... nice one.

Your IBT temps are better than mine now too , ... just can't seem to get mine lower .. but i've checked on the net and yeah bloomfield i7s are just hot puppies. I even lapped this chip i'm using now (see the thread) and yeah no big change , possibly my prior cpu was reasonably flat to start with.

Well done mate .. bring on Sandy !!

[technogiant]

Yeah I'm well pleased with that, wasn't expecting such a big drop.

I think your pushing your cpu harder than I am, hence my slightly lower temps. I suppose I could used the reduced temp to squeeze a little more out but tbh I think my ram is causing instability so I'm not going to bother any further with the Q6600....just wanted to try the new heatsink and try the air flow reversal on the old system so I had some frame of reference/comparison.

I've assembled my new 2700k system and put it in the chamber but have come to a slight problem.

The 24pin atx mobo block is in a different position on the new board and my cable wont reach, I'm going to have to order another extender. Just hope adding another extender won't affect my voltages because of there being more connector blocks...:-/

Additional Comment:

Well I've finally got my 2700k system up and running and I'm very very pleased with the results too.

My max stable overclock with intel burn test is 5.4GHz. That matched with my ram at 2133Mhz is pushing out a massive 128-130 Gigaflops......compared to 46 on my Q6600 at 3.7Ghz......that a massive increase.

I've disabled hyperthreading to get there....the burn test produces lower gigflops with hyperthreading for some reason anyway.

Also had to put the vcore to 1.51 to get stable which results in temps 60 - 70c!!!

I've saved those settings as an extreme bios profile and turned things down a little for a 5.0Ghz profile....1.38vcore and runs intelburn test at 30-40c and still produced 122 gigaflops. I don't want to burn out my new cpu...tbh....don't think there is much in the way of gaming that would be resrticted by a 2700k at 5.0ghz but I'll probably make a few intermediate profiles between 5.0 and 5.4 gig.....perhaps some with hyperthreading enabled also.


My board has got a facility to take screenies of the bios settings .....I'll have to post them up to see if I'm missing anything I should have altered.

PS....got it running at 5.5Ghz with hyperthreading stable in wprime 1024......from my results I'm in the top 50 worldwide for the 2700k.....just will not go passed the bios splash screen above 5.5 though....just freezes there and need cmos reset....went to 1.54vcore but still would not go....not sure how far you can push the vcore on these?

[Drewmeister]

Wow.. 5.5Ghz! That is just wicked TG!!! Congrats!

5Ghz at 1.38v is crazy too.. great job! Looks like you might have a golden chip there!

I've seen several ppl running up to 1.65v for benching on cool water.. some hwbot subs on air with 1.65-1.75V(not recommended lol). I don't think you need to try more than 1.6v with your cooling. Try booting at 5.5Ghz 1.55-1.6v and use software to take her up more in windows. If that doesn't work then either need colder or have reached limits of cpu. We can take a look at those bios settings too if you want.

It looks like you would annihilate just about all the air cooled benchmarks on the hwbot.. just not sure how fair that would be.

That's a huge improvement in temps with the new HS and reversing fans. Are you considering wc for the cpu?

I would love to see some bench screens.. also if you could test BF3 and let us know the improvement.

Fantastic job with all this TG!