Quad Head Dual Rotary Cooler Project

[Gray Mole]

So people say it's been quiet and I agree. There's also been a lack or particularly different custom stuff being made.

So here's something that hasn't been done, or at least not for a long while.

I'm converting a mach 2 case. I have a Lian Li kit for it, as well as front and top panels. So I'm chopping them up and making something pretty and functional.

He's got a pair of 6990's. They're a dual chip card, and it was originally just a dual head project, but with a dual card setup, that means 4 cores to cool

He really wants it pretty and with what I have it's looking that that'll work out well too.

So the project is a pair of dual head 1/2hp rotaries in an extended mach2 case with lian li faux front and top kit, so it's going on top of his case and the flexy's are going down.

Each head will likely be putting up with 300w at worst, depending on the OC. I've seen specs stating 300w stock per card, so I'm guessing this is 600w per unit max.

Most testing will be at around 240w per head though. Makes more sense to me to get a figure closer to the average-high range of OC. Won't be going too cold for the gfx so it's not a condensation/insulation headache, and it seems like around 50w idle per core, so 100w idle per core heat dump to the card means heater pads shouldn't be a requirement.

I'm aiming for around -10c per head idle temp evap. The core idle wattage + memory should make the card sit above where moisture is condensing, and the backplate foam will be there to help.

So specs...

1 x 1/2hp Rechi Rotary 31LRA
1 x 1/2hp Rechi Rotary 28LRA
(slight size difference, but I'll tune it so they give around the same temps. Higher spec on the primary card set)

2 x @itor gpu evap
2 x Chilly1 gpu evap
I'll likely use the better compressor on the Chilly1's. The @itors are a bit better, so will balance out I hope. A couple degrees difference will affect nothing.

2 x half condensors from a 12k btu 3 pass HE model. Real nice condensor, need to chop it and rebraze ends on the cuts. Paint afterwards.

12" tall, full Mach case width. 2 x 25mm fans per condensor for space.

It's a tight fit, and esp with the dual lines out, will split the lines for dual heads closer to the ends to save space and balance LP.

Small SLHX braze design. No room for real SLHX's but I can do the brazing to the Discharge post-filter/pre-captube or pre-filter/post-condensor. Prefer after the filter. May go dryerless on this one.

Gas is still to be determined. Will do a propane test and see what load = what temp per head so I know if I have to go down to 134a.

The rotaries are 11" tall. Case will be 13" tall. Space is non existant so it's all going to be very minimal.

Air in the back and front, out the top.

Window to be modded in, so I'll have to choose some paint schemes. He wants polished copper pipework though

Pics of the process so far, it's early days on it.











So I've been chopping the panels so I get a good front fascia match. Some paint to see what the final will look like, but it's just a mockup. Real paint later.

Need to cut down to a 13" front, and that should get that blank between the hole patterns pretty even, so less to make the seam obvious.

Still have to extend the case once I have the exact measure of the front.

3 of the heads are already prepped in the 'all in one' project from before which is cool, only the one to braze.

But I have to make some hold downs for them all. So I'm making up some ideas of how to do that easily and still have a nice hold down.

I'll keep posting as progress is made. I have a plan but I also have a lot of possibilities for colours, shapes, mesh types, etc. So many little things, the cooler is the easy part

Gray

[LRV]

Nice vid. To be continued?

[Buckeye]

So LRV in almost 3 years from your last post you end 2011 with this one ?

There is no vid in this thread

[Gray Mole]

So LRV in almost 3 years from your last post you end 2011 with this one ?

There is no vid in this thread

Was kind of wondering myself.

Back to work on it today, kinda crashed for a day or 2, long hours with no sleep has taken it's toll.

Hopefully some pics of a working cooler today sometime. I should at least have one of the dual head systems running.

Wish I didn't need sleep, things would go much faster that way.

Didn't make it in time to get the side panels on friday either, they closed at noon where I needed to go. Stupid holidays.

[MaRtIe]

another build mr mole

looking good. definitly different and gonna be interesting to see how the case turns out.

you going with lasercutting for panels?

[Gray Mole]

Yeah

I'm not that inclined to do custom stuff anymore, but this one's got some really interesting stuff in it, so I said I would.

Nope, no lasercutting here. I'm with Rakesh I figured I'd just chisel 'em out! (sorry Rakesh, kidding )

I've got a compound mitre that does really nice cuts with an Alu blade. For anything fancy I use the scroll saw. Just means grinding and polished by hand. I have die grinders and die polishers so it's a bit easier than hand filing.

He has a great laser etched panel he's going to install when it's there so I'll just do the side window for him, and prep it with some 3M tape.

But the toughest part IS the case. With a window it's got to look really nice inside. I'll have to fully polish and seal the copper tubing (no paint there unfortunately) and paint the compressors after I remove the stickers.

The condensors have to be painted though. Getting a perfect brazed finish isn't possible without a brazing chamber and I haven't set that up. I may sometime though. Not having to run nitrogen through a system, just into a chamber, would be really handy. When I get a better set of oxy/acetylene gear I might do that. I'll be looking at something very compact. I may look at oxy/mapp if it's a nicer fit.

The front panel (you saw that 2 piece front) is going to have a seam though. Not a lot I can do. I might look at a resin fill. I'd hate to mess up the look of it, but it's worth the risk. Once it's attached and strong it should hold the fill well enough to paint and get it smooth. We'll see.


Gray

[Gray Mole]

So this is what Willis was talkin' 'bout...




Making return bends and an endplate takes an hour or 2 if you're not rushing it, but it's a pain no matter what.

Cleaning the ends, bending on the vise, cleaning the condensor after cutting, it's all tedious but if you rush and miss anything you could end up with serious grief.

I did the harder one first, basically matching the second half's return bends. I'll braze it up and start working on the other one after I've brazed up set 1.

It's a pain but custom condensors are brutally expensive. Good reason to find a nice 'standard size' and just remake it for what you need.

This is a pic of the fins, and why I like this style of condensor. It has the riflebore tubing but the fins have venting for cross-fin airflow. So even without a real shroud, you still end up with full fin cooling.



I'll get a pic of the condensor after it's brazed and painted.

Gray

[runmc]

Looking good Grey. I love to sit down at the bench and cut/braze stuff up. If it goes right it's a treat. Time goes by soo fast.

[Gray Mole]

Yeah, it's a lot better when you're not trying to actually get anything done, though I wouldn't know what size I'd need so it's always when I'm trying to get something done

I ran out of black on the last project, so just for fun I sprayed it with some glossy chrome it hides nothing of the lumpy brazing but condensors aren't easy and I tend to heat until it's flowed, then cap it a bit just in case.

Makes for a not-so-pretty finish but the black keeps it out of sight, and even with the window the condensor isn't seen. If I had to show it and it really had to be pretty, I'd be ordering them.

But it's already a pretty expensive build with 4 heads and 2 compressors. Adding 200 bucks or so per custom condensor (short runs are expensive) just makes it way too much. But it's up to him, I can at least give a choice.

So a pic of it in shiney-shiney


Now if both sides looked that way, it would be tempting to use chrome

But...too rough for it.




The front is completely invisible in the case, with the cover on. The back has the fans and a secondary plate for 120mm fans so it can't be seen either.

Still going to comb out those fins though...

I pressure tested when I was done (last thing you want is paint when you have to rebraze ) and it's solid, so I'll be putting it together tonight, but it's been slow going today.

I tend to take a lot of time making condensors but it's easy to get leaks there, so you spend the time it takes.

Bit dizzy off that chrome paint though, it's nasty stuff.

Gray

[MaRtIe]

i admire you guys who cut these condensor's up and hand bend all the little elbows and braze them on. must take some time to do. do you use mapp gas or proper oxy setup?

what do you like to pressure test upto gray?

[Buckeye]

Nice work Gary !

I have often wondered how you guys go about brazing these up, don't you have to use a purge gas when doing this ?
How do you go about that with this ?

[Gray Mole]

If I'm using a condensor for a SS then it's only up to 200psi for leaks. Usually do a quick spray test, then let it sit for a couple hours. I've been tricked before, thinking after a spray it was fine/couldn't see anything, then it's down to 100psi in an hour and I'm hunting for what went wrong.

You definitely need to purge, I think it's the most important other than the evap to purge.

On a line if you do a quick 'in and out' braze you still get some crud in there, but it's minimal at least if you like to cut that corner. On an evap or condensor that's where the heat goes in or out. You foul up your heat exchange and you're going to be scratching your head later wondering what the hell went wrong.

So yeah, for normal use it's Nitrogen or Argon (no mixes with oxygen). If I have one that's been touched before and the insides are dirty from a bad brazing job, then I flush the propane through it, then burn off anything inside while propane purging. Grabs the carbon and burns it off. You can see when you purge propane, the flame (you have to burn the propane while you purge with it if you use that stuff) turns black when it's burning off carbon and it goes 'clean' again when it's gone or at least mostly gone.

That works for most thing that got carbon inside, you just have to get it red hot, but don't need to get it quite hot enough to melt the braze that's already there.

For me I'll usually do half the condensor with a chopped line (I have older access lines that are worn out and I use them to purge) then cool it, and do the same on the other half.

You can do it by attaching a large captube or 1/8" pipe to an access valve on the condensor outlet pipe, and braze that in for purging and do a cold towel wrap to protect your line too.

Loads of ways to do it I guess.

Oh, and I stopped using oxy/acetylene on small condensors. Too easy to blow a hole in the tubing they use. Not as thick as the coil tubing, made thinner for heat transfer.

Pain with Mapp though. I need a new Oxy/Mapp mixer.

Oh and I do all the return bends by hand on the vice. No easy way, you just clamp, bend, squeeze to round the pipe, and repeat until it's close to what you need. Thankfully it's a 1/4" discharge compressor, and the condensor is 5/16" so I could use 1/4" pipe to make my bends without affecting performance. I wouldn't make 3/8" bends. Probably wouldn't be able to stop them kinking.


Gray

[Gray Mole]

OK

So here's a mockup of the enclosures.

I saw pics of the 6990's and the regular enclosure size is too close to the plug at the back to use. I didn't want it to conflict since it's easier to make it smaller than needed than to find out it's wrong and have to redo an enclosure set, especially with there being 4 of them.

So I found some high density low expanding foam, that's what I'll be using inside. Some small dowels inside to set the height, then the foam will fix it permanently with the 2mm protrusion shown. I'd rather do a permanent enclosure with plate replacement than an install kit that's got to be assembled. Being round gives enough versatility that I'm not worried about future dual chip cards, and it's as small as it really can be done and still be insulated, short of welding thinner alu enclosures, and that's a lot more expensive.

The plate is just a quick mockup for what it's going to be, I'll be picking up some ALU for the panels, the fan transfer plates, and these hold downs and backing plates tomorrow.

The evap is set up for redoing the flex and captube entirely. I didn't like or trust the stuff that was there. No point in mistakes, the shipping's expensive for these things, hate to pay 3 times.







The outside will have a single wrap of armaflex as well as the back where the plate doesn't touch. Just in case, really.

I can't make the backplate permanent, since it's way better to 1. be able to rotate it and 2. so in future a new plate could be made should the need arise.

When I have the exact hole - to - hole dimensions I'll drill them into the new plate, and likely slot them a few mm so if there's a slight anomoly it'll still fit without any issues.

Outside diameter is 2 1/4", around 2.5" with the tape. Should be well within the limit. The spec on the foam I'm using sounds ok even with it thin, and the tape will take up the slack. The back has around 1/4" of insulation room too, so I'm less concerned with that than the sides.

Gray

[Gray Mole]

And now, the first lineset is set up, ready for final length of the drop (it's a top mount cooler in case that was missed) and the captube.

I'm doing the captube after this part, because it's so critical to get the heads aligned since they're on a card together. Can't make any mistakes or they won't sit straight. But they have a lot of play and sit pretty straight now naturally, so the mounts will pull them flat to the card together I didn't want to risk being distracted by the cap, and as well I'd rather have that area cool when I set up to braze the head. If it lets go it'll be a lot more painful to do a second time.

I'd forgotten how irritatingly precise this has to be. Been a while since the last multiple head project.





I'm thinking 7' of .031 per head for the wattage I'm after. I'd like 300w per head. The card's stock wattage is supposed to be 375w I'm told, so 600w total in case of overclock will do two things.

1. Ensure lots of overhead
And more importantly
2. Promote a balance between heads.

I always found that while matching the captube length and doing a split closer to the heads is a big part of the balancing act, having a high head pressure/upsizing capacity seems to also help with symetrical cooldown and head cooling performance, especially under load.

You'll notice I made my own 'T' out of a 90 bend. I didn't want a lot of extra solid in this line. Trying to keep the bulk out of it, as well as keep as much flex as I can. The first head has a very short flex for that fine tuning (could have made that solid really) and the other has all flex. I want that so there's room to use it on future cards. If they make the next dual 2" closer, no problem. There's at least 4-5" of travel on the second head, maybe more.

Gray

[n00b 0f l337]

Looks good. I almost would have considered a single mounting plate to hold both evaporators. Just a thought.

[Gray Mole]

I still could, but that's a risk I don't want to take.

If I had his PC here and could do all the measurements myself I could be sure.

But also, if the card's spec changes on the next one (I'm assuming 7990 will be dual again) then it's got to be separate mounts.

Don't know. If I can get a waterblock or stock cooler here or something to make a template I could do them as a single mount setup. Can always slice it in half later to use separately I guess.

But thanks. It's slow but I'm being more picky than I normally would be.

Besides. I can't easily buy a mount kit, prebrazed evap kit, etc etc. So everything is measured and made Too much like work. How did I get rooked into this again?

[Gray Mole]

Well a few more pics and progress for the thread.

Nice to see it coming along. Sometimes it's hard to face when you have so many things that have to be perfect.

I figured that while I'm focusing on the cooling portion I'll try not to think about the case. Compartmentalizing is always key in keeping a build moving forward.

Can't do that completely, since everything has to fit, but I marked the baseplate so I can be sure about the exit point, and I keep checking back on any measurements that might affect things.

So the line looks like this on the first one...


Cooler no. 1 looks like this in the case and the drop down...


And some frost




The bent tubing in the initial pic is to show how to set the drop. I like this tubing from Under-The-Ice (little plug there ) for being able to hold a shape on install and of course it's durable and easy to work with.

The line's too long on purpose, so if there's a different mobo used or the case ends up changing the height, it should cover some changes without an issue. Limits to that of course, make it too long and it's a mess and hard to get it to fit any case.

But the balance is phenomenal. Better than I'd hoped with almost no difference in cooldown speed or startup. I think I could feel the shorter head cool first but it was so fast I couldn't be 100% sure.

So in real use, it will be no issue in boot times. No waiting for an extra 30 seconds to be sure the heads are both cold. I like splitting the lines close to the head for that.

I'm hoping to see the ability to run the head with unbalanced load too. In previous systems and a lot of tests on prototype setups I've had no problem with around 150 to 200w difference using this kind of setup. I'm going to try a 250w test on these, running on unloaded, the other with that load, see if it can operate semi-indepedently if a single core is used on the card. I don't know if that happens on it, but I'd like it to be completely versitile that way.

The pressure is running around 2psi low, but it's just a fan trying to blow through, not a fan set like it'll have. I also haven't done the captube braze for it yet. Wanted to see it, check balance, and also the final leak test overnight. I did a 1hr vac (was purging the entire time so I doubt there'd be much air or moisture either way) and now I've pulled that and it's on a 250psi test. I don't want mistakes, but it's holding nicely so far

Oh yeah, the temp. I didn't worry about what temp it's at right now, more interested in the pressure. With it holding around the 2psi with marginal load from the air and fan, I'm expecting even lower with insulation and stuff.

I'm thinking it may be an idea to use 134a on this. Up to the guy, but if I go 404a it's going to be awfully cold at idle. I don't want condensation issues but still like it cold. So if it's running -20c or so idle with the wattage that a 6990 puts out (guessing 50-100 watts total per card/2 chips) then it shouldn't give any grief, especially with the card's components dumping heat into the pcb.


Gray

[rakesh_sharma23]

This is Cool...

[Gray Mole]

Thanks Rakesh

So I've been a bit slow the last few days, got myself a nasty cold and it's tough to do a lot.

Bit better today, and I've finally gotten back to it with purpose.

Not much to speak of yet, since I took pics of the last one and all I'm really doing is finishing up the second. But I did want to share something for those considering a dual head system.

You need balance on the captube, and that means being pretty precise on the length. I'm going with 7' of .031. The length isn't as important as each length being EXACTLY the same. Ideally, you braze your cap into the joiner (I'm using a short length of 1/4" pipe for the join) and then pull them tight, chopping them the same.

Now if you have accurate flow measuring tools then they're always the best, you can get a digital read on your flow and make sure it's the same. Not everyone has that. In those cases, nothing beats a good visual test, and propane is a great way to really see you have symetrical flow.

Just like you're hooking up for a purge, attach your line to the 1/4" pipe and flow just enough propane for a small but good visual flame for a check.



So it's not easy to see it in my terrible pic, but the flames are identical and that's what to look for. If you see any variation in flow, you're either got a slight length difference, or a bad cut on the captube or whatever and you've got to sort it out. Any difference in flow is only going to be worse in actual operation since the gas will flow to the easier path. There's always going to be variation in your return, but being shorter there will help.

If the cap flows identically to both then return variation will have a lot less effect.

Just my 2cents.

Gray

[SAE]

Very nice. Hope it performs well under load.

[Gray Mole]

The problem with this and most of the systems I put together these days, I already know it will. The limitations of each compressor will dictate it's actual temperature under load, but the evaps and captube setup allows for around a 350w max per head capacity. That's going mostly by the evap limitation.

The real challenge for the cooling system isn't there, other than to ensure precision balance in the system between heads so they'll cool equally and cool down evenly. That's where a system balance test is important, or you can end up with that issue when it's assembled and have to rework it.

I do miss the uncertainty that I had when I was doing it all for the first time. I posted the research units I was making, trying to build them to work with both symetrical and asymetrical loads. They were a lot of fun, and then I did feel like I was 'pioneering' in a way. Now it's more like 'just doing it' because it needs to be done.

Honestly, now it's more the challenge of making it 'fit' and look pretty when it's done, and that IS a challenge. After the cooling system test to make sure that the capacity and balance were there (extrapolated from pressure readings and temp testing) I've taken the first head setup off the cooler and finished the other one. They're both in the process of 'beautification', insulating and installing the probes and doing the sleeving.

On a dual head you've got to do that on the part that goes into the case before it's put back on the cooler, or you can't braid the longer part of the line (unless you buy the split braid, and I don't like the finish of it or the fact that over time it get's weak and starts to 'wing out' on the curves)

But thanks I'm getting it done, albeit slowly and I have to apologize for that. I get a bit dopey when I have a cold, and I'd hate to miss anything and have to redo things.

Gray

[MaRtIe]

cool trick to balance the cap's there. its obvious when you see it but not so obvious if you know what i mean. I would like to have a go at one of thses one day.
when you say a bad captube cut, i guess you mean when the tube sometimes folds on the end?


we should get you a donation for photography lessons

coming along nicely, looks a bugger to insulate

[Buckeye]

Gary do you run a load tester to actually get that 350watts of load ?

Thats a very high load number is why I am asking. As far as I haven seen the usual target number is 300watts and anything higher is a not only usually wasted effort but none have shown anything running with that high of a load.

[Gray Mole]

Oh no, when I refer to this one I mean it has a potential 350w per head maximum capacity. The captube and evap limitation is around that.

I'm charging it for a 300w per head capacity since with the captube mod it works amazingly well to have a slightly shorter than required captube length for capacity.

I use synthetic load testing for everything now. I have to rely on results I see to determine overall cpu requirements.

There is one chip that I and most are underestimating and that's the 12 core load of the 970+ intel chips.

I'm seeing a unit I tested at 400w drop-off just barely fail to keep the evap fully flooded under load. It's reading -94c (cascade) I'm told when loaded but the cpu temp is reading into the high 30s.

It's a 'top down' evap, captube is spraying at the top and coming out the column through the base.

I'm also told that measurements are showing close to 450w with the best estimations using Killawatts and OC idle/load differential.

That's a hell of a load, and I think that a change in evap design is about the only way to overcome it. Any evaps I've used aren't very efficient over 350w and while I've held 500w it's been with a huge temp delta base to back, as well as a terrible result base to resistor.

This card is rated around 375w total heatload from what I could find, though power wattage and heat wattage aren't always exactly correllated it's only safe to assume that it can be that high.

So giving a minimum of 500w is prudent, 600w means it's able to manage most overclocks possible without taking voltmods into account.

That would be 300w per head. So there's no real problem, other than the potential temperature of the system being relatively high. I'm estimating for -10 to -25c load temps per evap, but running a dual head means a shared capacity and that's a lot of load for the smaller compressor, so I can't guarantee the most extreme of cold temps.

Beyond that, a graphics card is a bit less easy to insulate, and there is a socket that can't easily be air sealed, so a warmer overall than it's sometimes nice to have is more ideal I think. I'd like to stay over -30c, and if I can get a good/low temp differential idle/load I'd be happier to see -20idle/-10load since the pcb temperature will be easier to manage and condensation shouldn't be the issue it would at lower temps.

Thanks Martin, and yes, it's a bit of a pain. Insulation isn't really that hard, but it's got to look nice and that's less easy. I'll get some pics of the insulated/sleeved lines so you can see what I mean.

The idea for the unit seems a bit like the cooling is nice and all, but it's more about having something that's a bit special, not common, and really nice to look at. More like a sweet watercooling setup than a serious benching machine. That means watching your total result, ensuring it's not too cold to make condensation a serious worry but having good temps to do a bit of oc play with. It also means that while it's a cooler and that's it's primary purpose in life, it's got to make the total PC package look better, not worse. Add to that the fact that the insides will be seen through a window, and you have a real project on your hands. Makes me tired every time I think about it. Maybe that's just the cold though

Gray

[Buckeye]

Sounds good Gary.

I have not heard of this synthetic load testing before.