Are you ready for Stargate Command?

I started this thread somewhere else, but have gotten no responses there. So I'm hoping that this will be a more active environment for discussing this idea. Here's the project:

Originally Posted by Shingoshi

Ok. Here's the comedy behind the question. I have this board (a Tyan S4985: I have since RMA'd this board and now considering the S4992 instead) that's way too large for just about every case on the market. The board measures 16x13". Yeah, it's a four-socket behemoth. Anyway, last year I thought about building a computer inside of a Pelican 1780NF transport case. You know the cases, even if you didn't know who made them. The were constantly shown on Stargate SG-1. They were used regularly in the filming of that show.

So here's the idea.
1.) Get one Pelican 1780NF case. (internal dimensions: 42" x 22" x 15.1")
2.) Build all of the mounts required for my 5 boards and accessories inside that single case.
3.) Get a Silverstone ST1500 power supply. Hell, maybe even two of them.
4.) Keep the liquid-cooling I have now on the Tyan S4985 board.
5.) Get eight Dynatron A5AG heatsinks (I have four Tyan S2912s as well).
6.) Get an (as yet) unspecified number of Koolance radiators.
7.) Get as many Scythe Kaze Maru @1900 fans as needed.
8.) Mount some of the fans to the radiators.
9.) Secure the radiator/fan combinations inside the case.
10.) Get one very large automobile radiator.
11.) Mount the automotive radiators to the top of the Pelican case.
12.) Drill pass-through holes for the tubing.
13.) Connect the internal Koolance radiators to the external automotive radiators.

The heat from inside the case will be transfered to the outside by means of heat exchange. The smaller Koolance radiators will act like the heatpipes attached to a processor. The fans attached to the radiators will force the cooler air created by the cooled fluid from the external radiators, and circulate that air throughout the Pelican case. The sound produced within this system should be fully insulated.

Everything (the motherboards) can be connected to the outside by wireless. The power connections can be your standard waterproof types used for passing power to outside appliances. So the case could remain completely sealed, only opened for servicing as needed. Even the graphics cards could use waterproof connections, mounted to the walls of the case. So anything that absolutely has to be physically connected to something outside, would pass through a waterproof interface.

Now how's that for a proof of concept system? This might even be more attractive than Amanda Tapping. But then, I might have Tapping on my shoulder, for the insult!

NASA, USAF, NSA etc, I'm available!
Shingoshi

Here's a picture of the case.

And now that I have pictures of it!


And for the sake of showing how much volume this case has!

And here's why I needed something so grand as this project to contain it!

Upon further consideration, I'm thinking about which cooling techniques should be used. The two most popular are liquid and phase-change. And there is also the possibility of combining them in one system.

The big problem here for me is that I have four Tyan S2912 dual-socket motherboards which use the earlier 4.1" heatsink mounting pattern. The standard that is widely used now, has only 3.5" between the two opposing mounting points. Needless to say, trying to find the earlier mounting pitch is becoming more difficult. That's the reason why I mentioned using the Dynatron A5AG heatsinks. The fans I thought about using are no longer around. And I need 8 units in total for all four S2912 boards.

But this complicates things. Heatsinks aren't very efficient alone, unless they have a large volume of forced air flowing over them. And in an enclosed system like the one proposed, air flow is a big issue. So I'm considering another technique, possibly to be used in conjunction with the liquid-cooling system. Because the fact is, I still need to cool the entire environment, and not just the processors. For instance, the power supply must be cooled, as well as the hard drives.

It should be noted that heatsinks without fans are typically only used in 1U servers. Where the air flow can be properly directed without obstructions. But that requires often contrarotating fans turning at rather high rpms, creating with that, high noise. This was why I chose to install a liquid-cooling system in a multiprocessor system.

To manage the full cooling requirements of this system, I'm thinking about placing an evaporator (like the ones used in freezers) in the ceiling of the case. A small air-conditioning unit can be modified for this purpose. The case would essentially become a mini-freezer. Then with the air inside the case kept at such low temperatures, there should be no problem cooling all of the peripheral components, along with the processors.

Having an evaporator mounted into the top of the case would allow any warmed (and rising) air to be rechilled and directed back downward to the components. Employing a large number of fans (especially the Kaze Marus) would ensure that air reached anywhere it is needed.

Now all of this could be averted by not using the S2912 motherboards. But since I already have four of them ready to go, I really don't want to waste. If I could find the proper mounting plates for them (which Koolance doesn't make), I would liquid-cool the entire processor complex. Having those plates made, may not be that big of a deal. They would simply have to have the 4.1" separation between their mounting points.

Anyway, for the time being, I'll leave this here and return later after further thought.

Shingoshi

Edit: I may not have hard drives inside this case at all. I may instead opt for ssd.

I had a build idea to use a pelican case too... except mine was to build a Mini-ITX system that could run on battery for 8+ hours (be portable) and record uncompressed 1080p video through HDMI, as well as have a built-in 15.6" screen for previewing (if you weren't running on battery). and have it priced under $1000, and have it fit on a carry-on (if you took out the car battery -_- ).. Lithium Ion packs were $300 and would only get me 5-6 hours realistically, while a $40 car battery would get a lot more...

But sorta gave that up since I don't film as much anymore, but it was a cool concept... your's is a bit over mine haha...

Can't wait for some pics...

Since writing about this earlier, I too considered the use of batteries, but for backup (in a controlled shutdown) not long term usage.

Now to continue with the project. I have considered using the majority of components I already have without adding much more. Just this morning I thought that I don't need to purchase additional hard drives, since I already own eight of them:
1.) Two Samsung F1 1TB (in RAID1)
2.) Two Western Digital 1TB (also in RAID1)
3.) Four Seagate Barracuda 320GB (in RAID5)

I have also realized that I can simply boot the other (S2912) boards from USB drives. I can then take the eight drives I have and put them into a Network Attached Storage system.

Now for a basic design outline:
1.) Having the Pelican case laying flat in front of you, so that the lid opens away from you. We now have our proper orientation for discussion.
2.) Arrange the motherboards so that the S4985 is in the center of the case.
3.) Use rigid 2" long tubing to space the S2912s into pairs. I should explain this:
a) Using a number of tubes equal to the number of mounting points on the motherboards, place 2.5" long 1/8" threaded shafts through each hole (passing through the tubes as well). In that manner, the S2912 boards are held together stacked in pairs.
4.) Position each set of the S2912s to each side of the mainboard.
5.) In front of the motherboards (all lined up) we now install our radiator/fan combinations.
a) I currently have two B&M Racing Supercoolers. They were intended for oil-cooling, but work well in my application. They are in fact what I use in my liquid-cooling system. They measure 11" square and are very efficient at dispersing heat. They work so well in fact, that I am using them passively with idle temperatures of @28C. But I will have to continue testing that setup as I add more processors.
6.) Connect all of the radiators in series from one end of the case to the other.
7.) Use a Full-House water filter housing like those from Lowe's as my reservoir.

I'll connect the entire system so the liquid is drawn from the from reservoir into the radiators. From there, the liquid will exit the radiators and be pumped into the cooling blocks. The liquid will then once more return to the filter housing to complete the loop. Fans will be attached to the Supercoolers so that they draw rather than push air through. The design of the Supercoolers is such that they would have the air moving at a vector from the axis of the fans. It's for this reason, it's best to draw the air through, so that it moves in a straight line from the fans to where directed.

Again as mentioned above. The internal radiators are cooled by the evaporator in the top of the case. As long as the internal atmosphere is kept very cool, there is no need for an external heat exchange from the radiators to the outside of the case.

I'll continue later...
Shingoshi

When I looked at the Koolance GPU-200 graphics card cooling-block, I was concerned whether it could possibly work. I was also concerned as to whether the block would fully cover the chip. But I just pulled one of my boards out of the closet, and examined it. According to the specs, the arms of this unit should indeed extend to cover the mounting points on the cpu. And the base of this unit should be more than adequate for this purpose. Using these devices would allow me to liquid-cool my entire system. I will purchase one of these to test, to determine if they will accomplish the job. In the long run, using these would be a cheaper solution. And the temperatures would also be much cooler system-wide.

Shingoshi

Given that I will likely now use the Koolance GPU-200 on my S2912 boards, I will increase the separation between them to be 4". That would likely give me enough room to change processors without having to separate the board pairs, instead leaving them intact. Maintenance shouldn't be an issue with that setup.

Shingoshi

Initially, when I started thinking about this project, I was under the impression only the S4985 would be capable of liquid-cooling. In fact, my previous cooling loop was a drop-in replacement for the S4985. There were no modifications required to accommodate the new board. The problem as I mentioned above was the wider separation between the mounting studs for the S2912 motherboards.
S4985 = 3.5"
S2912 = 4.1"
That difference may not seem like much. But when you consider that the holes on the Koolance CPU-330/340 are pre-drilled for the smaller spacing, the other boards are then eliminated from those cooling blocks as a solution.

The Koolance GPU-200 however has swinging arms which extend to cover rather large distances (@5"). I knew from the diagrams that the GPU-200's arms would reach the mounting posts. But I wasn't sure whether the arms would be subject to bending upon being mounted. If the arms and their attachment points are rigid enough, none of this will be an issue.

So now having a uniform option of using liquid-cooling for everything, some of the design parameters for this project have changed. One of the things I had previously desired, was to not have any permanent connections between the top and bottom halves of the case. That can somewhat be tackled by using quick disconnects between components.

I thought about using a phase-change system in conjunction with liquid-cooling. The idea is to superchill the liquid and then cool the internal components with it. I've since looked at the Koolance HD-60s as evaporators, and possibly submerge them in a tank (reservoir) where the cooling fluid would pass over it. Being that antifreeze would be used, the temperatures achieved could be very low indeed. Cooling my system with a fluid many degrees below zero centigrade, would be very appealing indeed.

Essentially the entire computer case would become a minifreezer, with the liquid-cooling fluid being chilled by the environment. The foam of the case would act as it's insulation. And I could use gel-packs to maintain the low temperatures, reducing the amount of work the refrigeration system would have to expend. If I do use this format, the phase-change system should reside completely in the top half of the case. The condenser would be mounted externally to the top as well. And it would likely make sense to have the compressor sit outside of the case as well. For this project, an automotive AC compressor would be a better match for the conditions it would have to work in. It would simply have to be electrically powered, like those units on refrigeration trailers.

Shingoshi

http://www.filtersfast.com/Pentek-15...er-Housing.asp
This would have to be the monster of all reservoirs. Think how much money most of you have spent for some small reservoir (by comparison). This thing will sit inside my case and be chilled by the lower temperatures inside it. And being that I would be using an automotive radiator as the only (external) cooling device for the LC loop, having something this large would be very nice. This would allow the liquid to be precooled before being deep-chilled by the evaporator.

And this brings up another point. I think we've all overlooked the fact that the entire system will be cooled by the LC loop, whether the A/C system is installed/running or not. At no point is the LC loop interrupted. It functions no differently than any other LC loop. The only difference here is that it's temperatures are forced lower by being exposed to the cold expanding gases from the evaporator. But like I said, even if the A/C weren't running, the LC loop will be sufficient for "normal" operation.

I've become rather familiar with using NPT fittings, thanks to Lowes.com. I've purchased virtually all of my plumbing for my LC loop from Lowes. Having 1 1/2" NPT is not a problem for me, now that I know how to use them. In fact, I'm using two B&M Racing 11x11" SuperCoolers (#70274) for my radiators in my existing system. And they use 3/4" NPT fittings, which work easily with 3/4" to 1/2" hose barb adapters.

Shingoshi

How about some pictures? /subscribed

Originally Posted by elite.mafia

How about some pictures? /subscribed

I still have to get my Tyan S4989 mainboard. That won't be until August, likely. In the meantime, I'm documenting the progress of my planning here.

It would be nice to see someone else try this also. Then, we could compare notes and benefit from each others progress.

You'll have to be patient, just like me!
Shingoshi

Wow, crazy build! So what are you going to be using these machines for?

Originally Posted by extide

Wow, crazy build! So what are you going to be using these machines for?

I hope that title for this post answers your question. I've packaged software for personal use for a long time. But I began investigating the possibility of building my own specialized Linux Distribution. Since my interest tends towards mathematics and analysis, I wanted to have a machine that would suit that purpose. Additionally, I knew that if I wanted to build a distribution for clusters, I needed to be operating one of my own for testing of the software. So that's the background for this project.

You might want to look at this link also. As it is a sub-thread of this topic:
http://www.hardforum.com/showthread.php?t=1431845

Shingoshi

One of the main features of this project, is that it is intended to be completely sealed. With that in mind, every connection to the case, must be waterproof. So I'm going to start listing the components I've found to facilitate that, here:

https://www.photonage.com.au/index.p...oducts_id=2951
http://taitek.manufacturer.globalsou...GetProduct.htm
http://www.techpowerup.com/97115/Ste...anless_PC.html

These are things you simply MUST see!
http://www.stealth.com/littlepc_Waterproofpc500F.htm

Frequently check this post. And if you have suggestions, please make them.
Shingoshi

I started considering using a different means of chilling my liquid-loop. Water coolers came to mind. So I searched Google and found this (and it's cheap!):
http://www.air-n-water.com/product/W...:referralID=NA

How much more could you want than that?

Edit: Upon closer examination, I realize this unit is so small because it's a thermoelectric (peltier) device. Not exactly what I was looking for. So back to searching.

Shingoshi

But I have to confess, this just looks too damn sharp!
http://www.banjovalves.com/1x1thrfla...ankflange.aspx
It has that look that just says, I dare you to try and break me!

Shingoshi

Now that we have a committed relationship, I hope we have a long life together!

I just ordered my case! You can see it here:
http://www.tactical-store.com/ts-pe-...transport.html
This should make for a very good home for us to live in. Plenty of room for the kids.

We'll have so much room to play with, that we just might get lost in the backyard. Ok. Enough with the silly puns. I'm pleased!

Shingoshi

But I just found this while looking for supplies for my system. I don't know how strong they are, but they are very attractive. I guess for most builds not running at excessively high pressures, these will certainly do the job.

http://www.marinedepot.com/ps_ViewIt...ct~FT1513.html

Shingoshi

To summarize phase-change chiller construction in 3 steps:

- Assess heatload and desired coolant temperature
- Size the compressor and evaporator (and choose evaporator design)
- Assemble, test, and tweak

I wouldn't just assume that the compressor for an A/C can handle the heatload without running the numbers. You have the option of temperature control as well (compressor need not be always on at idle, for instance).

Edit: I see that you intend for the LC to be capable of dissipating the heat alone. Still - sizing the compressor/evap appropriately toward your end goal is key (even if they are part of a prebuilt A/C)

Edit2: If you plan to chill the water directly, then you should know that the radiators will work against you. You don't want your cold water exposed to that amount of surface area (it should instead be insulated of course).
If you plan chill the water using chilled air forced through the radiators, then this is far less efficient, but has additional benefits. If you completely seal and insulate the case then your case temp will be roughly the same as the coolant temperature, and thus condensation may not be an issue. The mechanical drives will require their own chamber, however.

Here is an example of this: http://www.xtremesystems.org/forums/...d.php?t=160872 --- This is extreme, but it's well-executed. Instead of -40, you could shoot for 10C below ambient perhaps.

The reasons I don't quite like phase change anymore:

- noise and bulk of the system (compressors can be noisy)
- power consumption (weighs on my conscience that I don't need to be wasting all that power)
- Less friendly to 24/7 operation than regular watercooling
- I'd want a fully DIY design, and though I'd love to devote the time to it...I shouldn't

If you're serious about phase-change then you might find more useful information at Xtremesystems.

If I'm going to air-condition my room anyway then I might run the A/C air directly to the radiator first, but then again I don't have multiple systems in my case. I wouldn't want all that heat in my room... I'd exhaust it out.

Ok... I read through some of your Prius thread. That one is quite different.

I'll need to read it more in-depth, but:
(And I'm sure this is all obvious to you, but I should walk through it anyway)

- In general a chiller will require several passes to achieve sub-zero coolant temps. It will take more than a few minutes if you have a large volume of coolant (which you will have).
- The coolant temperature doesn't change nearly as much as the refrigerant temperature.
- If the coolant temperature is still below ambient after exchange with the condensor, then the radiator will hurt performance (however it's needed for regular LC operation I assume).
- Liquid cooled condensors are not uncommon, but they are generally cooled by separate loops. It would seem counterintuitive (you may say imaginative, if your design works) to reintroduce heat into the coolant loop from the heat exchanger responsible for chilling the coolant (creating that absence of heat).

Now if you have run the numbers on your design and it works, then ok. Unfortunately I don't have time to really analyze it right now, but you sound like you know what you're doing. There's no way you'd overlook trivialities such as zeroth and first laws of thermodynamics.

Edit: My tone sounds condescending here. Unintentional.

One tip: The Primochill typhoon 3 + Laing D5 for parallel water loops using a single pump. Unfortunately this is a reservoir/pump combo, but it's amazingly efficient, as tests will soon reveal.

But then again, that's novice within the scope of my knowledge. And there are more things that I need to know and take into consideration. So let me just say it right now, Thanks!

I appreciate the interest you have shown here. I will hopefully consider every piece of knowledge you or anyone else has to offer.

About the philosophy of this build. I keep changing things. This happens partly due to cost limitations. I simply don't have the means to explore the full extent of my ideas. I would be homeless if I did. And where would I put my computer then? So I have to consider the long-term consequences of my attempting this.

My other thread goes further into the details surrounding the actually cooling system itself. You can find it here:
http://www.hardforum.com/showthread.php?t=1431845

The reservoirs which I have yet to receive are very large. The sumps/tanks on them alone are 20" deep, and tapered downward from almost 8" in diameter. The two intakes on them are 1" FNPT. I'm thinking about replacing the heads with 1 1/2" fittings. You need to read my other thread to see what all is involved there. I got two of these Whole House Water Filters by Watts/Pentek (http://www.waterfilters.net/Big-Clea..._p_41-575.html). I made special effort to get the clear 20" housings which are soon to disappear (at least according to Pentek). I simply want to see what's going on inside the tanks. For one thing, I'm going to do something very radical. And it will involve an agitated visual experience like none other.

Provided that the bottoms of the reservoirs are semi-flat enough for me to work with, I'm going to drill a hole in the bottom of each of them. There's a special apparatus I want to install in them, sealed with Banjo bulkhead connectors. This thing is really going to look ALIEN!!

Most people will look at it and have no idea what's going on with all of the technologies that I'm incorporating in this design. And many won't be able to interpret the actual function of each component. But I'm getting ahead of myself to say that. I could be dead tomorrow, and none of this come to pass. So I should at least wait until I have the reservoirs to decide what truly can be done with this build.

Shingoshi