Ultra quiet 'All in one' Quad compressor unit

[Gray Mole]

So after the promising results of the 'crazy small gpu cooler', I wanted to try out the whole 'all in one' setup. I haven't made one like this before, so wanted to approach it more like a prototype job than just a test bed. Not that I'm planning a production line (no way. too much work) just that I wanted to put it together like it could be.

Finally got the last of the compressors I needed. Ended up doing a quick test on a whole lot of them before I found an assortment that fit the bill. Had to be small enough to fit the compact design, large enough displacement for the job, and most important, quiet.

All different ones though. Wanted to see which would stay quiet under load, and if the slight difference in brand and CC's would make much of a difference, considering they're all around the same BTU rating, other than the larger displacement CPU compressor.

Specs so far

GPU compressors
Tecumseh 2.4cc low temp
Danfoss 2.6cc low temp
Huayi 2.9cc low temp

CPU compressor
Samsung 1/5hp low temp

Common condensor 12000btu/4
10 passes per gpu
12 passes for cpu
14" x 13" x 2.5"
Works out to the equiv. of 6x8" gpu and 8x8" cpu condensors


Custom case
12" x 15" x 20"
Central outlet for flex lines, will be designed for a PC case to go on top with room to go beside the lines with the PC side panel removed, or cut the PC case and have the lines inside.
I fabricated the case with metal I had here. Bit of a variety of pieces but it's all welded so nice and strong and ready for panels. Haven't decided completely yet, but I'm tempted to go with Perspex and have it a smokey or black finish.


Won't see the insides when it's done, but I painted the chassis for now. Nicer that way. It's all galvanised metal so it was awful to weld. Came out ok, and the ugly bits aren't visible anyway.



The case and compressors are a perfect fit. That is, all the room is taken up so no room for error. I'll have plenty for wiring but as far as temperature displays go I'm going to have to go flush mount. Only ones I can find are the awful PC case things :p but they're ok to -50 and they'll do for reference anyways. A real 4 temp display would be better but I can't find one that really suits.


The evaps are going to be a Chilly1 head for the CPU and a pair of @itor GPU evaps along with the modified Kayl GPU evap I made for that other test unit.



The ideal thing about this is that it's getting away from the multihead unit where you need to deal with balance and have a huge noisy compressor to go with it. A waterchiller was a thought, sure, but direct die is more ideal I think. Smaller compressors in multiples can do the same job while being quieter are more efficient.

I'm shooting for 200w per head on all 3 gpu, with the CPU evap tuned for 260 watts.

The temps I'm expecting on the GPU heads under load at 260w are similar to the test unit I built, around the -5c mark, but I think that with the temps you'll see on the GPU it will only OC to a level that would pull 200w.

I may just tune them a bit heavier, or right up to the 260w levelon the GPU's.

The CPU I want full coverage, and I'm hoping to see -20 or so at 260 to 300w. Haven't used a Samsung compressor before, but it's bone silent If it gives a decent level of cooling, and can remain silent at the higher load/backpressure, I'm a happy guy.

The condensor will be running either 9x120mm fans or 4x140mm fans. I haven't decided yet. Better condensor coverage with the 120's, and can run them slower for noise levels. 140's are naturally quieter but the sizing is wrong, and I have no room for shrouding so I need a full cover of fans for sure. We'll see how it works out. I still need to buy them, but will stick a barrel fan on for testing if I don't have them yet.

So far the CPU phase unit is together with the Chilly1 head. The 1/5hp compressor is set at 1-3psi unloaded, and with a quick ‘flame test’ pushes the low pressure up to around 18psi. Frosting just at the compressor inlet with no floodback. I’m pretty sure that it has the capacity for 260w with that kind of pressure response but with a smaller compressor Ican’t be sure until it’s on the tester, possibly later tonight, possibly tomorrow.


I have the GPU lines left to assemble but the short flex’s are done, just need the larger flex line installed.


the plan is to have the more flexible and smaller flex to the gpu's, with the heavier flex directly attached that can be shaped to the exact fit. Should be a good combination of flexiblity, size and vibration dampening.

That's as far as I've gotten. I'll update when I get more done, but it's pretty close and once the other lines are brazed I can finish up the insulation and get to load testing.

I did add one accumulator per GPU due to the smaller pipe size and short suction lines. I didn't add one to the CPU line due to size limits, but it's got 3/8 pipe along with a long 5/8" flex, so volume isn't the issue like on the GPU lines.



So here is the current setup. The two-part flex lines worked very well, but in future i think it would be much easier to just use a smaller flex on the GPU's in one piece. This setup is really cool though, allows for a more solid run to the card, then a really flexy line for the finer adjustments for install.

I added a basic barrel/cube fan for testing until I've got the flat axial fans here. Runs quiet enough though not what I want, around 200cfm at low speed.

They're all maxed out for load now, ready for load testing.


It's funny that the Tecumseh is the smallest of the compressors but really the noisiest of all 3 gpu compressors. Figures though, Tecumseh always seems to be the loudest of compressors in any size and capacity range.


The Danfoss is the quietest of course but the Huayi isn't that far behind, and it's good to see since in actual production it could be a very inexpensive alternative. Knowing it's also capable at higher loads, I think it's a winner all around.


The Samsung isn't bad for noise either considering it's higher capacity. Under load is where it's most meaningful. Looking forward to finding out how much the noise increases with load.


So just one initial load test done, and it's about what I expected.

Idle/unloaded, around -30c with 8psi low.


260w worked and sat around 40psi low.


The front temp is the evap itself, and the back one is the actual plate temperature of the load tester. About a 20-25 degree difference across the thermal paste, so heat transfer isn't too bad. At that heatload I expected something in that region.

That was on the larger of the GPU compressors, the Huayi.

Results on the CPU unit...




So -19 at 300w now that I have the tester to set it up

Should be fine for Quad core now, and the -20 range is great
Really quiet even with load on it, still under 20db.
Running about 18psi low at full load

I don't have any real information on how high GPU heatload will go, and I haven't seen much of anyone showing GPU temps vs evap temps on a single stage cooler. Would be great to compare this info with actual installs.

Bit tired tonight, so I'll get on it again tomorrow. Should get all the load testing done.

I'm really hoping to get what I need to convert this load tester to variable this week. I think that 200 to 225w would be ideal for capacity on these small coolers. I doubt that a GPU will OC to a point higher than that heatload at projected temperatures. Hopefully I can arrange a test myself, but I don't really have any decent gear to work with.

I may just leave it at 260w capacity though. I'm hoping to see the other circuits have similar results, though perhaps not as good with the smaller compressors.

I think it ends up being somewhat ideal though, for a person wanting 'day to day' usage with a SS cooler. The temps aren't so low that they require hardcore insulation, and under even a small load, they should be around ambient on the exposed part of the card's board. I'm thinking a pretty standard insulation job of the core area, like any cpu install, would be sufficient.

Comparing results against water, assuming the full 260w, means a minimum of 20 to 25 degrees colder gpu temps which I think would translate into much better OC results. I think the difference would be much higher but if you have a high end water setup, able to maintain a 25c block temperature even under full load, how much higher would the gpu temp be? 20c? 25c? If so, and the gpu temp would average more like 50c at best under load with a high OC and high Vcore. A quick look at someone's 'roundup' of blocks seems to indicate 50c gpu temp at best with an overclock on 480gtx but no indication if that's under load.

Would like to see how it compares that way to water, get an idea of how beneficial this kind of small yet high capacity setup would be.

So tomorrow, I'll start with the CPU head for load testing, since I'm pretty curious about that one with the 1/5hp compressor on it.


Gray

[Unseen]

Nice one Gray. Its good to see that you are building something different.

[Buckeye]

LOL Nice one Gray

Thats a lot of cooling power in that box. Only comment I can make is be sure to use long suction lines, which I am sure you will

Can't wait to see the rig this is going to be used for.

Nice to see something different being built also

[zalbard]

Definitely a nice looking unit (check out the smiley, also )!
You customer will be happy, I bet, the box appears to pack a lot of punch!

[Gray Mole]

Heh, no customer for this one

I'm just building it, more or less for the sake of building it. I'm kind of pretending that it's a layout for future builds, and I can already see where I would make changes, what kind of configurations would be easier for actual production, etc.

The most important change would be the case, really. Having the 'base' structure welded, but make the top level a bolt-on design to allow for easier 'warranty work' kind of thing. Beyond that, preinstalling the full suction into each compressor with accumulators would have been much better, but it's still ok how it is. The case is as small as I could get away with, so a PC case on top wouldn't look too silly I'll get a pic of a case on there for reference. In future, I think I'd go just a bit bigger all around, to make the brazing in that much easier. I'm also considering the 3 compressor format in a cpu/sli or 3 way sli variant, and go a bit longer but much shorter case, or possibly a taller but narrower case with the condensor on the end, and the compressors stacked 3 high. Having the welder makes it much easier to play around with case setups that are much stronger than wooden plates.

The design is actually quite a short suction setup which is why I added accumulators for the GPU lines. There wasn't enough volume to charge them heavily.

I added the largest accumulator to the shortest line, and up from there, since I only had three accumulators and they were from larger (from a small rotary) to smaller (from a VapoSE).

The CPU line has the larger and longer suction, and I had no space left up top so it didn't get one, but after seeing the response to load from the basic flame test, I'm pretty sure i'll have the capacity without an accumulator. I hope so, but I can easily pull the compressors and redo the lines until I'm happy.

Not having someone that it's being built for, means no timeline, no pressure, just keep at it until I know it would be ok for someone.

Oh, and I'm making it for a 'case' setup, that is it's a 24/7 format. Instead of a bench unit, more like the 'Vapochill case under case' configuration. Interesting thing anbout it is that a person with one, two or three GPU's could set it up, and just not turn on the unused lines. Sort of a 'future proof' system. The lines are small enough as well that they can lay pretty much out of the way. I'm assuming that a person wouldn't buy something like this unless they have at least crossfire or sli gpu's so only the third line may go unused/stored though.

Once it's in the format that i like best, it's just a 'modular' cooler, 4 independent cooling units, currently sized to allow for up to 1/5hp compressors max size. I think that a small reworking could allow for a 1/3hp compressor at most, but I don't want to end up with too gigantic a case.

Besides that, it's being worked into the 24/7, quiet and general use format, so having the max cold isn't so important as low noise. The cpu compressor would be nice to have larger, but if the smallest of the gpu compressors is still able to hold a good 200w + capacity and nice temps, but also not be too cold at idle, then the graphics cards can be insulated 'normally' instead of huge prep work for the end user.

Honestly, it would be way easier to make it a bencher with very long lines. Having to work out exact length, that can be flexible enough to use into a case is a pain. They're good now though, and have the kind of flexbility to install in any case, any motherboard.

I'll get some pics when I get dressed of how the lines worked out, they're still uninsulated and easy to see

You're right though, it really IS nice to build something different

My brother has a 2 card sli and quad core in his PC so I may see if he'll be a tester but I don't know.

Will get a few more pics shortly so you can see where I'm going with it but I really like how it's turning out

Gray

[dejanbinladen]

i like it

[Gray Mole]

So I did a bit of playing with it yesterday (bit of running around to do) but found a couple things I wasn't happy with.

Mainly the speed of cooling on the Tecumseh compressor GPU line. That and the slightly higher noise of that compressor makes me think I'll probably change that one out.

I've got a Daewoo compressor I didn't even know they made compressors. It was slightly louder than the others in static tests, but if it's going to be similar to the Tecumseh then I may as well use it.

The Daewoo is a 2.6cc compressor, slightly larger and matches the others better than this one.

I'll do the load testing first though, see where that stands, and then look at the changes.

I'm hoping to find some cheapish perspex today for a case. The fans will be mounted to the case panel, and then bolted through to the condensor frame.

Sure hope it works out as well as it does in my head though :p


Gray

[wdrzal]

hi gray, I noticed your "-1 Kelvin club" myself and I bet the entire worldwide scientific community would like to know how you achived this and what happens to matter ?????????

[Gray Mole]

hi gray, I noticed your "-1 Kelvin club" myself and I bet the entire worldwide scientific community would like to know how you achived this and what happens to matter ?????????

Of all the guys on XS I was kind of hoping you'd be the one to notice that

With so many guys tacking on their own -(insert the coldest unit you've ever built here) I thought I'd poke a little fun at it for the sake of it

Oh, and -1 Kelvin was the approximate temperature of my Ex-Wife's heart I think. Or was it the air in the path of her stern look? I can't remember for sure


Gray

[Gray Mole]

Updated the first post again with the first GPU load test

Worked as well as I'd hoped, will try to get the rest of the load testing done tomorrow and then finish off the 'pretty stuff' of the insulating.

After that the hard part is coming up with nice GPU mounting systems. I have a few ideas, hopefully one of them will work but an easy mounting system that looks nice is pretty important.

Gray

[[XC] gomeler]

heh.. negative kelvin temperatures are actually possible. Read about them a few weeks back. Pretty interesting and mind boggling ideas -> https://secure.wikimedia.org/wikiped...ve_temperature

[Gray Mole]

heh.. negative kelvin temperatures are actually possible. Read about them a few weeks back. Pretty interesting and mind boggling ideas -> https://secure.wikimedia.org/wikiped...ve_temperature

Well that just sucks, totally kills the joke



Now I have to come up with something else that's both witty and scientifically impossible.

[wdrzal]

heh.. negative kelvin temperatures are actually possible. Read about them a few weeks back. Pretty interesting and mind boggling ideas -> https://secure.wikimedia.org/wikiped...ve_temperature

Their only saying in a fancy way one system can have a lower entropy then another.

You can not achieve a temperature below absolute zero........@ least as our understanding of thermodynamics stands.

One more thing....don't believe everthing you read on Wiki.

[Gray Mole]

Meh, even if it's theoretically possible, according to some quack scientist's perception or translation of the facts at hand...

It still kills it. Someone might actually take me seriously, and that would never do

Wish I could be a quack scientist...

[Gray Mole]

Results on the CPU unit...




So -19 at 300w now that I have the tester to set it up

Should be fine for Quad core now, and the -20 range is great
Really quiet even with load on it, still under 20db.
Running about 18psi low at full load

Very happy with that result, wasn't expecting it to be that cold honestly, more like -5 to -10 at best.

Pasted that into the first post, keep it all together


Gray

[Gray Mole]

I was curious about the power consumption of this thing, so I plugged it into the power meter and got pretty much the result I figured.

I only have one at the moment, so could only plug them all in and run them without any load, but I figure the cooldown gives a bit of an indication of what they might pull with a load strapped on.

The initial startup was 700w, which dropped after a second to around 550w.

After full cooldown on all heads, the running amps were 480, and I assume the CPU or largest compressor is pulling a little more so I'd estimate it at 135w, while the rest are probably about 115w each.

I can only guess based on their ratings, but most likely it will be pulling closer to 600w at full load on all heads, or approximately 5 amps or so total draw.

Just for comparison, a 3/4hp unit I tested on the meter was pulling 4 amps at light (cooldown) operation, likely closer to 5 amps full load.

So it's actually pretty good, considering. Running four coolers for the same power use as one large cooler isn't bad at all, around the cost of 4 or 5 light bulbs (regular ones, not the CCFL's)

I'm really liking these little compressors Small ones are far more power efficient than the big rotaries, and it's nice to see they can still eat up the high load, even if it's not quite as cold


Gray

[Buckeye]

Not to bad Gray, nice setup

So you saying 700w on startup, then 480w after its running. Is that 480w unloaded wattage ?

When I did Tri-Fire runs on older 4890's with a 955 on air I was hitting 720w's on full load Vantage. Hope this link is ok http://www.buckeyeoc.com/?p=737

So that would put the whole system at 1200w to 1420w max
Thats a lot of juice LOL

Now picture newer CPU/GPU's and I think it might get a fair bit higher than that.

[Gray Mole]

Oh, no that's just the cooler itself. Not putting in any kind of PC numbers in there.

The cooler looks like a 480w total draw unloaded, and likely as much as 600w when the system is fully loaded.

These little compressors would probably explode if they were pulling THAT hard

But yeah, with a full PC running at 800w or so, and this one running 600w or so, you are looking at a total of 1400w with everything running at full load.

I can't be blamed for the hardware I just make the coolers.

Funny, this system will pull around 600w of power, and if you add up the cooling it'll actually do around 900 to 1000w of cooling.

That's something I love about small, efficient compressors. You actually get more than 100% efficiency out of them

If you look at some of the really small water cooler/heater models out there that are entirely compressor cooled and heated, you can see up to 500w of heating from a 100w compressor draw. That's 500% efficiency Gotta love refrigeration!


Gray

[Buckeye]

Yeah gray it gets pretty crazy when you think about how much wattage highly OC'd systems use.

The house I am in now I have np running a ton of stuff, but my older place I could not do what I do now as I would be poping fuses everytime I turned stuff on.

Wasn't complaining about your system, it's about what is expected with these setups.

I like it and that thing is a monster cooler !!

[[XC] gomeler]

Not to bad Gray, nice setup

So you saying 700w on startup, then 480w after its running. Is that 480w unloaded wattage ?

When I did Tri-Fire runs on older 4890's with a 955 on air I was hitting 720w's on full load Vantage. Hope this link is ok http://www.buckeyeoc.com/?p=737

So that would put the whole system at 1200w to 1420w max
Thats a lot of juice LOL

Now picture newer CPU/GPU's and I think it might get a fair bit higher than that.

It gets bad when you are utilizing all the circuits in your tiny apartment to keep from blowing breakers Cascade + water chiller + 5870 CF was brutal trying to figure out where to plug all the extension cords into.

[Buckeye]

It gets bad when you are utilizing all the circuits in your tiny apartment to keep from blowing breakers Cascade + water chiller + 5870 CF was brutal trying to figure out where to plug all the extension cords into.

Oh man I bet chris.

I am pretty lucky here at this house I am at now. I have ran all that in this one room with np. But when the Cascade kicked on the lights did do a quick flicker LOL.

[Gray Mole]

It gets bad when you are utilizing all the circuits in your tiny apartment to keep from blowing breakers Cascade + water chiller + 5870 CF was brutal trying to figure out where to plug all the extension cords into.

Heh

You know you're extreme when you have to ask the neighbors to turn off their lights so you can use your cascade

'Excuse me sir, would you mind if I plugged this into your socket? Mine are a bit full'



Oh wait, that came out totally wrong

[SexyMF]

That's something I love about small, efficient compressors. You actually get more than 100% efficiency out of them

Very pleasing statement.

[wdrzal]

^^^^^^^^^ Ya,did everyone forget the correct term?

COP = coefficient of performance

[[XC] gomeler]

^^^^^^^^^ Ya,did everyone forget the correct term?

COP = coefficient of performance

I only read every fifth word and try to make an educated guess as to what people are trying to say

This is shockingly true.. I skim a lot and missed that bit