If you have been following my Xeon Phi experiments, you know that I've been having some fun playing around with these old compute cards. Recently thanks to a post on the homelab subreddit, I have managed to get my hands on an incredibly rare Xeon Phi x200 series coprocessor. Due to this being a different revision of Phi from my previous experiemtns though, I can't install it into my existing system. As such I have come to a need for a new server (again).

My previous case was an old Antc 3U unit I found used nearby and has been working perfectly. The best part of it being that it supports full height PCIe cards! Sadly however, this was a fluke of a find, and anything similar new here is just far too expensive to justify buying. Not to mention that there aren't many options supporting full height cards in a 3U case. Normally you could use a 4U case and call it a day, but unfortunately for me, I am out of rack space. With only 3U left, I had to find a solution

You'll note that I used 2'x2' handy panels; this was due to it being easier to find flat sheets in this size, however you can buy a larger sheet and have it cut down for cheaper in some cases. As for the 3/4" MDF, I used (and linked) a 2'x4' sheet as that is what I could find easily, I had it cut down to 2'x2' to fit my CNC. Each of the 2'x2' sheets of 3/4" MDF can make about 5 back panels. The number of inserts and fasteners needed will depend on how many of the holes you wish to populate for mounting. In my case I populated them all so I could test various isntallation methods.

Similarly for the standoffs, the number you need will depend on your motherboards size, and number of mounting holes. Any 19mm switch should work (just be careful about it's depth), athough most will require you to add your own wires and dupont connectors.

The panels can be cut with multiple panels to a 2'x2' sheet of 1/4" MDF. My personal arrangement was: Top Panel + Side Panel, Bottom Panel + Side Panel, Mid Panel + Front Panel, Back Panel. The only special care taken was to machine the insert holes before cutting out the profile, this way I could put screws through them iinstead of using holding tabs; and in the case of the back panel, it was flipped to have the rear PCIe panel pocketed out.

The first panel, and the most important one, is the bottom panel. This panel has the vast majority of the inserts and has a number of optional cuts. Firstly, not all the mounting holes need to be machined, only those for the size of ATX board being used, for panel mounting, and for however many of the modular attachment points are to be used. The modular attachment points themselves are the holes located within the rectangles in this design. The rectangles themselves do not need to be cut and are simply there for design purpouses. These holes are placed in such a way that 3D printed drive cages supporting up to 3.5" drives should fit without issue. I have currently created a single SSD mounting bracket using these, and had planned to use them for mounting for a PCIe support for horizontal PCIe installations. Finally, the small rectangle at the top of the file is a relief for PCIe mounting plates to be able to fit. If you plan to use the vertical backplate with PCIe cards this will need to be cut to a depth of 4.5mm. Currently mount points have only been added for standard ATX board layouts. Overcut will be needed for all box joints on all panels including this one.

The side panels are relatively simple, the holes are optional, and recommended if using the horizontal layout. Otherwise, all insert holes are needed, and again, screws through these holes were used in place of holding tabs to hold down the work piece.

The front panel has a large number of holes to cut and as such can take a while. I used the helical drill operation for this task, but pocketing is probably the better way to handle these. The rack ears and their holes are optional, if they are undesireable you can simply omit them in CAM. The box joints on the bottom do not need to be enclosed, this is simply an artifact from the design.

The top panel has an overlapping rectangle and holes. I would advise against trying to get both of these geometries at the same timw. Instead the holes are intended for use with the horizontal mount design as it has no rear exhaust fans. If using the vertical mount design, instead, you will need to cut the rectangle as a pocket with a depth of 4.5mm. Keep in mind that the depth of this pocket will set the limit for the maximum height PCIe expansion card that can be installed.

The mid panel acts as an internal support for the frame to help stiffen it, in addition to providing a location for mounting the PSU and fans. It is designed to be reversible so that the PSU can be installed on the left or right side of the case. I would recommend cutting all PSU holes incase you change your mind during the dry fit.

The vertical mount rear panel has a number of complex and overlapping geometries that will need to be handled in CAD. Firstly, the rear IO cutout will need overcut on the four corners; the PCIe and power cutouts do not need these however. Next, only the holes underneath the PCIe cutouts need to be drilled, the other geometry is a guide for where the PCIe retention bracket will need to be installed. I only recommend installing inserts for the bracket where they are not near a box joint cut out, as it is very easy to blow out the sides here. The retention bracket is strong enough to not need all of the screws. Finally, above the PCIe cutouts some cuts will be needed to install inserts for locking expansion cards into place. The small rectangle at the top needs to be cut to a depth of 15mm, following which, the small rectangles located within the large one just below that will need to be cut to a depth of 7.8mm. This provides a location to insert threaded inserts that will then retain the expansion cards. These inserts are just pushed in sideways using the 7.8mm deep holes. I recommend adding some cyanoacrylate glue to keep these in place and to help reinforce the MDF in this location (this is required if you pocket out the other side of the panel!). If you wish to pocket out the PCIe slots for easier access to the cards ports I recommend simply using the PCIe cutouts to guide a CAM toolpath to create a pocket over them.

The horizontal mount follows similarly to the vertical one with a few key differences, firstly, instead of the previous PCIe retainer, there is now a PCIe support brace for mounting cards horizontally. Only the holes are needed for this brace, and they will not take any threaded inserts, those instead go into the 3D printed bracket. This bracket does require a custom cut support plate that I have not yet designed, and will depend on the risers used. If you wish to use both sides of the bracket you will need left angle and right angle risers.holes for the PCIe card screws are cut similarly to before, but instead of a small pocket, the large shapes next to the PCIe cutouts will need to be pocketed to a depth of 15mm for installation purpouses. The benefits of this installation method are primarily to do with increased PCIe card height support.

The corner braces are simple to install, they simply go into the corners of the chassis and screw into the isntalled inserts. For the top panel the bottom of the bracket has an insert installed instead of the top, and the top is screwed on using it.

Identical to the corner braces, these are instead installed in the center span of the case.

The PCIe retainer is used for the vertical version of the case, and is screwed into the rear panel with the gapped side against it. This part doesn't take much load, so you do not need all of the screws to install it. It's primarily there to help constrain the PCIe plate.

The SSD mount is a simple bracket used to mount a single 2.5" SSD to the modular mounting holes in the bottom panel. 3.5" drives and multiple drive cages should be able to be installed here without much trouble.

The PCIe brace is not fully confirmed to work as I have not built a full version of the horizontal mount case, but it contains the retention needed for PCIe plates, as well as mounting holes for a 1/2" piece of MDF to be used as a holder for PCIe risers to be mounted to. I will likely desing it with modular mounts that allow a 3D printed spacer to mount the PCIe risers since they do not use a standardized pattern. If used, a PCIe support bracket will almost certainly need to be printed for use with the modular mounting holes if using a PCIe expnasion card of any significant weight.

The main panels for the frame (middle, bottom, sides) were able to be cut relatively quickly, and were used as references for verifying all the designs were correct prior to cutting the longer to cut pieces. This was fortunate, as I caught an error in the original desing of the bottom panel that casued it to have incorrectly aligned box joints.

Once the mistakes with the bottom panel were fixed I proceeded to test fit the panels using the corner braces as well as the PSU to ensure it sat flat against the bottom of the case. Afterwhich I installed the inserts into the panels; this is a task that generally can be done by fitting the insert into the hole then hammering it in (I recommend putting a piece of scrap wood under the panel and hammering it on the ground as it takes a fair bit of force). Sometimes however, the holes don't fit perfectly, in which case holding the insert with a pair of pliers to start the hammering process worked perfectly.

The rear panel was cut and the case was dry fitted to ensure that PCIe cards would fit correctly along with the board.

Here you can see that the board line sup nicely as well as testing the PCIe mounting alignment. Note that in this image the PCIe plate for the xeon phi is warped and is not misaligned.

Finally the case was ready to assemble and fill with parts. Firstly I soldered on and terminated the connections for the front panel power switch and power light. The power cable for the PSU had the end cut off and spade connectors were crimped on to connect to the new socket on the back of the case.then fans and the PSU were installed into the middle panel. The panels were then assembled and screwed together; unfortunately I was a touch short sighted in setting up the install locations of the screws for the front corner brackets, so I had to make use of a flexible shaft adapter for a screw driver and carefully fastened it all together. Then the SSD mount was installed along with a fan hub. Altogether the process of building the computer in the case was relatively simple aside from the difficulty of putting on the front panel.

Once installed in the rack I find it to look quite nice, and it fits within the 3U spec perfectly... except for the screws on top, but that can be dealt with by using countersunk screws.

I feel that the case came together quite nicely and it was a great project to get me using my CNC! Of course there are some issues, such as the installation of the front panel that can be addressed. There is also the risk of introducing so much wood into a computer environment, this can have potnetial risks, however I am not too concerned of this personally. Of course this is up to each individual, and if it is a concern fire reardant MDF is a product that may mitigate the risk significantly if not completely. As for grounding, the PSU should ground everything, however I may still try to add some foil tape underneath the standoffs and route it to under the PSU to help with this.

This case is excellent for its modularity and I hope to design more revisions of it along with more options for the modular mounting points, possibly to add more drive capacity. I do also plan to complete the design of the horizontal rear panel and its bracket in order to use it for in future. If you like my design and content, or if you use it and like it please feel free to leave a tip using my paypal link!