Bitcoin Forum

Post your Klondike custom case designs here!

steamboat's 3U 72xK16 proof-of-concept:
https://i.imgur.com/1ohwFnC.png
https://i.imgur.com/erBoEu8.png

marto74's 8xK16 with and without space for a PSU and router:
http://s4.postimg.org/z670utkk9/8xk16.jpghttp://s22.postimg.org/z857k2cb1/8xk16inabox.jpg

I'd love to see a design for a tower-style 2-wide-by-16-high.

http://i230.photobucket.com/albums/ee106/PFC4L1FE/MOCK_zps134cc5d8.png (http://s230.photobucket.com/user/PFC4L1FE/media/MOCK_zps134cc5d8.png.html)

Trying to fit 512 into a rack mounted server... wish me luck.

Looks good.
Appears to be 2 or 3U.
Possible to fit it all in 1u?

I think the PSU is not gonna take full length space, if thats so then ample room for some embedded computer to control these suckers...

Or a USB hub and a Pi/BeagleBoard hanging between the cables in the rack. Lots of wasted space there :)

2U it has to be I think for 512 and possibly 768... 1U for 256. Hard to find PSU to get the +1200 W that fits 1U and be able to power boards and arranged in 40x40cm (even if 1 layer design) 2 layers of 4 x K64. Depending on the heat sink size it might be possible to jam 512 in a 1U and I am trying to get a venturi effect to push that air hard through the tunnel I am creating I think it might be possible. The left over space could house controller... depending on what is used in front of the PSU on the corner next to the front fan bank.

Not much head room on top there... not sure that you'd want to put anything on top wires ok but nothing else going to be pretty warm there. I was thinking that the top of the case and bottom will need venting slats.

Front View:

http://i230.photobucket.com/albums/ee106/PFC4L1FE/MOCK2_zpse1ac9203.png (http://s230.photobucket.com/user/PFC4L1FE/media/MOCK2_zpse1ac9203.png.html)

That would defeat the point of putting it in a rackmount case. You'd need to leave space between the servers to exhaust the hot air.

Why not put bigger / additional fans to circulate the air on top?

Or, turn the design upside down (hot air will migrate to the tunnel) and let the fans push the hot air out through the tunnel.

I'd like to play as well. Would anyone care to share a model file for a single Klondike 16 board?

Github?

https://github.com/bkkcoins/klondike

No not really since the tunnel is sandwiched between the boards. The top does not get air flow so definitely need top slats bottom gets airflow so it is fine without. The majority of the the heat is the chips so again sticking something near them on top etc not a good plan. The idea is to stack the boards so airflow can be achieved in the smaller space so I will adjust the stanchion heights and heat sink heights till I can get a balance. A 1.5U could accommodate 512 chips max height for fans though still has to be 40mm or 50mm fans.

I am saying this because you are pushing air under the boards. If you flip it upside down (just to be clear: the whole case, not the insides), the hot air under will naturally rise to the tunnel.  You could also use a 2U design with 80mm fans.

Thinking... if the goal is co-location, i dont think datacenters  allow 1kw for 1 or 2u rackspace... its just too dense...

Checking that out. There are plenty of GPU servers that throw that sort of wattage and more so not sure that is a hard and fast rule depending on the nature of the data center and what services they provide.  http://www.nvidia.com/object/tesla-servers.html (http://www.nvidia.com/object/tesla-servers.html) If I need to go to a 1U K256 so be it that is still doable I just do not want to have to build a server room on my 3rd floor so this is really an experiment to see if I can get my chips into a rack mount configuration. If it works I will definitely be doing more of them in the future.

As for flipping the boards they are fin to fin so that the heat is pushed out through the fins getting the best airflow from of the fans. Also the Klego is easier to put togther in a fin to fin configuration. Access however is not and maintenance and inspection will require a lot of time.  But given this a stanchion solution it is easy enough to test both configurations by changing stanchion length. I can test both ways and see which is better. I have a feeling fin to fin is optimal if the air flow is constant and in that Venturi type configuration.

Thanks

which 3d modeling program did you use?

The thing is typically a datacenter will bill you per rack unit, and a particular amount of power(and hence cooling expense) comes with it. Generally people that need more power buy more rack units than they need. Ive never investigated this point, but thats my understanding.

hmm... after writing the above i googled a little... it seems power is not an issue.

https://www.stackpop.com/search/colocation  hit search without entering anything, choose cabinet space 1U .. there is a provider offering 15A even...

Power IS a problem. Standard racks are 20A redundant (230V). I have 1U servers that need a half rack just for themselves, it's getting ridiculous. You can ask for more power but the prices are through the roof.

I'm seriously thinking of getting a DC in Texas or Washington (state).

Not flipping the boards. Take the whole box *as is* and turn it upside down. You don't change the current layout, just put it on its head. :)

I really don't think this is the "venturi" layout you're looking for. There is no significant reduction in cross section, just a hell of a lot of turbulent flow. I really think that even with a middle row of fans you'd struggle to remove that crazy amount of heat from there.

I am in Indonesia and I know this is a huge issue then again our problems here are air conditioning as ambient is 28C to 33C everyday, power outages and security. So for the extra electricity costs it might be well work a collocation set up. I am checking that out now and will definitely get back to you guys about it. The other issue for me is I don't want to have to deal with having it in my school over the longer term more of a space issue. I can put a few of these K256's together but I really don't want to house them all here... I suspect that if you have any number of these the cost of water blocks and maintenance might make it worth a look to collocate if the price is right. Always about the price.

Again you can do a K256 @ 512 W and get what 72.192 GH/s in a 1U... so that is certainly not going to be an issue. As a solution 1U makes sense if you can get PSU and a empty 1U barebones chassis at a good price. Must be plenty of chassis laying around somewhere at a great price. Group Buy? I am still going to try for a more dense K512 in a 1.5U / 2U and test it out at school.

The heat would be trapped then... again having the larger space above will allow heat to dissipate faster... heat rises. Flipping the design as it is would trap more heat I think. Again I can test all this out and run various configurations and see. I think the best solution is to get airflow into both the upper and lower cavity and push the air out and I think I might have to given the ambient temperatures here in Indonesia. I am not interested in air-conditioning the floor just for a single unit so I will have to be very concerned with airflow. The 3 - X6500 I have at home seem to do well although they run at between 40C - 52C all the time. I suspect with the right heat sink, fan configuration the K256 in a 1U should be more than adequate without AC. If have multiple units however then I think the collocation will have to be the option or build my own data center on the 3rd floor.

The Avalon boxes have a lot of air blowing at chips and board as well as fins in their vertical configuration. The poor goop job results in less efficient heat transfer as indicated in the thread of the Avalon user who pulled the heat sinks from the PCB. But it seems Avalon is really trying to tunnel the air into a very very tight space I just want to see if I can compact this down into something that is manageable size wise. I am really keen on what might be possible with a GEN2 Avalon chip so Blade Servers or SATA Hot swappable cards could be something that works as well with the preexisting server chassis. We will see what BKKCoins finds when he gets testing boards and heat sinks but I think the Klondike is easily configured so testing this will be quite easy. Personally seeing things like the TESLA 8 GPU server in a 3U/4U configuration and some of the 2 GPU Teslas crammed into a 1U gives me confidence that if you get the air flow and heat sinks right it is possible to get this sort of density.

15% Air Flow Top (Left)
-----------
70% Air Flow Middle Fin-2-Fin Tunnel
-----------
15% Air Flow Bottom (Right)

That way you only need the fan banks front and back. 16 Fans for 16 boards seems right as that is what people will do in stack configuration anyhow. It is such an amazing design that BKKCoins has come up with. Even if you are not going to do a lot of DIY board building just configuring cases etc is going to be great fun and I bet GEN2 Klondikes will be even more versatile.

Isolated view of the low profile fin-2-fin tunnel concept for a K32 section:

http://i230.photobucket.com/albums/ee106/PFC4L1FE/K32VH_zpsf1606292.png (http://s230.photobucket.com/user/PFC4L1FE/media/K32VH_zpsf1606292.png.html)

It is smooth not very turbulent through the fins it is basically less turbulent than the Avalon design given the proximate distance to the fins and as the fans are face mounted this will provide a much more stable and direct airflow across the fins even without baffles. Unless you have a simulator shows airflow I suspect you are speculating just as much as I am?

Testing this will not be an issue since the modular design of the boards means break down and re-configuring is simple. So I will be testing it with and without baffles to see what happens and get temp readings. Having a larger opening (volume) at the front and smaller space (volume) at the back along with having 2 banks of fans I don't think there will be any issues given this is based right off other server designs with a lot more obstruction of the airflow. I don't see this being an issue at all but again I will test it out and provide feed back. If it works I'll throw up the SketchUp plans so people can modify them for KICAD or what not be keen to see others work on this as well.

SketchUp 8. I will post up some models in a few days when I get time.

http://www.sketchup.com/ (http://www.sketchup.com/)
http://sketchup.google.com/3dwarehouse/ (http://sketchup.google.com/3dwarehouse/)

You guys have some great case ideas! I have several hundred chips on order which are destined to be K16's and I can't wait to pick a beautiful home for them :)

Hi ,
Just small remark
I do not see the power connector in your 3D

Ya I didn't add that to the model... I was using an older BKKCoin design... need to add that for the next one. Also parts are not to scale just eyeballed except the QFN48 size. Just wanted something that I can mock up fast not too detailed. But yes need that has to be added just so that cabling can be assessed. Problem with the design as well boards are not oriented to the KLego design that BKKCoins specified. Just trying to fit the boards into the space and orient the heat sinks along the axis flow of the air.

Re: the rackmount chassis -- excellent design, great work. How long is that design? It'd be nice if there was some extra room for some sort of controller -- maybe we could cram a raspberry pi in there?

I think you're going to have too much heat near the rear of that... Maybe a second row of fans in the middle to help move the air?
I think the higher velocity fans are about twice as thick as those too...

That 512 chip monster would be pulling more than 1kW. That's a lot of heat. Yes, the rear will be very hot.

I'm working on my own design for four K16 boards in Sketchup now, will post it later tonight.

I am working on the idea of a 1U for 256+ chips the 512 would need to be at least 2U to fit 50mm fans. Search online for the specs as there are number of different configurations.

As for heat there are two banks of fans push and pull so there won't be too much of an issue with that given some of the server designs I have seen that utilize this same design even with more restricted airflow and generate as much or more heat given the PSUs they use as I stated above the GPU servers are much hotter. Have to remember the design is spread out over a much wider surface area and fans blowing over the entire area so heat should not be such an issue given the CFM is proabably more than sufficient. Anyhow having looked at various cases the max in a 1U could be a K256 with only 512W so no heat issue then.

I will post some new mock ups for that tomorrow.

Need something to fit in the case?  http://cubieboard.org/

beagleboard? coinninja looks to have a nice anubis + cgminer cum mobile phone distro.

Here's an idea I had. It's not really feasible since the heat sinks aren't readily available and would be crazy stupid to custom make. It was fun to put it together in Sketchup though.

What if you could make a heatsink like this?
https://i.imgur.com/UGjnkUwl.png (http://imgur.com/UGjnkUw)

They would fit together with the Klondike 16 board on the back, like this:
https://i.imgur.com/nALH704l.png (http://imgur.com/nALH704)
Notice how the pin placement is more dense towards one end of the heat sink.

When you put four together, they create a tunnel:
https://i.imgur.com/2p7XjEXl.png (http://imgur.com/2p7XjEX)

Less dense end. This is where you blow air in:
https://i.imgur.com/KIx6WYol.png (http://imgur.com/KIx6WYo)

Dense end. This is where you want the air to come out:
https://i.imgur.com/oVCXJvCl.png (http://imgur.com/oVCXJvC)

Now, since there are more pins towards one end of the tunnel, more heat will be given off in the denser end. Thus, the board temperature is the same in both ends.

So, make a case of plexiglass boards and some screws/rods:
https://i.imgur.com/BdTm78Ol.png (http://imgur.com/BdTm78O)

Put some fans and feet on, and you are good to go:
https://i.imgur.com/p3iOxVbl.png (http://imgur.com/p3iOxVb)

Those are two 140mm fans, which will ensure good airflow both inside the heat sink tunnel and over the top of the boards. Air is sucked in at the bottom, and pushed out on top, working with the natural flow of hot air.

Voilla! The K64 cube. ;)

So, the heatsink I made up for this design is, as I said, not feasible. However, it would be possible to use the same tunnel design with a normal heatsink - but the chips towards the top of the cube would get a little warmer than the chips on the bottom.

The cubes could be arranged in a square at least 2x2 for a 256W space heater.

Nice! That's genius :)

Revised 1U Rack Mount

http://i230.photobucket.com/albums/ee106/PFC4L1FE/1U_zps1f3bdb92.png (http://s230.photobucket.com/user/PFC4L1FE/media/1U_zps1f3bdb92.png.html)


Love the K64 CUBE... stacking them and placing them in a room you can save on heating costs or use them in greenhouses. I am imagining them in cylinders.
Those heat sinks might be very expensive but in a cold climate you could certainly use that extra heat. Might want to see if there are some off the shelf heat sinks that can do the same thing or can be extruded. Change the angle of the heat sinks and get penta, hex, hepta and octogonal shapes.

You can cut existing heat sinks on a bevel... and I bet you can even have that design in an extruded version without having to cut it. It definitely would not be as pretty but it could be done and it would look a bit more utilitarian but would work very well.

http://i230.photobucket.com/albums/ee106/PFC4L1FE/bevelheatsink_zpsb3ead942.png (http://s230.photobucket.com/user/PFC4L1FE/media/bevelheatsink_zpsb3ead942.png.html)

http://www.supermicro.com.tw/products/system/1u/6015/sys-6015x-8.cfm  <---- Casing for the 1U design. http://www.mkl.co.id/ <--- can get them in Indonesia here.

So can a Raspberry Pi, Arduino or BeagleBone run this server or the K64 Cube can you put it somewhere in your modular design?

An Arduino won't cut it, at least not a regular Uno or Due. There is an Arduino board that also has a Linux PC on board that might work, but a regular mini PC like a Raspberry Pi is way cheaper than that board. The cheapest mining-capable embedded Linux system that I know of is the TP-LINK TL-WR703N router. It can be had for ~30 EUR. It's used in the Avalon miners. (https://en.bitcoin.it/wiki/Avalon#The_TP-LINK_WR703N)

So Tosku... with your K64 Cube... where does the controller go? Interested to see that fit in the design and power supply.

It doesn't go anywhere. I designed the case thinking the controller and PSU would be modular, external units. If you wanted to, you could fasten a simple controller on the side of the cube.

http://www.curtpalme.com/forum/files/aleph_5_191.jpg

There are definitely extrusions for your K64 Cube! This one is inside out Tosku.


https://i.imgur.com/ZpItyB0.jpg

Might be good to consolidate all our various extra Klondike threads to one.

1. Casings
2. Controllers
3. PSUs and cabling
4. Firmware  / Mining Software
5. Miscellaneous
6. DIY Build information
7. Pricing and sourcing of PCB, BOM, etc.
8. Heat sinks and cooling

Google tells me
http://safebrowsing.clients.google.com/safebrowsing/diagnostic?site=http%3A%2F%2Fwww.coolinnovations.com%2Fincludes%2Fimages%2Fcomponents%2Fslideshow%2Fthumbs%2FFlared_III_Large.jpg&client=googlechrome&hl=en-US (http://safebrowsing.clients.google.com/safebrowsing/diagnostic?site=http%3A%2F%2Fwww.coolinnovations.com%2Fincludes%2Fimages%2Fcomponents%2Fslideshow%2Fthumbs%2FFlared_III_Large.jpg&client=googlechrome&hl=en-US)

I think better to remove link to them.

Yikes, can you imagine the noise in that thread with all the different topics. Methinks this might/should turn into its own subforum.

No not really... better to have slow and fast together... one repository. Harder to follow all these different threads especially with some of these get buried and lost especially with all the hate mail BFL threads messing up this subforum.

I'd prefer we house all the best information for the Klondike build in one thread.

I like this,
1 mod I would suggest is make the surface area of each pin equal.
That would mean that the shorter pins were progressively fatter.
That may give a more even heat draw from the board, also more even heat dissipation.

cheers,
Kev

+1 this is better than having that in the main thread...  but "4. Firmware  / Mining Software" would probably belong to main thread... dunno...

another idea could be spin off per topic threads, and getting bkk to list them in the OP of the main thread.  OR make a new master thread started by someone quite invested in Klondike (im looking at you Bicknellski  ::) ) . Where the first post can be summary of options, etc as the discussions continue. That way a new comer can simply read one place with links to individual threads if needed.

Having everything in one place has some drawbacks also. I may not be interested in all topics listed. For example im following the case topic simply out of curiosity and will not be interested if it gets information overload. I will most likely not use a case.  Im not worried much about topics sinking deep since i "watch" posts...

And if you're following more than one topic, the read becomes difficult. You need to jump posts to read on one topic then backtrack wherever the other one starts, again jump other posts, then jump back again for 3rd topic, etc. There are already 8 topics in the short list. This will get confusing really fast.

How about making 8 "official" threads instead? (I would still prefer a subforum, this is tuning into its own little community)

Made a Reference thread for DIY Klondike peoples.

I can compile all the important info and people can discuss how they want below. I will update often.

See now, that's cheating. You're making a subforum in a thread. Sheeshh. Next you'll claim it was your idea from the start ;)

You made me look.... damn YOU!

Back to the case design thread....

Anyone got any other off the shelf cases?

I got over my Apple obsession years ago, but I still like to keep up on their new stuff. You seen the new MacPro (http://www.apple.com/mac-pro/)? The round one? It's new "Unified Thermal Core" uses a triangular heatsink that splits the heat between the CPUs and GPUs, and blows air up thru the center. Very similar to the idea you have here.

http://i594.photobucket.com/albums/tt24/EugW/Computer/Apple/MacPro/MacPro-ThermalCore1_zpsee1fb7ad.jpg

I would consider this worth looking into more.

Not been stalking for a while. I'm not going to simulate this in CFD, because there are way too many variables to make the simulations tell us ANYTHING useful.

'Turbulence' essentially = Reynolds number, and is a function of length travelling over a surface. For heatsinks this large/long/arranged in a pattern, you're well into turbulent flow at any sensible air flow. You're going to need a hell of a lot of pressure to maintain the airflow you need to maintain the heat transfer you need, especially as the effective ambient will probably be 50C for the 2nd+ bank. Heat transfer is a function of temp difference, so each time you try and remove heat with the same air, it gets harder and harder.

The difference to commercial servers is:
1) Typically use centrifugal blowers, much higher air pressure [required to overcome frictional losses]
2) Are VERY VERY VERY VERY noisy, because the airflows required are insane.

Just look at a 1U heatsink fan on a typical Xeon. Even if the CPU is only 90W, you'll have a 45dba centrifugal heatsink for that ONE cpu, using a solid copper heatsink.

1U sounds noiser than vaccum cleaner when booting up or full load... Got one to office once to debug. Everyone looked up shocked when i turned the sucker on...

What is the difference between the heat sinks proposed and those?

Not going to care much about the noise... it will be 3rd floor till I can find a permanent home. Given the talks I have had with a local GPU farmer we are probably going to go into together and rent out space and create a farm center. Again I will test it out and see what we get and I am toying with the idea of having a local aluminum shop fabricate a my chassis as it would be cheaper. Again given the constraints and the heat once you get enough of these in one room you have to go to liquid cooling of some sort. At that point I think mineral oil is the best method and commercial cooling / chiller tower set up on the roof.

I will keep you posted on the build. I am also going to build the K64 Cube or maybe even a K96 Hexagon.

Really love that triangle based prism concept as well.

Yep. And thats for what, 100-200W? Imagine the density of some of the designs we've seen here.

Not all 1U servers are 100W... or 200W... come on. Average what 200 to 400W.

And I guess I shouldn't have bogarted the thread with a 1U server... we are supposed to be talking K16 Casing.

i believe when booting up all fans, etc are set to max rpm... before the controllers/sensors come online... but still noisy in moderate usage.

You have 2 kinds of people. Ones ( like Bicknellski) who plan on putting it in dedicated room where people wont be working, and dont mind the noise.... and people (like me) who would probably have one running in my bedroom or office and a vaccum cleaner would drive me nuts, but at the same time dont mind leaving it left caseless and MacGyvering external airflow...

Without any knowledge of thermodynamics... id guestimate atleast 2 120mm fans worth of airflow to remove heat from 4 K16 (128W)... with ambient < 25C

You can't run this 1U in your room... lol! Mine is going on the 3rd floor of my shop house school well away from the working floor of the school.

No no no.

I think you would have to stack like cairnsmore / x6500. Even then the heat might be a bit much. My ambient is 28 / 32 night / day.

We're guessing at what he just dragged into his office. I doubt he has a super dense 400W 1U server in a work space lol. Have fun trying to use a phone.

My comments were actually made in regard to http://i230.photobucket.com/albums/ee106/PFC4L1FE/MOCK_zps134cc5d8.png

Nah it wasnt super dense... I still have that box collecting dust for sentimental reasons (decomissioned years ago and i cant use it for fun cause noisy). I will look up PSU capacity and let u know. I believe its one of the early xeons with hyperthreading, and the board fits 2 cpus, but has only 1 installed.

I can bet anyone 1 satoshi that no datacenter is gonna accept a unit so dense. Personally i think its a little dangerous.

There are ones more dense than this in centers already. GPU servers and what wattage does a 4U server throw... density? I am thinking with the 55nm Avalon 2nd Generation this will be possible. That is what I am counting on but I want to do a 1U now to see how it all fits together. 3 Months till the next chip! We need to get back on thread though... K16 casings.

The rack in my shop begs to differ, and begs to hold as many of those as it can.

Which? 1U server idea? K256? 700W?

The DC's I have contacted recently will let you have a 2200W server per half rack.

edit: 2200W usual, 3500W peak

Depending on the data center, you'll find most rack densities in the 3-7kW per rack range. At 42u, that's 71-167w per U on average. You can get more dense if you want, but you're going to pay for the power / cooling that you use and it won't do you any good to make it more dense. Ultimately, the colo provider had to build their data center with a design power in mind and if you chew up power before the usable space, they can't sell just bare white space without the power / cooling capacity to support it. You really can't look at it in terms of u-space or even physical (white) space alone. Stick to the kW / rack or even watt-per-U. 100w per U is a pretty good number since it would give you 4.2kW / rack. if it was fully populated. Go higher and you're probably going to be using blanking plates.

Source: I'm an 18-year "critical power industry" veteran.

I can't remember how many W each K16 uses. How many K16s would 100W limit you to?

32W per K16 estimated

3x32W=96W
4x32W=128W

not too much eh..

https://www.stackpop.com/configure/colocation/singapore/singapore/singapore/19885
4 kva for 42U $2500 / month
125 x K16 = 4 kva
125 x K16 = ~3 K16 per 1U  (most likely there are some unusable slots, use for switches, spacing, etc)

https://www.stackpop.com/configure/colocation/united_states/virginia/ashburn/17237
24 kva 59U for $1000 per month (or $2000 unsure if power is included or extra)
750 x k16 = 24kva
750 x k16 = 12.7 k16 per 1U or ~51 per 4U  (most likely there are some unusable slots, use for switches, spacing, etc)

But these can only be used once stability of these devices is proven, to not need much babysitting by hand.

Cramming higher density into 1U is only viable if you are hosting the rack on your own premises and can arrange adequate cooling.. But then why would one use 1U and not bigger cases where u can have better airflow using more efficient bigger fans.

For something of the 1U I designed to go into a datacenter you will need 2nd Gen or 3rd Gen ASICs where power is a third it is now. Maybe the 55nm Avalons will work. Then it will be datacenter ready. That is why I will have to go cheaper route build my own data center? LOL... sounds funny to say but it will be.

This is a brilliant design! I especially like the idea since many cases already have high airflow designs for hard drive bays and extra fan mounts available stock.

So each 1U server would literally be housing 3-4 tiny little boards for only 15-20GH/s tops. That's not a whole lot, especially when you consider the cost of the 1U server chassis, the PSU, controller, etc.

I'd rather go with a 3or4U case that can handle a decent ATX PSU, and throw a whole lot more of those K16s in there, maybe sandwiched together into pairs with fans between them?

That would certainly be more cost effective.  I've seen one proposal that tries to fit 72 K16's in a single 3U case... at that density you'd run out of power after only 6U of the rack filled.  I think that 2.5KW in a single 3U case is way overboard, but even if you cut that in half to 36 K16's, you'd need just over 16KW for the whole rack, and that is unheard of.  It would be 2TH in a single rack though, which is pretty sweet.

It really depends on how much you want to cram into a rack. If you're just looking for a single, really dense unit, sure you could cram them in there until they don't fit anymore. At that point, physical placement is your bigger concern. If you're talking about populating a rull rack, power distribution becomes your limiting factor.

20 of them in a 4U chassis would be about 700W (including fans/controller), which is totally doable from a thermal perspective. A 4U chassis could vent a lot more than 700W of heat, esp if you use a lot of fans on the front and back in a push/pull. 10 of those in a rack would be about 1TH/s, at ~7KW. That would be manageable, I think.

Had a random thought.

This is only for on premise hosting... not datacenter.

If we use such dense layouts, airflow is clearly channeled. All of the cool air would be taken in from one side and hot air coming out the other end. Why not seal the outlets, attach ducts, and release the hot air to outside....  I guess the devices should run fine with 25 - 30C ambient in tropics. One would need air conditioners because the hot air released increases the ambient much higher, but that wouldnt happen if you channel the heat out... The hot air should be 5 - 10C warmer than ambient i presume...

I see hosting in datacenter or controling airflow the only reasons for even needing a case.. if no airflow control is needed just keep naked pcbs...

I drew this little diagram in another thread, but thought it might be applicable here.

If you're not in a datacenter, then sealing everything up and channeling your airflow is the best way to go. For proper cooling, you need a low ambient temp. If the miner is dumping 1000+W into your spare bedroom, then the ambient is going to skyrocket. You could throw an AC into the room to combat the increase in temp, or you could throw all the heat out the window, literally, and just keep a fresh supply of cooler air coming in.

This thread (https://bitcointalk.org/index.php?topic=200306.0;all) has some good ideas for properly insulating and exhausting your rigs. At that point, you don't need to worry about giant AC units, you would just need a steady stream of cool, outside air. I live in the NE USA, and even in the summer, it barely gets 25C, and hardly ever to 30C, even in the summer. In the winter, it's not uncommon to see below 0C. Fresh, cool air isn't a huge concern for me. ;)

      COLD             HOT
       AIR               AIR
        |                      ^
          V                      |          OUTSIDE
----FAN-----------FAN---------------
        |                      ^           INSIDE 
        V                      |
         --> ASICs -->

I guess in winter you have free heating by removing the ducts...

Here the ambient can go as high as 35C!

The question is... if all the heat is thrown out the window, would 35C peak ambient need to be cooled using AC? If hot air is thrown out, naturally the room would suck in fresh ambient air from other sources to recoup maintain the pressure...

While this seems good in principle, there are some things to keep in mind when using outside air for cooling.

1) Just exhausting air to the outside and using the room as the supply means that the negative pressure you're creating draws the air from somewhere else in the house / building. Exhausting the air out your window means the air to make up for what you're exhausting is going to be coming into the house from somewhere else. You may not feel it but it's happening.
2) Using an inlet and an outlet will remove this but then you have to worry about air quality, humidity, and even potentially rain. If you have a long enough duct or have one of those dryer vent covers, you'll avoid rain but humidity (specifically condensation) is a more difficult challenge. The last thing you want to have happen is for the temperature to fall below dewpoint and you start getting condensation. Probably not going to be a huge issue since the temperature change across the device is going to be positive therefore producing exhaust air that is capable of holding more moisture than the intake air; but if it's enclosed and you're taking in cold air in the winter, it could potentially cause condensation on the exterior that could mess some things up. Additionally there's debris to worry about like dust, pollen, or whatever else. This can accumulate in the heat sinks or other airflow areas and either block airflow, foul heat transfer surfaces (meaning the internal temperature rises), or even accumulate and pose a fire risk.
3) In addition to humidity, electronics may experience damage or small cracks due to excessive rate of change of temperature (called heatup and cooldown rates). ASHRAE TC 9.9 (http://www.eni.com/green-data-center/it_IT/static/pdf/ASHRAE_1.pdf) (This is the 2011 version which was recently superseded by the 2012 version, though it's not publicly available from what I could find) is the industry standard for this type of thing. This is all to prolong life and prevent damage or danger to the environment. The point here is that if you have this running in the winter (presumably powered by an external fan) and the hash rate goes down (pool goes down, lose connection to internet, or any number of other reasons), if you continue to pump cold air through there, that could stress sensitive components and ultimately lead to premature failure.

There are some other considerations but these are the top things to keep in mind. It's not as simple as just rejecting heat to the outside. It can be done but if it was simply that easy, every data center would be doing it.

That is what i was counting on.. the house/building temperature is not that warm, close to ambient.
It doesnt get cold, and never ever close to dewpoint, but the problem is humidity and/or pollution/dust. I feel using internal building air should be fine, since its not directly coming from outside...
Interesting. Will go thru the doc after a while. Hadnt thought about heat-cool cycle causing problems.

In fact i thought about this because I remembered hearing about someone (it was either amazon or facebook unsure) throwing away the air from the hot isle, and using fresh air for further cooling and using. They obviously have some expensive equipment in place to treat the air before unleashing it onto the servers....

Oh for sure, they do these kinds of things and they can be done but they are in situations where they're basically using commodity hardware (for them) and have environments where the failure of a machine, rack, row, or even in extreme cases entire data centers doesn't translate to lost service to customers. They can afford to be pioneers and push the envelopes. Most of us simply don't have this type of ability in our houses. They save money on power / cooling equipment and just accept a potentially higher MTBF of server equipment. I worked at a "big company's" data center in Seattle for a while where they had a tent outside that was basically protecting servers from the rain. Intel did similar things. These types of endeavors are what have helped open up the TC 9.9 operating window. It used to be a max of 68F back in 2008 (iirc) but it has since been raised to 80.6F for the cold aisle (inlet) and even in some classes of machines as high as 104F.

I've been thinking some more about the K64 cube heat sink. The one I originally used is very complicated. Here's a model using a 90x100mm heatsink, similar to off the shelf-sinks available today:

https://i.imgur.com/oJFoWyT.png?1 (http://imgur.com/oJFoWyT)

The heat sink I made up only has fins over the actual Avalon IC:s. Such a heat sink could be cheaper, but the airflow between fins would be decreased by the large opening in the middle. Maybe it's better to use a heat sink with fins everywhere?

Here's a model using a "perfect" heat sink I made up:

https://i.imgur.com/8JOPEHV.png?1 (http://imgur.com/8JOPEHV)

The idea is to increase the heat exchange in the far end of the tunnel, which would be pointing upwards when in use. Unfortunately, again, a heat sink like that would be complicated and expensive to make. It looks darn slick, though :)

~800 W
That appears to be 3U half rack unit. or is it 4U?

I think needs some space for
1) Controler - raspberry pi
2) USB hub(s)
3) PSU

So, stick in 2 of them in single 3U (full length) case, remove about 2 or 4 K16, stuck in PSU, hubs, rpi. Add in external power/network ports. Then find datacenter that'll take 1.5 KW in 3 RU...  then profit.

It's a 3U case. It'll have the PSU and controller (an Ubuntu box) separately as I won't be using hosted rackspace. Otherwise, yeah you could sacrifice some K16 space for the things you mentioned to create a self-contained unit.

I think we should post designes of cases, but builded with materials that many of us allready have them.. or they are easy to find and to buy... not fancy CNC drilled cases made in a fancy 3D software

I think we should post designs that we are building for ourselves in hopes that it may help anyone else irrespective of that they use to design it or what materials it uses.

I think you can think whatever you want - regardless if you know what you're talking about or not.

The case I posted is readily available for internet order and would only require knocking a few holes in metal plate for fan cutouts - which could be done with a jigsaw in a pinch.

I thought it might be useful for others considering similar designs, sorry you don't find it useful.

yes indeed the design is verry OK... but my point was a little bit of something different  :)

I think one challenge in your deployment will be mounting. You would probably need a lot of hex spacers of exact specific lengths, and then drill holes in in the case and mount off that asif its 3 pillars.

OR mount them off the big side (which in the pic is not shown), but its going to be tricky IMHO.

Whats your plan regarding mounting?

https://www.stackpop.com/configure/colocation/singapore/singapore/singapore/19885
4 kva for 42U $2500 / month
125 x K16 = 4 kva
125 x K16 = ~3 K16 per 1U  (most likely there are some unusable slots, use for switches, spacing, etc)

https://www.stackpop.com/configure/colocation/united_states/virginia/ashburn/17237
24 kva 59U for $1000 per month (or $2000 unsure if power is included or extra)
750 x k16 = 24kva
750 x k16 = 12.7 k16 per 1U or ~51 per 4U  (most likely there are some unusable slots, use for switches, spacing, etc)

But these can only be used once stability of these devices is proven, to not need much babysitting by hand.

Cramming higher density into 1U is only viable if you are hosting the rack on your own premises and can arrange adequate cooling.. But then why would one use 1U and not bigger cases where u can have better airflow using more efficient bigger fans.
[/quote]



come and host them in ROMANIA-EU

you will pay 125x16K=4kva= 503$ / month

what's the DC in Romania?

i presume you just calculated electricity cost only... not factoring in cooling or rackspace or other facility charges.

Sheet metal frames to lock in columns of three K16s, as well as to channel airflow - essentially creating a "blade" design. Frames will likely be screwed into top/bottom plates. I'll post pictures once I'm finished the CAD work.

503$ only electricity 4kva

but in near future I will have a temperature control cooled room and I will open a hosting service... ofcourse not for 503$/month...but it will be an verry atractive price considering that I will provide UPS, chilled room, internet backup, and allmost 24/7 man if something needs to be done!, remote control of your device!

The links I posted are about proper data centers, one in our own cities which we trust (they handle much more expensive equipment). one which we can visit. One that has high gurantees and SLAs... And offers good physical security.

Personally I would host the mining equipment in a different city than one im based in...

k16 is too big (100x100) and heatsink is max 50x50:
http://www.fischerelektronik.de/pim42/upload/fischerData/image/web/lam5.jpg
http://www.fischerelektronik.de/web_fischer/de_DE/K%C3%BChlk%C3%B6rper/D02/Miniaturl%C3%BCfteraggregate/PR/LAM5_/index.xhtml (http://www.fischerelektronik.de/web_fischer/de_DE/K%C3%BChlk%C3%B6rper/D02/Miniaturl%C3%BCfteraggregate/PR/LAM5_/index.xhtml)

or sub-optimal solution:
http://www.fischerelektronik.de/pim42/upload/fischerData/image/web/la5.jpg
http://www.fischerelektronik.de/web_fischer/de_DE/K%C3%BChlk%C3%B6rper/D04/L%C3%BCfteraggregate%20mit%20Axiall%C3%BCfter/PR/LA5_/index.xhtml (http://www.fischerelektronik.de/web_fischer/de_DE/K%C3%BChlk%C3%B6rper/D04/L%C3%BCfteraggregate%20mit%20Axiall%C3%BCfter/PR/LA5_/index.xhtml)

You could use join 4 of the first one to make a cube

math :)

price at rs-components is 30,56eur for 1 with 12V fan.
not really "cheap" but I will give them a try when everything is ready.

https://i.imgur.com/DbQYe7n.jpg
Assembled "Blade" consists of three K16s and an aluminum frame. The frame has two features to secure it to the chassis: a tab on the bottom and a flange with a hole for a screw.

https://i.imgur.com/aofWL7X.jpg
Blade tab fits into the slot in the chassis board and a machine screw is used to secure it down. This method is quick & dirty, but not very robust - that flange isn't going to keep things rattling around if the enclosure gets bounced around. For my purposes it's fine as I don't plan on transporting the finished enclosure anywhere.

I am surprised to see so many push pull fan setups in a closed case. If the pull fan sucks out more air than the push fan pushes in, the air pressure will be low in the case and there will be less air molecules to pickup and transport the heat.

I expect a push only fan setup to be much better, there will be lots of air molecules to pick up the heat and the air has no way to go except out.

Maybe even forcing a high pressure in the case by limiting the outgoing openings of the case. For example a 12cm fan blowing towards a 8x8 cm opening with smooth transition. I would like to use 12cm, so the board is 1cm away from the housing, creating some airflow over the remaining electronics on top of the board too.

push-pull is good if you can't get enough airflow to keep the air cool enough across the case so that the components farthest from the push fans can still be cooled adequately.

The push-only makes sense... Thats how most rack cases are designed also...

How about push from both ends, and vent from top? would that cause bad circulation for lower boards?
Imagine if there is an ATX PSU on the top back (like in a regular PC case), Thru the  psu is the only way for air to get out... All other fans potentially RPM controlled based on temp....

Even for push-pull config, one of the pull fans could be removed... to provide space for cables, etc...

Interesting read here: http://martinsliquidlab.i4memory.com/Radiator-Fan-Orientation-And-Shroud-Testing-Review.html
My gut feeling was wrong. Push + pull is better than just push. And adding two shrouds to it makes it even better.

Shrouds may not be needed. The center lane of the heatsink would not need much heat dissipation. Because of the design of K16, the heat is near corners of the boards, hence the outer portions of the heatsink will get hotter. So if the fan motor blocks some airflow in the center lane... then its a feature not a bug.

Need some help, When i downloaded the files from https://github.com/bkkcoins/klondike i converted the KiCad file to vrml and later to sldprt(Solidworks part) and that looks like this http://imgur.com/Fg3O9kn which look good but there is two problems, the file is close to 100MB which is very large for a CAD file and it is laggy to work with it, and i can´t edit it or measure it.
From the pictures I´ve seen in this thread it has been very few of the components on the PCB, where did you get these "more simple" files, could someone send it to me(i would prefer a sldprt file) or give me the measure of the PCB on some of the components.

PCB card, 100mmx100mm height=?
Diameter of holes and how far the holes is from the edge of PCB?
The length,height,width and position of the rest of components(chips etc)

Another question, How do i connect/power the Klondike to a computer, could it be USB, and which software does it use(is it just plug it in and mine or do I need to configure something?)

man in SW all you need for real part design is to make a plate part:
2 mm thick, 100x100. Mounting holes 90x90 , heat sink holes 62x55 all 4 mm DIA
That's all.
If you want to make rendered image this is different story

also, from main thread, keep in mind pcb is designed to be mounted on heatsink, and not the other way round. i.e. pcb cannot take the weight of the heatsink.

Also use M3 screws and not M4 for mounting to account for expansion under heat, etc.

VMRL import in SolidWorks is okay for basic components, but doesn't manage complex assemblies well. You may have better luck with the scanto3D plugin (which is what I used). You'll need a ton of RAM to import the full K16 model, I think I topped off at ~6GB.

Here's a link to the sldprt file of the bare board with no components:
https://www.dropbox.com/s/ks2hh709gpppx5d/K16-PCB-bare.SLDPRT

I believe the part is actually 1.6mm, rather than 2mm as marto stated - at least as per the VRML file that Bkk provided. It makes little difference either way.

yes  it is , but all I needed there was a representation to be able to check the design of heat sink and enclosure

Thanks :D , but unfortunately I have Solidworks 2012 SP03, and I can´t open the file, it says it is a "future version", I don´t know how to deal with this.
If you either don´t know can you tell me all the measures or upload a picture of all the measures and the PCB. I got the board 100x100x1.6 but I don´t have a measures of the holes and chips etc(it is enough with the holes but it could be good the have the other components too)

Try this one: https://www.dropbox.com/s/11eznnw47fyk5wt/K16-PCB-bare.STEP

Open it as a STEP file, and it should import fine.

Sorry for the offtopic, but you case designers, would you be willing to design a computer case, according to some specifics/requirements provided by private customers, and how much would you require payment?

EXCELENT stuff i'm seeing!

EDIT: designer not needed anymore as for now.

//offtopic

Thanks you very very much, everything works fine, now i can start with the case 

http://imgur.com/QyMdL2S (eyesore warning)

Thats the limits of my artistic abilities... i am design challenged. If this idea has merit perhaps one of you could illustrate it better using some fancy 3D tool. Goal is easy maintenance and spacious layout.

What im thinking... have blades of 4 K16 in a row. The entire blade is the heatsink with holes in appropriate places. This shouldnt be difficult since aluminium is extruded, and easy to get in whatever length needed. (Its sold in very very long bars and cut to spec). Mount the K16s onto the heatsink, So now you effectively have ~45cm long k64 with some mounting mechanism(not clearly illustrated) in the ends.

Place 2 metal bars ~40 cm apart and mount blades vertically to it. provide fans between blades.

The first picture is the full installation of blades on the rods. The next is the front side of the blade with the pcb facing you. The last is the back side (with heat sink finns).

The smooth surface of the blade could either have M3 screw holes so the pcb can be mounted using normal M3 screw, OR it could have a thru hole where the PCB is mounted using nut + bolt.

Just find a cabinet (or frame or box) or something half meter deep, and you can stuff loads of these into it. To go more fancy, make a backplane on on one end with connectors, and handles on the other end.

A 3U rack case (or even normal size pc case) could easily fit 6 of these . = 6 X 4 X K16 = 384 chips. and remain be very spacious.

So what's the purpose of the mounting holes, then?

maybe if u use a light heatsink... 32W of heat does not need very thick heatsinks as most people are planing...

Do I understand this right that the PCB is 1.6mm thick?

Ente

In theory, you are right - a gas will have less molecules per volume at less pressure, and we need as many molecules to transport that heat away as possible.

However, the kinds of fans we are talking about have very, very low "pressure" or "pull". We are not talking "vacuum cleaner" here, we are talking "to fan with a sheet of paper". Which means that
1) the "pressure-reduction" *any* fan-setup can archieve is very small. I would guess less than 1% reduction in pressure, molecules, heatcapacity.
2) The volume throughput will go down very quick, as soon as there is the slightest resistance for the fan. Meaning if the fan-datasheet states 100 cubic-something per minute, it will be reduced to 90 by just blowing into an empty case with a large hole at the opposite side. I would guess in real setups you have anything between 80% and 25% of the stated "free-running" air-throughput.
Now you can have more throughput with two fans. And push-pull helps more with the resistance-issue than it eats through the molecule-density-issue. By some magnitudes.

Also, there are axial and radial fans, the latter with way higher pressure.

Ente

That is my understanding.

Bicknellski... i like your rack-plan. I would love to see an offer in the forum where one offers such a 2U (or whatever it becomes) and includes the cables needed, a matching psu that can handle overclocking too and a raspberry or something to run the software needed. Most probably a software will be needed to administrate everything from the net too.

Maybe even create a similar thing for burnins bitburner.

The datacenters seems not too expensive for such things.

I hope something is developing here. I would buy.

Does one have to build such cases or is it prebuilt to buy somewhere? I probably dont get where to find such things or how they are named.

I assume the designer's goal was to build it themselves.

If i needed to stuff them in cases, what id do is goto a shop selling second hand equipment (plenty of large shops selling used computers (p2/p3/p4), audio/video equipment, etc almost at scrap value) , and find some case that roughly matches the width and is off appropriate depth.

5 years ago i bought some pentium 4 computers for office at ~$60 each ... i am still harvesting them for fans/wires/etc...

But the thing is that this is a 19"-enclosure to be put into a server rack. i doubt its easily possible to change a tower into such an enclosure.

Sab

Looking at the overclocking potential I think mineral oil and something along the lines of the http://www.grcooling.com/ guys set up is what I will eventually have going in the 1U variety / Gold Pan K256. No fan, keep the heat sinks. I will keep you posted as I get closer to a finished K256 and post everything up. In terms of sales... might be best to have it as kit or something where people finish them locally. Shipping anything this size will cost an arm a leg and a few internal organs plus given the issues with customs and tariffs? Not worth it but maybe a hosting situation where you buy the unit and we host it here in Indonesia might be promising but given the difficulty rise I can only see one way to go and that is group buy coop's.

http://www.youtube.com/watch?feature=player_embedded&v=6IX9U2zaI_I

Thanks for answering. When you go the mineral route dont you think the hosting will be considerably more expensive? And it will be hard to find a hoster anyway i guess. Plus i read mineral oil isnt meant for running 24hours 7 days a week because its only made for dissipating heat fast. So over time the oil will become hotter and hotter, so the oil has to be cooled down too then.

Regarding DIY... i already search such rack shelfs but cant find anything looking near your and the other picture shown. But maybe i only dont have the correct searchwords.

32W per Klondike means a 1000W psu is only good for a couple of miners. On top you need to have around 4A at 12V for each of the miners. I believe most psus will be able to power less miners than they could by wattage because the Ampere on 12V is lower than this amount.
But i believe chosing the correct psu is only possible once burnin and bkkcoins tested the overclocking abilities correctly. If there are differences the price MH/$ will maybe change again too.

I searched a bit for colocation racks and found that hetzner, in germany has this offers: http://www.hetzner.de/hosting/produkte_colocation/basic and they claim a maximum server depth of 750mm. Shouldnt this be played out or is it so unusual that only hetzner offers it? They even have racks where the depth could be up to 1150mm.

Cool designs

I searched for a cheap solution and im wondering what you think about this. And very sorry for the very poor skills in gimp... :)

Its based on 2 of these 2U racks: Link (http://www.ebay.de/itm/2-HE-19-Rackablage-2-Unit-Ablage-Tragereinschub-Rackwanne-NEU-/200733902723?pt=DE_Computing_Rackmontage_Schr%C3%A4nke_Rahmen&hash=item2ebcac4783)

https://i.imgur.com/LNorzrp.jpg (http://imgur.com/LNorzrp)

Since each is 375mm depth i thought chaining 2 would maximize the 750mm hetzner is offering. I think they could be chained by bending the front flags that are made to mount them with screws on the colocation rack to the side and attach it to the other rack unit at the bottom. At the top i would use a metalpiece to connect both parts with screws. It should be possible to make it strong enough. If the border at the bottom end of the first rack disturbs then it could be bend too.

Then i would take 2 miners. Burnin or klondike and put both heatsinks together. Then attach a fan and make something around so that the air is forced through the heatsinks and cant escape. Then put more miners after this and so on. At the end another fan like in bicknellskis image.
All miners then are attached to the bottom with screws.

The image shows 4 pcbs without heatsink mounted to the bottom and 2 pcbs with 2 heatsinks each put together and mounted to the bottom. Again, sorry for the bad art. :)

Regarding the psus... 2U should be the standard height of ATX-Power supplies and those are the most cost efficient because they are widely bought and cheap produced. So i would add one or more of those at the end and put the pull fan at each minerrow a bit more to the beginning so that the airflow still is possible and the psu has enough space.

Its a cheap solution, the rack would cost 2 x 20€ only. 20€ if you would go the standard 37.5mm

Will this work?

But even when... how to remotely administrate it? Is one raspberry enough to run cgminer for all miners? And can this raspberry be reached from the net to change something? Maybe restart and so on? Anyone has a solution for the administration?
Maybe its even better to have 2 raspberries? One that can restart the first when it stopped working? Or is this overload and a cheaper unit could be used for this?

I think ur thinking of 3U. 2U feels small to fit in standard ATX PSU. Even if u can cram it in, the air inlet would probably be sealed tight by the case.
Keep in mind its not only the rackspace ur renting. You are also buying electricity (and the cooling capacity to get rid of the heat). The datacenter will probably give you low power... and you would need to negotiate for more watts.


Linux... so ssh... Raspberry pi even has a watchdog with can make the pi reboot itself for most failures... worst case, you can ask the datacenter to power cycle your device. Some datacenters provide power strips u can power cycle remotely.

2U would be 88.9mm after standard. I checked my seasonic and i found the heigth is 86mm.
I dont think the airflow will be broken too hard with it because i dont want to create a wall of psus at the end. I only want to place one or 2 psu one after another at the end of one of the middle miner-rows maybe. So that the pulling fans can still throw the air around the psus.

Yes, i know about the power. In fact a full colocation rack with over 300 klondikes will have the powercost as biggest factor. Thats why i already look for countries with cheap power to search colocation hosting there. The rent itself is the lowest cost.

So regarding the raspberry... you say one unit could be enough for one 2u-rack? Or even the whole colocation rack maybe? And it would be possible to check over the net how everything works?
Is this safe then or might someone be able to attack from the net and do something bad to the miners? I dont have a clue how this would work.

The pi is probably one of the cheapest components in the setup... In theory it could possibly run the entire rack... but id suggest maybe put 1 pi per case.. or something. This needs testing how many K16s can single pi handle... It boils down to distributing the risk by having multiple isolated systems vs one large system.

Nothing is safe. Everything that can break will break. Everything that can be attacked will be attacked, use your own judgement. I would say if setup correctly the odds of being pwnd is greatly reduced. Seems like you need to wait for someone to make raspberry pi image for k16 (maybe minepeon or something) and trust they did the right thing.

Yes, i guess i have to wait someone providing something. I only thought since the raspberry is directly on the net, and the admin has to reach it through the net too, it sounds dangerous. But maybe the raspberry is capable of protecting itself good enough.

I think I have personally assessed the data center idea for the current K256 idea and at least in Indonesia it is actually cheaper to rent space and build a customized data center. Indonesia has a lot of positives for going this route. Manpower is abundant especially technical know how. Space in terms of commercial buildings with access to sufficient power. Finally given fabrication of PCBs can happen at a reasonable cost I think the COOP we have set up might be a way to get things jumped started here in Indonesia. If we can see Generation 2 chips from Avalon and BKKCoins can modify his current design to suit I think then a 1U will be possible for a standard data center. Given the overclocking the 1U I propose will need 700W and the oil submerged system, with heat exchange and an external cooling tower. Given the heat in Indonesia this might just be the most cost effective system in the longer term.

My plan.

1 - K256 prototype air-cooled (600-750W PSU)
15 - K16s on stanchions with my first batch no cases
44 - K1 nanos

I do love some of the case ideas here... especially the CUBES.

Interesting info on oil choices:
IEC 60422, IEC 60296 (Google search should turn up some useful links) and

http://www.power-eng.com/whitepapers/2012/shell.whitepaperpdf.render.pdf (http://www.power-eng.com/whitepapers/2012/shell.whitepaperpdf.render.pdf)

I think oil and cubes design could be integrated. If you couple the heated oil to water, you can generate hot water and save on that as well. Some ideas to ponder from the world of power transformers http://www.electrical4u.com/transformer-cooling-system-and-methods/ (http://www.electrical4u.com/transformer-cooling-system-and-methods/)

Not to be contentious, but Sebastian I think you are mistaken regarding the operation of equipment 24/7. Intel and many other data centers have done exactly that. Intel did a one year pilot study and follow up failure analysis indicated nothing of mention. There are certainly other issues in terms of maintenance, which is where good design can make a significant difference. Modular mining units (the blade designs appeal too for this reason) that have snap on feature seem like a good idea IMHO. If you search for the Intel pilot study, it is also interesting to note the significant decrease in power consumption. Considering that these miners will have a carbon footprint the likes of Godzilla rampaging through downtown Tokyo, this is a good thing *financially* for miners as well. Less cost up front.

That being said, there is the oil and despite the videos demonstrating ease of service http://www.youtube.com/watch?v=U5zoIEjo1Zk (http://www.youtube.com/watch?v=U5zoIEjo1Zk) there's still the oil. The fewer small parts to drop, misplace, recover from the bottom of the oil reservoir, the better.

Lots of design opportunities here... I don't think ordering a GRCooling Carnojet system is necessarily the most prudent solution for everyone... as pointed out previously the shipping costs are likely crippling.

Can you tell where this power decrease comes from and in what range it lies?

Air conditioning most likely... maintaining adequate temp gradient between environment and hardware.

http://phys.org/news/2012-09-intel-math-oil-dunk-cooler-servers.html
http://www.datacenterdynamics.com/focus/archive/2012/03/cgg-veritas-uses-liquid-cooled-servers-hpc-environment

Range... would be worth looking at heat production of existing Avalon systems. Need to poke around on forum for that... and electricity cost is another variable. Commercial/industrial rates are typically cheaper than residential.

GREAT INFO there FullFathom...

Love the re-purpose of the heat to get a warm shower brilliant.

...and when you want to remove the oil... for reselling or relocating...

http://www.youtube.com/watch?v=huGl2hfzL90

 :o

Intel Uses Mineral Oil to Cool Servers, Finds Success

By Tuan Mai

SEPTEMBER 5, 2012 10:00 AM - Source: Wired

http://www.tomshardware.com/news/Server-Cooling-Hardware-mineral-oil,17348.html (http://www.tomshardware.com/news/Server-Cooling-Hardware-mineral-oil,17348.html)

The 3RU case can be bought from any number of vendors. I went with Metcase's "Combimet" series: http://www.metcase.co.uk/enclosures/combimet19.htm

I'm having parts machined to customize the enclosure to house 24 x K16s in 8 x 3 "blade" rows.

https://i.imgur.com/aofWL7X.jpg

I have designed and will be offering for sale:

Klondike 16 to 3.5" HDD Universal Adapter Mount (allows horizontal or vertical mounting)
Klondike 16 to PCIE x16 and x8 (maybe x4/x1 if they are strong enough) slot Adapter Mount
Klondike 16 to PCI upper-rail support (like those used in certain 4U rackmount cases).  For example there's an $85 Chenbro 4U case on Newegg that could mount at least 13 K16 boards using 5x 3.5" adapters and 8x PCIE upper rail adapters.

These will be 3D-printed.  I've already printed and tested them.   They work very well.  

Also, might sell 'kits' with I2C comm cables, and PCIE 6 pin power cables for the K16's.

Pics later.   I'd like to know if there's any interest before going too far down this road.

Prices would be cheap.  *very rough guess* $8 per adapter, with $5 shipping in CONUS.  Kits for $12 or $15, bulk discounts, etc.

I like the idea of using a 3.5" hdd adapter to fill empty drive bays in my existing pc. Definitely interested.

Klondike 16 Mounting Adapters here with pictures:

https://bitcointalk.org/index.php?topic=246638.0

Nice simple solution... I have added your bracket mounting solution to the DIY Reference thread. https://bitcointalk.org/index.php?topic=227186.0

I'd like to see your mock up for the cabling and with heat sink attached to see the spacing you can achieve in the 3.5 HDD bays. Also our coop might be interested in bulk orders of this.

It's my understanding that as of now, there's no standard heatsink for Klondike.  That's one reason I haven't gone further in my designs.    If the heatsink + PCB height is 20mm or less, it should be possible to fit one K16 per 3.5" HDD bay.

Mock it up.

Given that most heat sinks are going to work to remove the heat we are not constrained really. But it be nice to see a mock up with cables etc.

This is how you could mount K16's (and K64's) into a standard PC case using Micro-ATX form factor. 

http://www.forked.net/~apex/k16/K16_micro_atx_x16.png

http://www.forked.net/~apex/k16/K16_micro_atx_populated.png

Heatsinks are 20mm tall.

From the looks of it, it seems that the PCB will not be able to tolerate the weight of the heatsink. They would be too heavy. The heatsink would need to be mounted on the case, and the PCB mounted on the heatsink.

Your adapter idea game me a somewhat related idea. Find heatsink thats the size of a HDD and drill holes into it. The thin pcb on top of it. That should fit everywhere a HDD can go... Length of 3.5" HDD is 146mm so make it a little longer, and stick 2 K16 per heatsink.

I could certainly test the strength - I have a lot of heatsinks laying around.  But I think it will be more than adequate.  

If it is an issue the adapters can be extended upward, and the heatsink bolted directly to the adapter, thereby sandwiching the PCB between the two.


[edit]  Picture with heatsink support:

https://bitcointalk.org/index.php?topic=246638.msg2621173#msg2621173

Hmm, I like your idea of placing two K16's lengthwise into a 3.5" HDD bay... :)

42 K16 in 1 3RU Rack with 355mm depth

http://abload.de/img/entwurf2qxske.jpeg

http://abload.de/img/entwurf36fsax.jpeg

I wonder is it doable this way with cables attached? And i think the airflow to cool the chips isnt optimized. There is space between the heatsinks so the air isnt forced through the heatsinks and you would need a cover at the top to prevent the air from going the easier way. Only my thoughts because the new known overclocking capability will be used by many i believe.

These heatsinks aren't optimized for airflow, so it won't have such a great effect.

If someone contemplates to purchase such a 3U case, i could provide a service of building and shipping.

Does someone know what material would be best to form an air tunnel? Im not sure whats best. It should be flexible so that i can bend it to create a tunnel. For example tape it on the fan so that the air flows in it, then connect to the heatsinks and so on.
The material would need to be dynamic enough but stable too. And im not sure how to fasten it. Glue strip? It should be tight so no air loss happens, it shouldnt create shortcuts for sure an so on.
I have some ESD-glue-strip here, maybe using this to fasten the tunnel material?

Is there a Sketchup model of the K-16 board?

Interesting... should I ask Spondoolies if they stole my crappy design idea?

LOL.

http://i230.photobucket.com/albums/ee106/PFC4L1FE/spondoolies1_zps41c98e3b.jpg

Knew it was possible.

LOL,
You are inventing the real Bick  ;D
Please take my apologies but while you are posting pictures and blinking LEDs other folks do real work.
Read this once again
http://www.spondoolies-tech.com/pages/team
I am sure that they knew that this case is possible but they knew exactly why.
That is main difference between you and everybody
You knew EVERYTHING but you do not know why  ;D
I am sure they stole your brilliant idea ;D I do strongly suggest to give spondtech negative rating right away and bring red  light to lost soles ;D
Or even better   stop wasping around with their chips immediately. They do not deserve it right?
Buy the way how is your AMT PARTNER rating doing lately? They do desperately need your POSITIVE RATING.

Please hurry up and save Zipkin ass ;D  
We all know that everything is possible but someone has to do the job at the end.
That is something to discus on your next meeting ;D
Just for the record 512 avalon gen 1 are about 3200-3500 wats
And you knew that you need at least 2 psu's for that 512 box which are missing .....
And I know the your box was about to melt right away like AMT CRAP IS MELTING ;D
Just focus on blinking  LEDs  Bick. That is where you are damn good ;D