[Build log] Desert Sapphire - External water cooling [Finished! (2016-Jan-03)]

[brandishwar]

Time to introduce my first client build. All of the primary hardware has been purchased and validated, and I'm just waiting on water cooling parts to arrive. This one is called Desert Sapphire. It'll have a mostly blue theme and will be water cooled.

Mainboard: ASRock X99 Extreme4
CPU: Intel i7-5820k
RAM: 16GB EVGA DDR4-2800 SuperSC (4x4GB)
Storage: Samsung 850 EVO M.2 SSD
Graphics: EVGA GTX 980 SC ACX 2.0
OS: Windows 10 Home
Chassis: Corsair 750D

This is going to be a very high-end build, and it'll be the first truly high-end system the client has ever owned, and he's very excited about getting this system. Currently he's running a second-generation Intel i3 with a Zotac GT620, so the difference here is like night and day. There's also an interesting story behind this build.

The client had actually been out of work for a while but wasn't really actively looking for work -- he's a younger man in his early 20s still living at home. So I proposed to him that I'd build him a system if he found a job and put up some of the cost. So he's put up about half the cost of the system, and I'm absorbing the rest. The chassis is actually the 750D from β Ori. (link in signature) that will be available once I move that system into a 19" rack.

And as I said, the system is going to be water cooled as well. Many of the parts have already been settled and/or ordered, at least on brands.

CPU block: EK Supremacy EVO
GPU block: Aquacomputer kryographics with passive backplate
Radiators: AlphaCool ST30, size to be determined
Pump: Koolance PMP-420 or PMP-500

This system will actually be externally water-cooled. There will be an inlet and outlet to the chassis that will be connected via Koolance quick disconnects, very similar to what I did for Colony West. The question that's up in the air still is what to use to house the external water cooling setup.

And I've had a few ideas, all of which so far involve modifying something from off the shelf. The idea on that mark is I want it to look like a piece of furniture with the tubing and cabling going from the back of the computer to the box. I've looked at night stands and small cabinets mostly.

I'm doing this for a couple reasons. First, the client lives in Nevada -- i.e. desert, hence the name. This means that there will be what some would consider over-kill radiator capacity. I'm currently planning on 6x120mm. The pump is also a little up in the air as I'm considering going with something more powerful than a D5 to ensure there won't be any flow problems. On that I'm considering two pumps from Koolance: PMP-420 and PMP-500. I may end up going with a D5 in the end, simply but again flow is the primary concern.

The other reason I'm doing this is simply because I can. The client has voiced no objection to going that direction with the loop and trusts my judgment on this. But at the same time I'm wanting to demonstrate what I talked about in a segment of the β Ori. build log and that AntVenom demonstrated with his setup:

The box will contain the pump, reservoir, and radiators. Coolant will pass from the pump up to the system, through the blocks, then back out to the box where it will pass through the radiators and go back to the reservoir. The quick disconnects will be color coordinated to ensure that the flow goes in the right direction. There will be a small ATX power supply controlling the fans and pump that will be synchronized to the main power supply. More on that when I get there.

And here's the best part: the cooling system is independent of the chassis. This means that the client could later add a second GTX 980 without affecting the radiator box. It'd affect only the main chassis, and not really all that much. Plus with all of the radiators external to the system, the radiators will have completely open airflow, meaning maximum cooling potential. The chassis could even be changed out later without having to worry about the radiators, and the radiators could be upgraded or additional ones added without worrying about the chassis.

Doing this provides for so much flexibility I'm surprised so few people are doing this. While there are some complications with going this route, they are relatively easy to overcome, provided you actually put some thinking into it.

The only part of owning a water cooled system that the setup would complicate is maintenance, but even then not really by much. It is something I'll have to keep in mind when building the box, though, as the client will likely be the one maintaining this system unless I agree to go back out there about once a year. Aside from that, all he'd have to do is keep the fans and radiators cleared of dust.

That's it for this part of the build log. Some of the water cooling parts have been ordered, but with Thanksgiving around the corner, it'll be interesting to see when they arrive. The Aquacomputer block was also ordered direct from Aquacomputer in Germany, so hopefully it won't take over a month to receive it. But if that happens, there will still be plenty to work on in the mean time.

[brandishwar]

I mentioned in the previous segment that water cooling parts were on their way. The first order from Performance-PCs arrived the day after Thanksgiving -- thank you FedEx for not screwing up this time.

The Performance-PCs order contained two radiators, two bulkhead fittings, and a collection of Koolance quick disconnects:

• AlphaCool ST30 double and triple-120mm radiator
• Koolance QD3-FS10X13 and QD3-FS10X13-BK
• Koolance QD3-MSG4 and QD3-MSG4-BK
• Koolance BKT-PCI-G
• AlphaCool Bulkhead Chrome and Black

One of the items listed is the Koolance L-Bracket with two G¼ pass-through fittings. Provided the quick disconnects could fit, I intended to use this for the pass-through to avoid modding the case. I know I've said previously that fittings shouldn't be ordered early in a build. This was a necessary exception as it would affect planning if this didn't fit. Plus I knew I'd be using these specific quick disconnects, so there wasn't any risk ordering these up front.

Thankfully it did fit, but really only just. The female side of the quick disconnects have 3/8"x1/2" compression fittings. Would it have worked if the quick disconnects were the larger 1/2"x3/4" compression fittings? That's difficult to say. It certainly would've been a tighter fit.

I really wish that Koolance had a second version of this pass-through that didn't have the DB-15 slot on it and had the G¼" fittings spread out a little more. I'd also like to see a low-profile version of this bracket similar to the one they make for the since-discontinued VL2 line of quick disconnects.

Note as well that I'm using off-colored fittings. And as you can see from the picture above, I'm using one chrome and one black quick disconnect. This is simply to ensure that everything stays coordinated for flow. The flow must go in a particular direction through the CPU block, so this helps ensure that the flow stays coordinated between the cooling box and the blocks.

Second order

The next payday saw another order to Performance PCs, again ordering what I knew I'd need.

• Koolance PMP-420
• EK Supremacy EVO -- copper/acetal, solid black
• Swiftech 3/8"x1/2" compression fittings - 7 chrome and 1 black
• Add2PSU
• PrimoChill PETG tubing - 3/8"x1/2"
• Alphacool HF L-Connector (90-degree rotary fittings)
• Tygon E-3603 3/8" ID (1/2" OD) Clear Tubing
• Mayhems X1 Clear Concentrate

Everything in the 750D will use soft tubing, while the water box will use PETG. I have quite a few PrimoChill revolver fittings, so this gives me an opportunity to use those up. Plus it'll make the water box a little easier to maintain. If there's any part of this build I want to be able to last a long time, it's that. I'm considering ordering the ratcheting cutting tool that MNPCTech sells, but I hear that you can use a standard tubing cutter on PETG, so I'll see how well that works first, and I have a hacksaw if need be.

The AlphaCool 90-degree fittings are new. I typically go with Swiftech or Koolance, but these were 2.50 USD less expensive than the Swiftech fittings, 3 USD less expensive than Koolance, so I decided to try them.

But as you can see, I opted for the PMP-420 from Koolance as the pump. Again I wanted something better than the D5 Vario. The question is noise. I'll be using this to flush pumps and blocks, so I'll be able to test it out. If the noise is a bit much, I can see about dampening it. And the Add2PSU is to synchronize the power supplies. The exact cabling I'll be using to plug everything up is still up in the air, but given how some LP4 to P4 converter cables are wired, I'm heavily leaning toward using a P4 cable.

Otherwise the rest of that inventory is pretty self-explanatory. Really about all that's left to order is the fans for the radiators -- likely going with Cougar CF-V12HPB since I've had good luck with them already -- and the FlexATX power supply for powering the water box. I have another FlexATX power supply I might re-use for this. I'll just need to first test it to see how noisy it was, so I'll likely use it as part of testing the pump for flushing components.

Thankfully the Aquacomputer block and backplate didn't take too long getting over to the US. Still could've been faster, but it would've required paying double the shipping rate for UPS. I basically had about everything I'd need when I'd need it.

The beauty of all of this? I can install blocks and tube up the loop in a smaller case before moving it to the 750D, and I won't have to change a thing when moving it. Whereas with a loop where everything is inside the chassis, you may need to cut new tubing, and you'd certainly need to consider compatibility with your radiators before buying.

The cabinet

The next step really is figuring out what I'm going to do for the cabinet and radiator configuration. I've been browsing around for cabinets to see what is available, what might be easily adapted. And more and more I'm leaning more toward building a custom cabinet for this instead of using something off the shelf. I'm having a little difficulty finding cabinets that will fit what I'm wanting to do.

So instead I'm going to buy oak or pine boards and build one. It'll be 30" long by 15" deep, and about 24" tall. The inside of the cabinet will be only 18" tall, providing for 6" beneath the cabinet. This will be good since the radiator side of the cabinet will actually be bottomless, well almost. The radiators will be all but completely exposed, allowing for the most airflow through the fans.

The doors to the cabinet will also be solid, hence why the radiator side needs to be bottomless. If the door is closed, there won't be any airflow if the cabinet was a typical cabinet. However with a bottomless cabinet, the radiators will still be able to pull air through. And the cabinet door can always be opened to allow for even more airflow if necessary.

The fans will have a controller as well. The Phobya fan controller I previously used in Absinthe will be used here to allow the client to control the fan speeds when desirable. It'll also serve as an immediate tachometer readout on the fans and pump as well. I forget whether it has an audible alarm, though.

Now in talking about "sides" to the cabinet, there will obviously be a dividing wall in the middle. Currently I'm planning to set it up so the left side of the cabinet has the fan controller and temperature readout for the coolant while the right side has the radiators, pump, and reservoir. The pump has a heatsink and being passively cooled by the fans moving air through to the radiators should suffice.

Otherwise, this is how the build currently looks in its temporary chassis:

[brandishwar]

Ah I love international orders...

Thankfully everything in the box was unharmed. The only casualty was a corner of the thermal pad for the passive backplate. Not anything that's concerning, though. The package from Performance-PCs came later in the week. That was the interesting one: fittings, pump, CPU block, tubing, fittings, fittings, and more fittings...

One thing I need to say about the pump: how in the hell is it capable of the specifications Koolance declares? And I mean that will all sincerity. If you look at the specification drawing, it's about 47mm square. I was expecting this to be almost as large as a D5. 47mm, for those in the US who don't want to pull out a calculator, is just under 2". And no I didn't look at the dimensions ahead of buying it.

On the plus side, it uses just 1.8A and runs off a 3-pin fan header, meaning it can be powered by the fan controller, if I decided to do that. The maximum flow rate is about 25% less than the D5 at max speed, but I bought this for the head pressure. So when the cabinet is built we'll see how well it'll perform.

Next step, though, was Home Depot to pick up some of their "random length" oak boards for making the cabinet.

Building the cabinet

I'll have plans available toward the end of the build log for those who may want to build one of your own. The design here is for three double-120mm radiators on one side with everything else on the other and one side of the cabinet being bottomless to allow for airflow while the cabinet still looks like a cabinet.

As I mentioned, initially I thought about making it out of red oak, until I looked again at the price of it and decided to go with yellow pine, which cuts the cost of the cabinet in half. I bought two 18"x48" laminated pine panels along with a 6' length of 1"x3" pine. Nominal width is 17.25", which puts it a little wider than what I was aiming for. This would provide <em>most</em> of what I'd need to build the cabinet.

I started by ripping 18" off one of the panels to give the 30" long tabletop. The second panel was ripped in half to make two 24"x18" pieces for the sides panels. This will allow for a 6" clearance under the cabinet for airflow into the bottomless side of the cabinet. I'm sure my neighbors didn't hugely appreciate the table saw (read a review of it here), but I can say that at least it's not nearly as ear-piercing of a sound as my circular saw. And easier to clean up.

Though ripping the piece in half down to two approximate 24" side panels required a little ingenuity, and a little cleanup of the cut afterward to ensure the boards were about as square as I could make them. The table saw has a maximum rip fence distance of only 18", which was good for the first cut, but not so much for the second. Basically the clean-up involved clamping an adjustable T-square I bought at Harbor Freight at the right spot to guide my circular saw, then using an aluminum compound square from Home Depot for the little bit at the end the T-square blocked.

I initially thought of using dowels for this as well, but opted instead for #10 wood screws. To avoid problems, I wanted to also drill pilot holes in the end grain of the side panels, and #10 screws use a ⅛" pilot hole, the smallest size supported by my V-Drillguide. Through holes on the panels were 3/16", which I would later discover is too skinny for the #10 screws to glide through cleanly. Not a problem as it means I can start the screws to line up and glue the panels before driving them home.

The holes are spaced every 3.5". The downside was I didn't have a self-centering dowel jig that could go down to 1/8" -- I don't even think anyone makes one -- so I had to improvise a little.

In cutting the bottom for the cabinet, I hit a little snag. I initially tried to cut a hole about 1" in from the front and side and 3" in from the back. Unfortunately in cutting out the hole using a jigsaw, I forgot to account for the weight of the board in the middle and when it got loose enough, it took out much of the side of the board. So instead of having a clean hole, I have a U. Oh well. It'll still look good either way, and there will be plenty of support in the 3" section for the radiators, especially since the radiators the radiators will be supported using brackets at the top and bottom.

So while waiting for the brackets to arrive from Performance-PCs, I tried to get the cabinet pieces to a point where all that would be left to drill were the holes for everything to be mounted. This also meant needing to figure out how and where to mount the pump and reservoir. My back made sure to complain later from being bent over the boards cutting and sanding.

Radiator brackets

Before finishing the cabinet, structurally at least, I needed to get the XS-PC brackets in hand and mounted to a radiator for a visual test for marking the screws. I surmised looking at the brackets that the screws would be best spaced about where the fan screws are, which would put them about 225mm apart, but as much as 240mm apart. The radiators themselves will be positioned such that the fans are centered on the "ledge".

To mount the brackets, I drilled holes for 1" long #8 countersunk machine screws. I needed the holes on the cabinet top and bottom. Then I turned my attention to drilling the holes for the bulkhead fittings. Here is where things got a little interesting.

With how the bracket is positioned and using an AlphaCool 90-degree rotary fitting, the intake for the radiator measured to about 1.75" from the back. Measuring using diagrams for the bracket and radiator puts the center of the fitting at about 1.5" up from the bottom (1.54" when converting mm to in, so close enough). So I measured and drilled the side panel for that.

Only there were a couple minor hitches I ran into. I initially tried to drill with a 3/4" spade bit, thinking that would be wide enough. Not quite. If I knew that was going to be a concern, I would've picked up a set of spade bits from Home Depot or Harbor Freight since I only have a 1/2", 3/4", and 1" spade bit -- something I'll likely do in the near future anyway.

But I did have step bits, one that went to 7/8". That allowed for the bulkhead fitting to sit, but not completely the way I wanted. It was just barely long enough, so I needed to sink it. That's where another step bit came in handy, drilling a 1" sink around the outside, allowing the fitting to sit flush with with the side panel.

The fitting's position should allow the return line to go straight to the bottom radiator. Unfortunately I was not going to know until the cabinet is put together whether that is how it turned out.

With the bulkhead fitting set, though, I didn't turn my attention to the second bulkhead fitting. Not immediately at least. I planned to place it immediately above the return fitting, though. But first I'd need to mount the pump and reservoir to know where it would go.

Knowing where the bulkhead fitting would sit made measuring for most of that significantly easier. I chose to line the reservoir up so it would be centered along the same line as the bulkhead fittings. It would just make it much easier and leave the pump as the only complication.

And my thanks to Koolance and Bitspower for accurate drawings on their website. And thanks to my primary and secondary school teachers for the math education I received. No joke on that either. Let me show you what I mean.

I was able to determine from Bitspower's website that the reservoir mount I'm using was designed to fit onto an 80mm fan mount. And when mounts are designed to fit onto a standard layout such as that, it always makes measuring layouts so much easier.

Koolance's diagram of the pump and 90-degree fitting allowed me to determine how far off the center line I'd be mounting the pump: about 40mm from the center of the bulkhead to the center of the pump, and shy of 1/2" from the bulkhead center to the center on the closest mounting screws.

To figure out where on the board I'd be mounting it, I needed to know some details about the two EK 45-degree fittings I'd be using to connect the pump to the reservoir. Except there wasn't any documentation available showing any kind of dimensions on the fittings, a situation that EK should rectify since everyone else from what I've seen makes such diagrams available. The diagrams are invaluable when making a custom project like this one. So I had no choice but to eyeball it.

Once I had the pump mounted, I needed some ingenuity to figure out where to put the bulkhead fitting. Sure I could've just tried to line my square up under it, but I wanted to be a little more precise. From what I could gather, the pump top is about 18mm long, and the G1/4" port appears to be middle-aligned, putting the center about 9mm in from the edge. This means that from the center of the lower screw hole to the center of the fitting port is about 46mm.

So I measured and drilled, following a procedure similar to what I did for the first bulkhead fitting.

So it should be clear how this is going to run. The pump outlet is going right to the bulkhead fitting that will feed to the system. And the inlet to the box will go straight to the radiators. And the radiator will feed straight to the reservoir.

Unfortunately, though, this is much too close to make use of hard tubing fittings. So instead I'm going to use a Swiftech SLI fitting. I just need to get one from my local Micro Center when they're next open.

The only question really still up in the air is how I'm going to power all of this. Current plan is to use a small FlexATX power supply I have laying around, provided it is quiet enough for the job. We'll see when I flush the radiators and blocks later in the week. And I have an Add2PSU for synchronizing it with the main system, if I decide to do that.

But what you see above isn't how the pump is going to be mounted in the final configuration, as I plan to use some 00 rubber washers for vibration isolation plus a couple 1/4" nylon spacers. That should get the pump's inlet lined up with the reservoir, and I'll use PETG tubing as a short channel between them. But the 1" screws I have are too short for that. And given a lot of this was all done on Christmas Eve night, I couldn't just run to Home Depot to get the screws I needed. That will have to wait till Sunday after all of the Christmas festivities with mine and my wife's families.

So until then, Merry Christmas.

[brandishwar]

So the build continues...

The day after Christmas, my wife and I drove out into Kansas to visit family. Given the weather forecast for the next day compared to the weather while we were driving back, I decided to run to Micro Center instead of going directly home. This was to pick up the SLI fitting plus a couple other things to finish the build.

When I got home, after unloading from Christmas, I took the SLI fitting and put it into place. This basically demonstrated conclusively that all of my measurements regarding the pump were spot in.

With almost every fit confirmed, the only other thing to do was glue and screw the cabinet together. The glue is Titebond No-Run No-Drip, which is perfect for end grain applications such as what I'm doing here, but requires that I leave everything for about 24 hours to cure. Since I was doing this on a Sunday afternoon, that would mean I couldn't touch it till I got home from work the next day. Even though I also screwed the boards into the end, I decided to leave it all clamped overnight.

One thing I will say about this: nothing really prepares you for assembling it all together. When everything you're seeing in your head or on paper actually meets the wood to demonstrate whether your idea has merit. So seeing this all come together seeing everything line up... it was such a relief to see it all work out.

Revisiting the pump

The noise level on the pump was not promising. The initial testing with the pump doing a 2-gallon push flush on the new radiator showed the noise to be concerning. I'm not sure if it was noise from air -- wherein if I had it mounted in some fashion and let all air bleed the noise would've died away -- or just the noise to expect from the pump.

I preemptively purchased a DDC pump from my local Micro Center -- specifically the Swiftech MCP35X. They also had the MCP355, but I chose the MCP35X to hopefully retain fitting options. We'll see if I will need it. I won't know until I have at least the radiators mounted so I can do an isolated leak test on just those.

Swapping the pump isn't out of the question -- though it'll be interesting given the cabinet is already prepared for the PMP-420. So if I can adapt for the noise, that'll be great. Perhaps it'll run quiet when the pump is mounted to the cabinet on vibration isolation with it pushing fluid for a while.

The power supply I was considering, however, will work just fine. It's just a matter now of where to mount it.

Mounting and joining the radiators

You know the old saying of "measure twice, cut once". Well sometimes it doesn't keep you getting something wrong. In my case it was the holes in the table top for holding the radiator.

So after leaving the cabinet clamped and upside down all night, I took off the clamps.... and it promptly fell apart. Just kidding... I re-measured for the holes on the top and drilled and counter-sunk new holes. Then I mounted the bottom radiator and the top brackets.

One of the radiators would be hanging from the cabinet top directly above and -- hopefully -- directly lined up with the one mounted to the floor. The one in the middle would require a little ingenuity in the form of 1/8" aluminum from Home Depot and a jigsaw. To figure out where to hang it, I needed to do a little math. Feel free to check this if you think it's wrong. I based these numbers off the diagrams for the XS-PC bracket and AlphaCool radiator.

The AlphaCool radiator is 124mm wide, with a 9.5mm distance between the center of the fan screw and the edge. Hanging on the bracket gives a total radiator height of about 132mm for the top and bottom radiators, 264mm between. The third radiator is out in the open for another 124mm. This gives a total radiator width of 388mm. The gap between the top and bottom boards on the cabinet is 450mm.

This leaves a total space of 62mm around the middle radiator, or 31mm between the radiators. Adding 19mm to this means it will be 50mm between the screws. This is just shy of 2". The widest piece of aluminum that Home Depot sells in any store near me is 2".

But all is not lost. I ordered a couple pieces of 3" wide 6061 aluminum from Grainger (item no. 2EYU9) for pickup... the day before we leave for delivery. In the mean time I can use the 2"x36" aluminum I did buy to make a temporary solution for holding the radiator.

I cut a 2-1/2" long piece of aluminum and drilled four 1/8" holes that formed a rectangle about 2" long and 15mm wide. Two of these should be enough to hold the center radiator and allow me to finish the loop.

My jigsaw is actually the MATRIX Jigsaw attachment from Black and Decker for the 20V MATRIX cordless drill. The blades are I used are Bosch T121AF3 jigsaw blades. This combination could make it through the 1/8" aluminum. For drilling the holes, I used a 1/8" cobalt drill bit. Make sure to keep your vacuum handy for the aluminum shards.

I wanted to get as far as I could before I had no choice but to take down Beta Orionis to get the other two radiators from that system. I mentioned in the first segment of the build log that the chassis for this build will be the Corsair 750D that I used for β Ori. I think what I failed to mention was that the other two double-120mm radiators, along with some of the fittings, would also be coming out of that build.

But first, I started laying some pipe in the cabinet. And the first run was mildly nerve-wracking since it, again, would tell me if I was spot-on with my measurements and mathematics. The first tubing ran from the cabinet's inlet to the bottom radiator. It's a straight run, straight line to the radiator from the bulkhead fitting. Or that's how I measured it.

And that's how it turned out. About 14-1/2". I love it when things come together like this. This also tells me that the line going from the top radiator back to the reservoir should work as I expect. The next piece of pipe was from the reservoir to the pump. Straight line again, but this one much shorter.

By the way, one little trick about cutting PETG: use the same kind of cutters you would for soft tubing. You'll get a very clean cut.

With this piece in place, it was time.

[brandishwar]

The project is now in the client's hands. Let's go over the last steps and the final details.

First, the PMP-420 was a no-go. I replaced it with a DDC pump that is still noisy, but not nearly as much as the PMP-420. I still want to figure out how to quiet the pump, so I will be in touch with the client about what to do for the noise. But given the client didn't mention the pump while standing inches from the cabinet, I'm going to presume it's not going to be an issue for him. If it is, then we can discuss it.

But again, let's talk last steps. Last I left I said that I got far enough with the system that "it was time". What I meant on that was tearing down Beta Orionis and stripping everything out of it. And I do mean everything. Alice decided to help. I actually didn't know that a 7 month-old kitten could fit into the triple 5-1/4" drive bays...







Once everything was out of the case, and Alice gave her approval, I fetched the new power supply. Corsair RM-1000x. I selected this to give the client a very, very good power supply that was going to also be quiet. The pump itself was loud enough. The PSU didn't need to contribute to it at all. I ran the cabling I knew would be necessary and then fetched the rest of the system from the blue case.









As you can see, I initially mounted the AIO in place along with the graphics card with its air cooler. Since I knew it wouldn't take long, I wanted to wait before putting on the blocks till either there was nothing left to do, or my back was going to require it. Instead I turned my attention to flushing and mounting the radiators.

And the aluminum piece I mentioned earlier worked like a charm. Remember, though, how I said I had a longer strip of aluminum on order from Grainger? Yeah it didn't arrive in time... So instead I had no choice but to stick with the piece of aluminum because of the looming delivery date. But it worked exactly as expected and, again, validated all my measurements and math.



This gave me the chance to cut the rest of the tubing, validating the rest of my math and giving me a small loop I could leak test. While it was leak testing, I turned my attention to the second aluminum bracket and got it into place. With that, I turned my attention to the CPU and graphics card and tubing up the chassis.



Then it was a matter of cutting the tubing for the quick disconnects and connecting the computer system to its cooling system for a full leak test and some initial temperature testing with straight distilled water. The temperatures were quite phenomenal. Recall that there is an i7-5820k and a GTX 980 connected to 6x120mm of radiator capacity. This is the GPU temperature after Heaven had been running for at least 20 minutes. And no, I'm not making this up.





Unsurprisingly the CPU ran hotter when I had Folding@Home's StressCPU utility running on all 12 threads, but still struggled to get even into the 40s. The temperatures were a little higher doing the final stress testing in my hotel room in the client's home town, but not by much.

Retrospective

There are so many things about this project that could've gone better to allow more time for the build. For starters, once I discovered that there weren't really any pre-built cabinet options, I should've started on the cabinet sooner. Had I done that, the rest of the system may not have been rushed and all of the planned items would likely have been accomplished. As in I was also looking at kitchen cabinet options at Home Depot seeing if there was something I could adapt instead of looking at the lumber section to see what materials I could meld into a cabinet...

What helped save time on the system, though, once I had dimensions in mind was going with the laminated panels from Home Depot. It saved money as well. So much so that I'm going to be using those panels to make a couple rack cabinets, including one that will replace the rack I built for the Colony West project. You build, you learn. Again, though, I should've started building the cabinet sooner than I did, though the delay with the USPS cost me a day as well.

But easily this shows that cooling a system in this fashion is doable, provided you're willing to expend a little effort. If you can modify a pre-fab cabinet to mount the radiators, then that will probably be the better option. Or you can build something like I did here.

In hindsight, a more impressive demonstration would've been using the Zalman Z12 that's been featured on this blog before. It has virtually no radiator capacity, so seeing that chassis connected up to the cabinet would show quite well that you don't need a large fucking case to get good radiator capacity.

Instead all you need is just a little ingenuity and imagination.

[Mnpctech]

Sweet project build and choice of components, and you're right about PETG, the ratchet gun PETG cutter is worth investment to people who think they will do several builds with PETG and want to save time, but a jewelers hack saw works great too, I've probably cut several hundred feet of PETG tubing over the past 6 months with this gun