Overkill Watercool MORA Build - Go Big or Go Home

[mattlach]

So, I’ve been working on a LONG and SLOW build now for a while now. I started documenting this build on two other forums I frequent, but during the very extended process of working on this build, I have gradually spent more time over here on the Level1techs forums, so I decided to add this location to my build log as well.

So, the first several posts will be playing catch up with the build log on the other forums. Once I have caught up, all new updates will be posted to all three forums (Hardforums, TheFPSReview Forums and here) at the same time.

First, some background:

The last build involved customizing fan mounts to fit two 480 and one 420 radiator in a Corsair 1000D.

It was a pain in the ass, and a lot of stuff went wrong (only a minority of which were my fault).

In the end I wound up with this:

Five years later, I’m pretty much still running the same build, except, I had to make some minor modifications to fit two interim GPU upgrades in there (first 6900xt, then 4090)

Except even with the huge radiator capacity, and 22 radiator fans + 2 case fans, it was still too loud for my preferences at full gaming load (probably due to the combination of the Threadripper and the 4090 pulling like a kilowatt at the wall when at full load) so I decided to get some active fiber-optic display and USB cables and just move the whole thing across the house and stick it next to my server rack, to finally have some peace and quiet in my office.

Which brings us to the starting point for this project.

[mattlach]

The reason for Build 3.0
This project kicks off for a few reasons:

1.) S.T.A.L.K.E.R. 2

I have been waiting for S.T.A.L.K.E.R. 2 for damn near 15 years, and I want to play it in all its glory, without “fake pixels” or “fake frames” at 4k Ultra.

The 4090 can almost pull this off, but really requires at least a little bit of DLSS scaling for it to be ideal. If this were the only problem I had, I would probably be playing STALKER2 right now instead of scheming this build.

Unfortunately, in STALKER2 the trusty old Threadripper 3960x has finally met its match. I don’t know what the deal with this game is, if they really just messed up the code and it is horribly inefficient, or if they are mining Ethereum in the background during gameplay, but it is extremely hard on the CPU. I’ve been thinking about upgrading the Threadripper for some time now, but Ive been frustrated. New Threadrippers are very expensive and have moved away from HEDT and into real workstation territory, meaning they perform like shit in games. Consumer platforms - however - have way too few PCIe lanes for my desire. Hoping and searching for some compromise solution I was happy with has delayed the upgrade.

With the launch of STALKER2 - however - there is no more ignoring the fact that this system can no longer keep up. I tested the prologue of the game, and was limited to framerates in the 50’s. Enabling frame generation made it look smooth, and gave me 100+fps framerates, but input lag was horrendous. The game was almost playable, but for a game I have waited this long for “almost playable” is not going to be good enough. It was time for an upgrade

2.) HEDT is dead. It’s consumer vs workstation now.
My build philosophy (at least since 2011, if not before) has been to build a “no compromises” system that does everything well. As much per thread performance as possible. Extra RAM channels if possible, as many PCIe lanes as I can get my hands on, more than typical desktop cores if possible. The problem is that since there is no more HEDT, this is no longer possible.

If I go for more RAM channels, gobs of PCIe lanes, etc. this puts me in workstation territory, and the per core performance in games is atrocious on those boards.

Furthermore, keeping up with a system that can sort of handle the latest titles in a workstation board, gets expensive very fast due to frequent upgrades. My Threadripper is still fine for my workstation purposes. If it werent for games, there would be no need to upgrade it at all with new workstation parts.

I’ve always hated the concept of a “gaming PC” but I slowly came to the inevitable conclusion that I am probably going to need two builds. One workstation build that uses expensive enterprise-like parts, but is infrequently upgraded, and one dedicated PC for games that uses comparably cheap desktop parts, so if I need to upgrade frequently, the cost is not a big deal.

So I guess that is what I am doing. Two builds instead of one.

3.) The desktop is already on the floor next to the rack, and I have lots of free space in the rack. Why not - you know - rackmount it?
I don’t really think I need to add more to this one. When this project is complete, I will have two builds, both in my rack.

4.) Watercool MORA radiators are awesome.

So, being a huge fan of I have been eying Watercool MORA radiators for years now, but never gotten around to actually buying one and doing anything with it.

Rackmountable cases are cool, but I am already frustrated with the cooling capacity of my “overkill” loop in my Corsair 1000D, and rackmountable cases generally have even less space in them for cooling, so cooling was going to have to be external.

Then it struck me, what if I mount one or two MORA radiators to the side panel of the rack?
In my head I had a theoretical build with two MORA 3 radiators, each with 9 140mm fans in them, hanging on the side of the rack.

But then MORA 3 parts started going out of stock on the Watercool webpage, and I started getting worried. I had heard that the watercooling industry had been hit hard last year (just look at what EK is going through, though that is compounded by mismanagement). I was concerned the MORA line was disappearing on me. To my relief it turns out they were just moving to a new generation, the MORA 4. And with this generation they introduced a new larger size, the MORA IV 600, which is HUGE.

I didn’t realize at first. I saw the mounts for 9 fans, and figured it was the same size as before.

Then I noticed that those 9 fans were all 200mm fans. They are huge.

Next up, some more detailed build plans.

[mattlach]

3.0 Build plans

So, the plan is for there to be a dedicated game system and a dedicated workstation, both in the rack, with external cooling in the form of two MORA IV 600 radiators hanging from the side of the rack.

The Workstation (Phase 1)

This system will take the motherboard, RAM and most of the drives from the current Desktop Threadripper build.

The 4090 GPU will be wasted in this workstation, so I will buy a lower end GPU and install it without watercooling. Maybe a Radeon RX 7600 (non-XT). Not sure yet.

In this build, only the CPU will be watercooled.

The current plan is for them to be built into a Sliger CX4150a 4U case:

Some holes will be cut in the back to allow the watercooling tubes to exit.

These will be connected using 3/4" OD, 1/2" ID soft tubing and Koolance’s QD4 QDC’s to allow for diconnecting reconnecting, and future upgrades, etc:

The Game System (Phase 2)

The game system will be all new parts. When I first started planning this build it was going to keep the 4090, and get a new consumer motherboard, CPU and RAM, but at this point, the 5xxx series is right around the corner, and the 4090 just isn’t keeping up with S.T.A.L.K.E.R 2 so I guess I will just go with a new GPU as well.

GPU Blocks usually have their ports raised up above the video card, which means a 4U case is probably not going to cut it. I may wind up going with one of Silverstone’s 5U or 6U cases just to fit the GPU, but their pricing seems rather excessive for what you get. I am continuing to do research on what case to use here, and I am going to need to get some measurements to finalize my determination on just how tall of a case I need, or if I cam maybe use some sort of adapter to mount the GPU on its side.

Just as with the Workstation build however, it will be connected using 3/4" OD, 1/2" ID tubing running out to the MORA radiators on the outside, and will have Koolance QD4 QDC’s. This system will have a water block on both the CPU and the GPU

The Plan

As is suggested above, this will go in phases. In part because I need some of the parts (pumps, reservoir, etc.) from the current build to make the full cooling system function, and in part because I need my workstation for work, and cannot accept much downtime.

I am going to start by installing half the radiator capacity (one MORA IV 600) and routing it up to just the workstation build. Once this is done, I am going to install the second MORA radiator, and only then build the game system.

The plan is for both radiators to cool both systems (shared radiator capacity FTW) so I will be utilizing QDC’s in key places to be able to keep it modular and expand the loop once I am ready.

Bear with me on this build. I am going to do it little by little, so it may take some time. I will post updated posts as I go along.

[mattlach]

Planned Loop Routing

I know this is still somewhat controversial in PC cooling circles, but I did it with my current build and it has worked really well, so I am going to do it again, just on steroids: The final loop will consist of 4 separate smaller loops that all mix in two reservoirs.

I really would have preferred doing this in only 3 loops, but I just could not find a reservoir with enough inlets and outlets to handle this, so I will instead be using two reservoirs (EK-RES X4 250)

I chose this EK res, not because I have any love for EK, but because I already happen to have one in my system, and didn’t want to buy more shit than necessary here. It also happens to have a large number of ports, which is what I need. This reservoir is now discontinued, but I was able to find a second one, so I have what I need.

Loop1:
This will be a simple loop just handling one of the two radiators, with a single D5 pump:

Reservoir1 → Single D5 Pump → MORA1 → Reservoir2

Loop 2:
This loop will be very similar to the first one:

Reservoir2 → Single D5 Pump → MORA2 → Reservoir1

Loop3:
This loop will cool the workstation:

Reservoir1 → Single D5 Pump → Workstation CPU block (Existing Heatkiller IV 4 block) → Reservoir1.

Loop4:
This loop will cool the game system:

Reservoir2 → Dual D5 Pumps → Game system GPU block → Game System CPU block → Reservoir2

The thought process here is that since the coolant mixes in the reservoirs, and continuously moves back and forth between the two reservoirs, both systems can take advantage of the outrageously massive radiator capacity of both MORA radiators, when the other system is idle or off.

The Game system will be at the very bottom of the rack. In case it leaks, I don’t want coolant running down on anything else more important than games, and destroying it. The workstation system will be right above the Game system, followed by a couple of U’s “safety gap” after which the UPS:es will go on top of that, and finally the more important network hardware and NAS upstream from them, so that no water ever gets on those no matter what happens (unless I have some sort of freaking fountain)

Fan and pump control will happen using 2 Aquaero 6 LT’s.

One Aquaero will be dedicated to controlling the speed of the two pumps cycling the radiators and their 18 200mm fans.

The other Aquaero will handle the speed of the dual pumps of the game system.

The workstation pump - since it only has one block - is a bit simpler, and will be controlled by the motherboard fan header based on CPU temp. (if this doesn’t work well, I do have a third Aquaero kicking around here I can use, but I am trying to minimize complexity in an already very complex system.

The Aquaero’s, pumps and radiator fans will have their own PSU. I will use some sort of relay rigged up such that it is on, whenever either of the systems are powered on. I was going to use one of those mining rig PSU controllers you see around, but I have been unable to find any designed to take input from more than one master system. I did buy one of those Arduino/Raspberry Pi oriented wall outlet relays a few years back for another project that never happened, so I might just use that instead.

Either way, I am going to have to do some custom wiring to make sure I don’t have backflow from one system to the other. It will include joining 12v power and ground from both systems, but through a diode so nothing flows the wrong way.

A company named Roadmaster makes a diode called the 790 Hy-Power that has two inputs and one output, and is thus perfect for what I am trying to do. It is intended for towing cars behind an RV, while linking their lighting systems, but without having power from one going to the other.

I think I can just put one of these on a 12V wire from a 4-pin “molex” (not really a molex connector, but everyone calls them that, so whatever) from each system, and then just tie their grounds together, as those are supposed to be at the same (0) voltage anyway. The combined 12v signal can then be used to control the relay that turns the cooling loop on and off.

[mattlach]

Arrival of the Radiators:

These came on December 23rd. I originally ordered them in early November, but the wall mounts were back-ordered (and there was no mention of this on the Watercool store page). Eventually they transited the Atlantic and arrived at my house.

The holidays being a busy season, I didn’t really have time to open them until today.

I knew they were huge, but DAMN, reading dimensions does not prepare you for seeing them in person:

16oz “tallboy” and 12oz Diet Coke (sorry no banana) for scale.

These things pretty much take over my office…

[mattlach]

Some Prepwork

I spent some time looking at the rack a few days ago, and noticed that the side panel I was planning on hanging everything from is really only held on by some plastic latch handles:

Those two little black handles, and a wimpy lock (for which I do not have the key) are what secure the whole thing in place.

I put some test weight on it (me, hanging from the side of the thing), and damn, the panel started to pull away from the rest of the rack. So I needed to reinforce that puppy before I start hanging heavy water filled radiators and reservoirs from it.

Looking at the rack from the inside I noticed that the back of the main rails (where you screw in the sliding rackount kits) has 1/4" screw holes about every half inch running top to bottom on both the front and rear of the unit.

I decided to drill matching holes on the panel itself, and install some machine screws through the holes with bolts on the back. I figured that should secure it up nicely.

I figured I’d pout one hole front and back up top, in the middle and down at the bottom, 6 bolts in total. That ought to do it.

I noticed the rack shelves were in the way up top and on the bottom, so they would need to come out for me to gain access.

These things are heavier than they look:

They are made of 1/8" thick sheet metal, with reinforcement in many places. Real heavy.

They came in the rack, and probably date back to the 90’s. I haven’t seen modern rack shelves (this is the one and only rack I’ve had the pleasure of working with) but I am going to go out on a limb and suggest “they probably don’t make them like this anymore”.

After a quick trip to home depot to buy the fasteners, bolts (and some lock washers for good measure) I started the drilling. I bought a new kit of drill bits when I was there as I did a massive home improvement project this summer, and have either lost or burnt out most of my previous drill bits.

I wound up with some Milwaukee “Shockwave Titanium” bits, because they were on sale. (I have no tool brand loyalties). They claim they stay sharper longer, and I was actually pretty impressed with the first hole. I applied a little cutting lubricant to the bit and started drilling, and it went through the sheet metal like butter.

I’m not sure about the “stays sharper longer” claim though, as already by the second hole it took some more force. By the last, I had to put some muscle into it.

But all the holes are drilled, and bolts installed, and the panel is now very secure.

Because I was worried about the bolts interfering with the shelves (or other things) when I put them back in, I flush cut the back of the screws with the tightened down nut using my oscillating tool and an “extreme titanium metal/whatever” blade.

In this process I learned that oscillating tools can loosen tightened bolts (even with a lock-washer in place) so once done, I re-tightened them all.

Here are the top four screws in place. (the bottom two are hiding behind the Corsair 1000D)

That’s all I got to today.

I am going to leave the shelves out for now, as if I put them back in, they are probably going to get in the way, when I go to mount the radiator wall mount brackets.

Unfortunately, in one of my many trips behind the rack, I must have tripped over and broken one of my 65ft display port cables connected to the back of the Corsair 1000D

Now I am going to be without one of my side screens until I can get and route a replacement.

That’s a bummer, especially since they are like $90 each, and a pain in the ass to route across the house to my office… Oh well. Shit happens. I’m sure this won’t be the last mishap before this project is done, and if it is the worst, then I will have been pretty damn lucky!

[mattlach]

So, I decided to take out one of the radiators and start to flush it today.

When I did? Bummer.

The first one I opened had quite a dent on it. Not quite sure how transatlantic RMA works, especially on something this big, but I guess I am about to find out.

I’m not sure how easily that cover might separate from the rest of the unit, but maybe if it is fine on the inside, they can just ship me that replacement piece, and I can install it.

The box was not terribly damaged, so I am not even sure how it happened. There were those nylon(I think) straps around it when it arrived, and I wonder if the straps maybe did that.

Otherwise if that was an inspection escape, that’s a pretty bad one.

I’m going to reach out to their customer service and I will let you guys know what they say.

[mattlach]

Meanwhile, I moved on to the next radiator. I took it out of the packaging and it looks fine.

In another thread cpufrost recommended a flushing method I had never thought of, that sure might save on water and effort.

Essentially, get a couple of household water filters, and barb tubing adapters for them, and put them in a loop and flush with distilled water.

First in line is a mesh 50 micron sediment filter. This ought to capture the big stuff.

Next in line is a 0.5 micron water quality filter. It does a lot of stuff I don’t care about (active carbon taste improvement, etc.) but it should also filter down microbes and finer particulate.

Last, - for good measure - I attached a filter that claims to go down to 0.1 micron. I figured, why the hell not.

A previous owner of our house built this in-law apartment in the basement where there is a spare (and probably illegal) kitchen. I set up my flush loop in this kitchen where I am just going to let it run for a while.

Filling this beast takes ~3L of water according to Watercool. So we are talking almost a gallon. Then add in the filter housings getting in the way, and the water they take to fill, and this was a tricky loop to get all the air out of. It was also a tricky loop to find good adapter fittings for that didn’t leak.

Much cursing and sopping up distilled water with rags later, I was able to get it tight, and flushing the radiator.

I’m just going to let it run here for a while. I am going slow on this project, especially since next week is audit week at work (who doesn’t love a good ISO notified body audit?) so that is the priority right now, but I figured it wouldn’t take much time to hook up the flush loop and get it going (I was wrong) and then the radiator can flush while we are being audited.

[mattlach]

We had an audit at work this week, which pretty much sucked up all of my time, but having passed with with flying colors by midday Thursday, I decided to take Friday off, and while I mostly caught up on sleep, I had a little time to make progress on the project.

Having had to lift and move these radiators around when I discovered the dent, and started the flush of the un-dented one, I am once again concerned, that the little I did to secure the rack side panel may not be enough to hold the weight of these things, especially when full of coolant.

So, it is belt and suspenders time.

The wall mount brackets look like this.

One quarter inch sized hole at the top and bottom, intended to screw the bracket into the wall, and then some prongs that screw into the radiator hang from the tear-drop shaped openings.

My original intent was just to drill holes in the panel, and screw these brackets straight to the sheet metal, but I am thinking maybe I should reinforce it more.

This is what the back of that panel looks like. Note those holes that I used to secure the panel.

I was thinking if I got some metal flat stock, and I mount it across between the front and back through those holes, that will give me a solid frame to anchor the radiators to.

Anyone who notes that this would result in it being offset from the back of the panel, and who has taken an Engineering class in Statics will note the problem with that though.

Allow me to illustrate:

This is what my force components are going to look like if I bolt it directly to the sheet metal.

There is - of course - a downward force component at each hole, but the top hole has an outward force component, whereas the bottom hole has an inward force component.

This is not terrible, as long as the panel can carry the weight, which I am not sure it can.

If I instead install the two horizontal bars and bolt into those, I get the illustration below:

The positive part of this is that we mostly remove the F+x force component from the sheet metal up top, as it is pulling on the bar instead of the sheet metal.

The screw that goes through the bar - however - is likely going to tilet downward with the weight, and then the downward force components (F-y) both top and bottom are still going to be pushing on the sheet metal.

So, it might help a little but probably not a whole lot.

Then I though, what if I add a second nut to the other side of the steel bars.

If it is tight, and the bars are rigid enough, it ought keep the screws mostly straight, and offload at least some of that that force on the sheet metal.

So I wandered the metal flat/bar stock aisle at the Home Depot and found some some 36" long 1/8" thick C-channel stock:

The C-channel gives it MUCH stronger torsional rigidity than the flat stock I had originally intended to use, so it winds up being pretty much perfect for my application.

(I had initially considered some rectangular tube-stock, but everything I could find was too thick, and wouldn’t fit in the rack how I intended.)

36" is a little too long for me though.

I must have something like 15 measuring tapes at this point, but for whatever reason I couldn’t find any of them, but I did find my laser measuring device:

From front to back I measured about 30"
Nut that is from where the sheet metal bends, not from where the holes are.

I figured I could take of ~2" on either side and still have them be long enough, but also short enough to make them easier to finagle into place.

This is where I miss having my nice large oversized two-car garage from my old house. I no longer have a garage. I have a shed, where I have set up a workbench with a grinding wheel, a drill press and a scroll saw, but I have no more space on it. So I was going to have to do this one outside in the cold.

But this is the warmest day in over a week, and the icy cold that is hitting further south is coming our way early next week, so it was pretty much now or never.

Besides, it wasn’t that cold. Only ~34°F and sunny, so it wasn’t even really jacket weather. The fleece sweatshirt did the trick.

I hadn’t used the trusty Harbor Freight cut-off wheel in a while, so first, I decided to verify everything was straight.

Some very loud noises, and flying sparks later, the cuts were good and straight, but I wound up with some serious burrs.

I decided to take the burrs off, and dull off those sharp edges with my Dremel, as I know if I don’t I am going to cut myself on them later…

They wound up maybe not super-clean, but certainly fit for purpose.

Next steps are going to be to drill mounting holes in them so I can mount them inside the rack.

I need to do some thinking about how I mitigate rust on them after that. I could clean them and blast them with some Rustoleum, or maybe rub them with some light oil. Or maybe clean them off and do a gun-style bluing operation, but then I am going to need to find something large enough to quench them in, that can handle hot steel…

I’d value any suggestions here.

Or maybe I am over-thinking it. They will be going indoors, so serious rusting will likely not be a problem, but it does get a bit humid in my basement…

[mattlach]

@mattlach

So, I decided to take out one of the radiators and start to flush it today.

When I did? Bummer.

The first one I opened had quite a dent on it. Not quite sure how transatlantic RMA works, especially on something this big, but I guess I am about to find out.

I’m not sure how easily that cover might separate from the rest of the unit, but maybe if it is fine on the inside, they can just ship me that replacement piece, and I can install it.

The box was not terribly damaged, so I am not even sure how it happened. There were those nylon(I think) straps around it when it arrived, and I wonder if the straps maybe did that.

Otherwise if that was an inspection escape, that’s a pretty bad one.

I’m going to reach out to their customer service and I will let you guys know what they say.

Oh, and to follow up on the dented radiator.

Watercool demonstrated some excellent customer service. They replied to my email as soon as it was business hours in Germany, and asked if it was OK if they shipped me a new cover for me to install myself. I said that would be ideal, as shipping large heavy things internationally is costly, as long as there is no internal damage that could cause a leak.

I also asked if there were any tricks to install it.

That was on Monday. Same day, they replied with a couple of videos with install instructions, and by Wednesday I received notification that it had shipped. It arrived today, two days later.

I haven’t had a chance to open it yet, but the box for the replacement part has a huge gouge on it. I am hopeful that the content is not damaged.

It seems safe to say that DHL has some serious handling issues somewhere on route between Watercool in Germany and my house in the Northeast U.S, as of right now they are 0 for 2, at least from a box damage perspective.

I am way too tired to open and inspect it for damage right now, so I guess I’ll do that in the next day or two.

I am totally thrilled with Watercool’s customer service. They seem like a really standup company the way they handled this.

DHL? Not so much.

[mattlach]

Alrighty, since my last updates I have had some setbacks, and come up with some solutions, but first the positive news:

Despite DHL seemingly trying their hardest to derail my project, the replacement cover from Watercool arrived in one piece, due in no small part to the Watercool shipping department going absolutely nuts with packaging materials. (I now have shock absorbing packing materials to last a couple of years if I need to ship stuff )

Watercool went all out and provided me with video instructions on how to take the cover off:

MO RA

…and how to reinstall it:

MO RA 2

In order to get to the step where they show you how to pry off the cover, you have to first unscrew the side covers, and when I did this I heard something rattle and fall further into the radiator:

It was the little light diffuser for the MORA logo, if you use the optional LED lighting (which I won’t be)

I was able to poke it back in, and it seems to stay put with the tape they used.

This leads me to believe that DHL really manhandled my package. It look like it saw some serious forces in transit. Besides the bend, we have the logo fall off, and there are also some bent fins (presumably from the accessories box moving around inside the box) I carefully adjusted them back.

I have a set of fin combs from working on HVAC units, but none of the sizes I had seem to fit the pattern on the MORA. Or maybe that was because they used a serated pattern instead of the more common straight fins. SO I had to go in with a tweezer and manually straighten them. They never look quite the same again, but they should not be an issue from an air-flow perspective, and once installed, the fins will either be up against the side of the rack, or behind the fans and fan covers, so it’s not like they are visible anyway.

The first step of prying the cover off (after removing the two side covers) was a little tricky. It takes a good amount of force, at the same time you have to be careful to not hit the fins and bend them, but I got it done.

Here are some “naked” pics:

In the end I was able to reassemble it and it is looking good!

(I’m not even going to bother wiping down finger prints on these things until I am completely done)

[mattlach]

Also, before getting tot the setbacks, I have ordered the case I am going to use for the workstation.

Instead of the 4U Sliger CX4150a I mentioned in my plans, I decided to step it up a couple of sizes to the larger 4U Sliger CX4200a, entirely because upon reading the detailed specs, my existing Asus TRX40 ROG Zenith II Extreme Alpha is slightly too wide for the smaller model.

They look almost identical, but this one is not only deeper, but fits slightly larger motherboards.

For the gaming part of this build (which will come later, after I am done with the first half of the loop and have moved the workstation over to the rack, I am likely going to use the 3U Sliger CX3171a. It is a 3U variant of the same case, but this particular variant comes with a PCIe flexible riser and the option to install a GPU sideways.

Right now their website says it comes with a Gen4 riser, but I have been in touch with their sales, and they have told me a Gen5 riser will be available in a couple of days.

This overcomes the height limitation when installing a GPU. The 3U case is just barely tall enough to be able to insert a full height PCIe card, but most GPU’s have coolers that are taller than the board, and are thus impossible to install in a 3U case like this. Some GPU’s would fit in the straight up orientation in the 4U case, but most water blocks have the inlet and outlet ports stick up on top, making them even taller, and not even allowing them to fit in the 4U case. Since the 3U is the only variant they sell with the sideways GPU mount, that is the one I am going to have to get. It also winds up saving me 1U of rack space (not that I am out of rack space yet, but, you never know what the future holds)

I briefly considered using the 3U with the horizontal GPU mount for the workstation as well to save an additional “U”, but then I realized that this takes up almost all of the PCIe slot space, so I wouldn’t be able to use my PCIe slots, and I stick a bunch of stuff in those, so that was not an option.

The Sliger cases come in a series of festive colors if you are the festive type. I’m not really festive, so I ordered black for the workstation. I’ll probably go with black for the game machine too, when the time comes.

[mattlach]

Sliger offers to sell you Noctua NF-A12x25 PWM fans with your case, but given that this rack is out of earshot and I cant hear it anyway, I wanted something with a little bit more power. After some brief research (and this thread) I landed on the Silverstone FHS 120X. They are 38mm thick, they hit 4000rpm, and they move a lot of air with a surprisingly high amount of static pressure.

They are probably a little loud, even with the door shut to the server room, but that is on the other side of the house, and they will unlikely run at full speed all that often.

A nice touch is that they came with 8 fan screws each, a 120mm fan grill, as well as a breakout adapter in case the 0.96A the fan draws is too much for your motherboard fan headers. It splits off power to a SATA connector, but allows you to still listen for a PWM signal and report an RPM signal to the motherboard.

As mentioned, at full speed they are a little loud, but I think I can live with that

Here is the sound profile:

Silverstone FH120X Sound

At minimum speed (or at least minimum speed I could achieve with the old (Sunbeam?) rheostat fan bay controller I was using for testing, PWM usually allows for lower speeds that with voltage based fan control) they were very reasonable, and did not show up on the decibel meter above ambient noise.

Silverstone FH120X Noise (min speed)

(sorry for the several seconds of nothing in the beginning as I was slowly dialing up the rheostat to see where it started moving. I don’t feel like video editing right now to remove it. Ain’t nobody got time for that.)

Speaking of Sunbeam…

They can’t be the same company that made irons can they?

This lovely relic of the 3.5" floppy bay era may no longer be useful in a case, but it is great for fan testing.

[mattlach]

Alright, so to the setbacks.

Setback #1

This is a real facepalm one. Remember my fancy filtered radiator flushing system from several posts up?

@mattlach

Meanwhile, I moved on to the next radiator. I took it out of the packaging and it looks fine.

In another thread cpufrost recommended a flushing method I had never thought of, that sure might save on water and effort.

Essentially, get a couple of household water filters, and barb tubing adapters for them, and put them in a loop and flush with distilled water.

First in line is a mesh 50 micron sediment filter. This ought to capture the big stuff.

Next in line is a 0.5 micron water quality filter. It does a lot of stuff I don’t care about (active carbon taste improvement, etc.) but it should also filter down microbes and finer particulate.

Last, - for good measure - I attached a filter that claims to go down to 0.1 micron. I figured, why the hell not.

A previous owner of our house built this in-law apartment in the basement where there is a spare (and probably illegal) kitchen. I set up my flush loop in this kitchen where I am just going to let it run for a while.

Filling this beast takes ~3L of water according to Watercool. So we are talking almost a gallon. Then add in the filter housings getting in the way, and the water they take to fill, and this was a tricky loop to get all the air out of. It was also a tricky loop to find good adapter fittings for that didn’t leak.

Much cursing and sopping up distilled water with rags later, I was able to get it tight, and flushing the radiator.

I’m just going to let it run here for a while. I am going slow on this project, especially since next week is audit week at work (who doesn’t love a good ISO notified body audit?) so that is the priority right now, but I figured it wouldn’t take much time to hook up the flush loop and get it going (I was wrong) and then the radiator can flush while we are being audited.

Well, I didn’t actually check that I was getting any flow.

It turns out that a 0.5micron filter followed by a 0.1micron filter are just too much for a single D5 to handle pushing through. I was getting no flow at all through it.

I wound up removing the blue filter (0.1 micron) and now there is a little flow. Not as much as I would want for what I am calling a “flush” mind you, but there is flow. Hopefully this will be enough to get any leftover crap out of it. I’ll likely just make up for the rather limited flow rate, by just letting it sit a little longer.

Setback #2
Remember those C-channels I was going all amateur metal worker on?

@mattlach

So I wandered the metal flat/bar stock aisle at the Home Depot and found some some 36" long 1/8" thick C-channel stock:

The C-channel gives it MUCH stronger torsional rigidity than the flat stock I had originally intended to use, so it winds up being pretty much perfect for my application.

(I had initially considered some rectangular tube-stock, but everything I could find was too thick, and wouldn’t fit in the rack how I intended.)

36" is a little too long for me though.

I must have something like 15 measuring tapes at this point, but for whatever reason I couldn’t find any of them, but I did find my laser measuring device:

From front to back I measured about 30"
Nut that is from where the sheet metal bends, not from where the holes are.

I figured I could take of ~2" on either side and still have them be long enough, but also short enough to make them easier to finagle into place.

This is where I miss having my nice large oversized two-car garage from my old house. I no longer have a garage. I have a shed, where I have set up a workbench with a grinding wheel, a drill press and a scroll saw, but I have no more space on it. So I was going to have to do this one outside in the cold.

But this is the warmest day in over a week, and the icy cold that is hitting further south is coming our way early next week, so it was pretty much now or never.

Besides, it wasn’t that cold. Only ~34°F and sunny, so it wasn’t even really jacket weather. The fleece sweatshirt did the trick.

I hadn’t used the trusty Harbor Freight cut-off wheel in a while, so first, I decided to verify everything was straight.

Some very loud noises, and flying sparks later, the cuts were good and straight, but I wound up with some serious burrs.

I decided to take the burrs off, and dull off those sharp edges with my Dremel, as I know if I don’t I am going to cut myself on them later…

They wound up maybe not super-clean, but certainly fit for purpose.

Next steps are going to be to drill mounting holes in them so I can mount them inside the rack.

I need to do some thinking about how I mitigate rust on them after that. I could clean them and blast them with some Rustoleum, or maybe rub them with some light oil. Or maybe clean them off and do a gun-style bluing operation, but then I am going to need to find something large enough to quench them in, that can handle hot steel…

I’d value any suggestions here.

Or maybe I am over-thinking it. They will be going indoors, so serious rusting will likely not be a problem, but it does get a bit humid in my basement…

Turns out that’s not going to work at all.

It turns out the spacing between the two wall mount brackets winds up such that no matter what, I am, going to have to drill through the rails, not mount between them.

This means a little bit of wasted effort, and wasted money on flat stock, and I am going to have to drill through stuff I didn’t want to drill through, but it is not the end of the world.

Setback #3
It turns out in order to screw the wall mount brackets in tightly enough that they don’t move, I actually bend the rails on the rack, which is a little bit surprising, as they are rather thick.

I don’t have a picture of it, but picture a crew going through the outside panel, and into the rail in this pic. They get squeezed together and the rail bends.

Luckily, they were not permanently bent, just flexed, so when I released the screw it is straight again.

[mattlach]

So, while the flushing continues due to setback #1, I started working on a solution for setback #3, and it involves more amateur metal work!

I measured the distance between the rails and the panel as 1-1/4", so I went to Home Depot to the plumbing aisle and bought myself an 18" length of 0.5" galvanized steel pipe.

Using the same method as in the C-channel cuts, cutting them with my Harbor Freight Cut-off saw, and deburring with my Dremel, I made myself some spacers:

It’s not professional work, but it will do the job.

I’m going to stick these between the panel and the rails, sticking the screw through them, and then I can tighten the screws as much as I damn well please. Nothing is going to bow.

Yes. It’s still winter, but today was a balmy 39°F, practically t-shirt weather, so working outside wasn’t too bad.

[mattlach]

Alright, so due to Setback #1, I still don’t have a flushed radiator that is ready, but that doesn’t mean I can’t start measuring and drilling holes to attach the mounting brackets.

The brackets are designed to mount vertically to your wall, being screwed in place through two holes, one at the top and one at the bottom.

They have upside down tear drop shaped cutouts where prongs that stick out from the radiators stick through, and then the weight of the unit pulls it down to the bottom of the cut-out, self centering it and holding it in place, all while allowing a bit of loose tolerancing so your mounting does not need to be perfect.

The prongs are not pre-installed on the radiators (as many or most use them with feet rather than mount them to the wall). They come with the wall mounting kit, and you install them to the radiator yourself.

You do this by removing four chassis screws from the side of the radiator you plan on facing the wall, and replacing them with the prongs.

This is fairly easy. Only minor annoyance is that the screws are different. The case screws are some size of Torx, whereas the screws for the prongs are Allen keyed. This meant switching my bit back and forth for each screw. Small annoyance, but not a big deal.

The fan bracket is configurable. You can easily move it to the side that makes sense for your build. Same with the prongs, they can be installed on either side.

This is how the brackets attach to the radiators:

[mattlach]

The next part is a little tricky.

I scoured the manual, but nowhere in that thing did I find the actual measurement of the space between the two brackets. Since all of these pieces are loose (when not mounted) and don’t stay in place, it was rather difficult to measure the hole spacing for drilling.

It turns out the screw holes are symmetrical, so if you turn a wall mount bracket upside down, so it can hang under its own weight by the prongs on the radiator, the holes are still in the correct place as when it is right side up.

So, the way I mounted mine was by first drilling the top left hole where I thought it should go. (keep in mind that the top of the radiator winds up being a couple of inches higher than the top screw hole)

This involved silicon oil on my drill bit, safety glasses, hearing protection, a shop fact up against my drilling area (to avoid conductive dust getting into the servers) etc. etc.

Then with the bracket now screwed to the side of the rack by one hole, and hanging straight down, I used by large framing square to make sure it was straight, and then marked a spot for the second hole at the bottom, and drilled that out.

After that I went and grabbed one of the radiators and popped it in the first bracket by both the top and bottom holes, and simultaneously supported the weight of the right side where there was no bracket mounted. With that bracket hanging upside down by the prongs, my level balanced on top of the radiator (amusingly not going for straight, but going for similarly un-level as the rack itself, so that it is square) I was able to support and slightly lift the right side and mark the spot to drill the holes through the second bracket with a very very short pencil. I cut one off really short with my oscillating tool that just happened to be within arms reach, to be able to get in there.

Unfortunately I didn’t take any pictures of the above, as I simply did not have enough hands during this process. I would suggest anyone who does this has the help of a second set of hands. I was able to rest a lot of load on the wall mounting bracket on the one side, because mine was screwed into metal, but if these were just two drywall anchors, it might just pull out. if resting on just one side.

After this, I drilled the remaining two holes, and did a test fitting.

The spacers, they worked perfectly. The rails do not bow at all now, and I can tighten the screws so that the brackets don’t move at all.

We are nowhere near done, but it is starting to look like something!

And yes, it is huge. That’s a Corsair 1000D, a monstrously large case for size comparison.

The second one will eventually go right under it, with only enough clearance to allow for lifting it up and removing it without interfering with the radiator above. This will bring the lower radiator uncomfortably close to the floor, but I’ll just have to be careful to not backing into it, and check it for dust more often than I otherwise would.

[mattlach]

So,

With my status now being “waiting for my radiator to flush” as well as “waiting for several things I ordered to show up” and "saying ‘damn’ as I remember things I forgot to order that will delay me even further, I wasn’t sure what I was going to accomplish on the project today.

Then my better half and her mother (who is visiting) announced they would be going out shopping, and I suddenly had an idea.

I know for most of you “having the house to yourself” doesn’t automatically translate into “time to do electrical work”, but that is certainly the case for me, as those other people in the house generally do not appreciate it when I shut off the power.

You see my “server room” used to be a second laundry room. At some point under some previous owner, the finished basement of our house was some sort of in-law apartment, and that apartment had its own laundry room. That is the space I am using for my rack (and miscellaneous storage)

There are only two electrical outlets in this room, one regular 15 amp outlet (halfway up the wall, presumably intended for washing machine power) and one big four prong 30amp 240v NEMA 14-30 dryer outlet.

All of this time I have been running my rack, containing all of my switches and two servers, and more recently (since the aforementioned remote desktop project) my desktop have all been sharing the same 15A receptacle… on a 15A circuit… shared with lights and ceiling fans throughout the house.

I have for a long time felt that this was asking a lot of that particular outlet, and have wanted to do something about that.

In fact, I’ve been planning this for a long time, but just rarely have both the house to myself and the time to do it at the same time.

So, this NEMA 14-30 receptacle was wired with #6 3-wire (plus ground) Romex to the main panel, where it was connected to a good old fashioned 30amp dipole breaker.

A while back I bought

• A 20amp combination arc fault dipole breaker (Siemens Q220AFC, because that is what is listed for my panel)
• A multi-pack of 20amp GFCI receptacles
• A matching 2-gang faceplate
• Some assorted wire nuts.

It was time to swoop into action.

I prepped all of my tools and made sure I knew I had everything handy. With some luck this would go relatively fast (before my UPS:es run out of run time)

I took the panel cover off, with it still live to buy myself some time.

I switched off the main breaker. Quickly removed the old 30amp dipole breaker, installed the new 20amp arc fault breaker, started putting the panel cover back on, and then Murphy’s law struck.

You see, at some point, one of the previous owners (or a professional they hired, but judging by the state of the electrical, I am going to go with “not professional work”) stripped one of the panel cover screws. I had made a mental note of “I’m going to have to fix this some day”, but had just kind of left it semi attached in the screw hole.

Well, during the re-attachment of the panel cover, the screw decided to make a run for it. Some tension in the panel cover released, catapulting the screw off the panel, and right into a hole in the wall, where it disappeared in an irretrievable location.

Well, I could have just left one screw detached, and lived with it for now. Many people would be tempted to use a sheet metal screw to just keep it on there, but that is definitely not per code (panel screws are not allowed to be sharp) I decided I wanted it done right. So I temporarily flipped the power back on to allow my UPS:es to charge, and ran off to Home Depot and bought a 5-pack of 0.5" long 1/4" screws. I’m not sure exactly what dimension the old panel screws were, but they were smaller than 1/4".

For a project a while back, I bough this Greenlee Drill/Tap kit, so I figured it was now going to come in handy for a second time.

I flipped the main breaker off again (Just in case) and one by one drilled and tapped out the old screw holes to 1/4". I was now able to reattach the panel cover with all four screws.

And yes. It’s a Murray panel (now defunct, but still supported by Siemens breakers) with many breakers that don’t belong, and and some questionable wiring. I’m working on it bit by bit and improving it.

Once all of that was done, I switched off the new breaker I just installed, and went to work on the receptacles.

I removed the single 240v 30amp NEMA 14-30 receptacle, and replaced it with two 20amp receptacles.

I wired them as a MWBC (Multi-Wire Branch Circuit) able to give me an equivalent of two 20amp circuits in that same box with the existing wiring.

Now I have plenty of power for my many things that are going in the server room.

Some day, if I get a 240v or 120v 30amp UPS, I may rewire it again to 120v or 240v 30amp, depending on what kind of UPS I get a good deal on.

Alright, I think that is enough for one weekend.

[mattlach]

I have been considering making some changes to the water loop layout as well.

Consideration #1:

Splitting up dual D5’s to two separate D5’s and having each power just one block of the game machine (one for CPU, one for GPU.)

My plan has originally been to dedicate a dual pump top of the two blocks in the game machine, but I have been thinking. Dual pumps (in series) DO scale, but they do not scale well. Generally if you want better flow, it is better to reduce head pressure. I bet I could get better flow over both the blocks, if I run them as separate sub-loops from the same reservoir, with just one pump each.

I’ll have to think about this. I’ll have to buy more tubing, and one more set of QDC’s but that might be a small price to pay…

Consideration #2:

I currently have all D5’s. I was going to continue using D5’s, but then I thought of something. I’ll be using 5 D5 pumps. (one per radiator, one for the workstation CPU block, one for the game machine CPU block and one for the game machine GPU block)

D5’s dump their heat into the loop. D5’s can consume up to 25W each. That’s an additional 125w of heat being dumped into my loop.

DDC pumps do not dump their heat into the loop. They can even be equipped with small heat sinks on the bottom to help them dispatch their heat tot he surrounding air.

DDC pumps also do not achieve as high of a flow rate as D5’s do (unless under very high head pressure).

I still want my flow to be exceptionally high across the game systems blocks (and I should be able to accomplish that with a one D5 pump to one block ratio) but the other pumps output are a lot less flow dependent. Maybe I would switch out the radiator pumps and the pump for the workstation to all be DDC pumps? That would certainly limit the system wide pump heat load…

DDC pumps seem to ave become less popular lately though. I’ve been searching, and there aren’t too many options that are available and in stock from trusted brands.

I am going to have to give this some thought.

[mattlach]

So, I am considering buying some Bykski DDC pumps.

As a reminder, I am looking for a few DDC pumps, because I plan on using 5 pumps, and D5 pumps each dump as much as 25w into the loop. That’s an extra 125w the loop has to handle, which just doesn’t seem efficient. The D5 pumps will get better flow, so I still plan to use them for the parts that benefit the most from high flow (game system CPU block, game system GPU block) but the remaining three pumps (workstation CPU block, and one each per radiator) could go to slower DDC pumps that dump their heat external to the loop without a problem.
Anyway, the above statement is a very weird one for me. Normally Bykski wouldn’t even be one of the brands on my shopping list, as I have been very wary of Chinese brands in the past.

My biggest concern has been that anything Chinese that connects to the internet is a potential spying tool, but that obviously isnt the case with a DDC pump. My second concern is that while I know Chinese stuff has gotten better in the last 20 years, I still have that nagging voice in the back of my head that says “it is Chinese crap. Everyone knows they make garbage. What if it leaks or fails after the first couple of weeks?”
I guess in this case, the pumps will all be pretty far from the sensitive and expensive PC parts, so leaks - while annoying - would be less critical.

I also don’t want to reward the Chinese copies of western designs. That is a big pet peeve of mine.

But looking at all of my usual sources (Amazon, Tinanrig, Modmymods, Performance-PC’s, Newegg), I just cant find any other fully dressed (pwm controlled pump, pump top, case, pump bottom with heatsink) DDC pumps in stock, so I guess I don’t have much of a choice.

I have found a couple of others without a heatsink on the bottom, but from my reading, you really want a heatsink on DDC pumps that see a lot of use, or they won’t last long.

They have a kind of ugly (but practical) top that is designed to either accept a screw in cylinder reservoir, or be used standalone with a G1/4 inlet.

I don’t really want to buy this pump, but I don’t see many other options. DDC pumps seem to have fallen out of favor in the water cooling community in later years, with most either discontinued or just not available.

I also have next to zero DDC experience. Every water cooling system I’ve built in the last decade has used D5 pumps. Maybe I can get the pump motor and tops and bottoms separately and assemble myself?

I’d appreciate any thoughts or recommendations for alternates from anyone who has DDC pump experience.