3D-Printed Rack Mounts for Mini PCs
Six machines, one rack unit, and a spool of filament that cost less than a shelf
Contents
The mini PC has quietly won the homelab. It is quiet, it idles at single-digit watts, it costs less than the RAM in a used enterprise server, and six of them will do everything a 2U machine did while being inaudible. There is exactly one problem: nobody makes a decent way to put them in a rack. The vendor sells a VESA plate. The rack accessory market sells a shelf, and a shelf means you now have six identical black bricks in a heap with twelve cables entering from random directions, which is aesthetically depressing and operationally worse when you need to identify which one to reboot.
So print the mounts. This is the single most satisfying thing a printer has done for my rack, and it costs about €4 of filament per rack unit.
Why the market doesn’t serve you
Racks are 19 inches wide and standardised down to fractions of a millimetre by EIA-310. Mini PCs are whatever size the manufacturer felt like this year, and the dimensions change between revisions of the same model. A vendor cannot build an inventory of mounts for four hundred slightly different bricks with a market of about nine thousand enthusiasts. The economics simply refuse.
A printer inverts this. The bracket is a trivial part — a flat plate, two ears with rack-hole spacing, and a pocket the shape of your specific machine — and the design cost is amortised across exactly one user, which is fine when the design cost is an evening. This is the same argument I made in 3D printing functional parts: the printer earns its keep on the objects that could never justify a mould.
Before designing anything, decide the orientation. Mini PCs mounted flat against a plate means one machine per 1U and a lot of wasted rack. Mounted vertically, on their edge, side by side, a standard 100 × 100 × 40 mm machine gets you four or five across a 19-inch opening in 3U of height, with all the ports facing forward where you can read the labels. Vertical is right, and it took me a wasted print to work that out.
Numbers that matter
The rack side of the design is fixed and non-negotiable. Get these right and the print bolts up first time:
- Rack unit height: 44.45 mm. Never 44.5, never 44.
- Mounting hole pattern: each U has three holes, spaced 15.875 mm, 15.875 mm, and 12.7 mm to the next U’s first hole. It is not evenly spaced, and assuming it is evenly spaced is the classic first mistake.
- Ear width: the panel is 482.6 mm across, with holes centred 465.1 mm apart — that is 232.55 mm either side of the centreline.
- Hole size: 7 mm clearance for M6 cage nuts. Print them at 7.4 mm because a printed hole comes out undersized.
- Leave 0.5 mm off the height. A 3U panel should be 132.85 mm rather than 133.35 mm, so it drops in without fighting its neighbour.
The machine side you measure yourself, with callipers, three times. Add clearance: 0.4 mm per side for a snug fit that still slides, 0.6 mm if you want to insert it one-handed in a dim rack. Printers systematically undershoot holes and overshoot posts by roughly 0.2 mm because of extrusion width and corner rounding, and the number varies with your printer and your filament. Print a 20 mm calibration cube and measure it before you commit to a six-hour print.
Ventilation is design, rather than decoration. Mini PCs breathe from the bottom and exhaust at the back, and a solid printed plate under one is a machine that will thermally throttle in July. Cut the plate away to about 40% material under the intake. I lost an afternoon to a beautifully solid bracket that pushed a node from 62°C to 81°C under sustained load, which the CPU handled by quietly halving its clock speed for a fortnight before I noticed it in the metrics.
Drawing it without becoming a CAD person
You need very little CAD ability for this. The part is a plate, two ears and some pockets, and any parametric modeller will do it in under an hour once you stop being frightened of the interface. Start from a sketch of the rack panel — 482.6 mm wide, some multiple of 44.45 mm tall minus 0.5 — and drive everything from named parameters rather than typed-in numbers. Make machine_width, machine_depth, clearance and wall variables at the top of the model.
The reason is practical. Six months from now you will buy a slightly different mini PC, and a parametric model means you change one number and re-export. A model with 100.4 typed into eleven places means you start again. This is the single habit that turns a printer from a novelty into a tool, and it applies to every functional part you will ever draw.
Print one ear first. A 40 × 60 mm test piece with two rack holes in it takes eleven minutes and tells you whether your hole spacing, your clearance compensation and your rack’s actual geometry agree. I skipped this step exactly once, on a five-hour print, and the resulting object is now a rather expensive plastic doorstop that fits nothing in this flat.
Material choice, which is the whole game
This is where prints fail, and the failure is slow and boring rather than dramatic.
PLA will fail. Not immediately. PLA has a glass transition temperature around 60°C, and a bracket bolted to a rack, holding a warm machine, in a cabinet running 12°C above a summer ambient of 26°C, will spend real time above 45°C. PLA creeps at temperatures well below its Tg — it deforms slowly, permanently, under constant load — and a PLA bracket sags visibly over a summer. Mine did. The machine did not fall out, and it did end up at a distinctly unimpressive angle, and the plastic had gone from rigid to slightly rubbery.
PETG is the correct default. Tg around 80°C, tough rather than brittle, prints on any machine at 240°C with a 80°C bed, cheap, and it tolerates the constant tensile load of a hanging machine without creeping. It is stringy and prints slower than PLA, and neither matters for a bracket. Print it and stop thinking about it.
ASA if the rack lives somewhere hot, or if it gets any UV. ABS’s better-behaved cousin: Tg around 100°C, needs an enclosure, smells industrial. Worth it in a loft, unnecessary in a flat.
Never nylon. It absorbs atmospheric moisture, swells, and your carefully measured 0.4 mm clearance becomes 0.1 mm and then negative.
Print settings that matter, and only these:
| |
The orientation line is the one that matters most. FDM parts are 40–60% weaker along the Z axis, because layer adhesion is worse than the plastic itself. A bracket printed standing up will shear along a layer line at the ear, cleanly, under load. Printed flat, the load runs across the layers and the part is as strong as the material. If a part must be strong in a direction, print the layers perpendicular to that direction. Everything else in the profile is negotiable; this is not.
Perimeters carry structural load in FDM parts and infill mostly stops the top surface sagging. Four walls at 0.45 mm gives you a 1.8 mm solid shell, which is stiff. Going from 25% to 60% infill adds forty minutes and eleven grams and almost no strength.
The bits worth adding once it fits
The bracket is the excuse; the surrounding details are where the value accumulates.
Print the labels into the part. An embossed name on the front face, 4 mm tall, 0.6 mm proud, costs nothing and survives forever, unlike the label maker’s output which curls off in a warm rack within a year. Recess the text 0.2 mm below a raised border if you want it legible without a colour change.
Integrate the cable strain relief. A small printed hook or a 6 mm slot on the rear edge of each pocket, and the power lead is captive. This sounds trivial. It means that pulling a node forward for maintenance no longer unplugs the node beside it, which is the failure that turns a two-minute job into a twenty-minute one and an incident report.
Add a blanking plate for the empty slots. Print a flat panel for the gaps you have not filled yet. It maintains the airflow path through the cabinet — an open slot is a short circuit that lets your intake air bypass every machine — and it makes the rack look finished, which matters more to morale than I would like to admit.
Consider a printed drip tray if there is anything above it. I have never had a leak in a rack. I have had a drink above one.
Troubleshooting
The ears don’t line up with the cage nuts. Ninety per cent of the time, the non-uniform hole spacing. Measure your actual rack, because cheap racks are cheap: I have one where the hole spacing wanders by 0.8 mm over 12U, which no CAD model will predict. Design the ear holes as 8 mm slots rather than 7 mm circles and the problem evaporates permanently.
The machine is a hair too tight to insert. Do not force it. PETG has enough give that forcing works and leaves the part under permanent stress, which is exactly the condition that makes plastic creep. Reprint with 0.2 mm more clearance, or take a file to it, which takes ninety seconds and works fine.
The bracket sags after a few months. PLA, or too few perimeters, or the load is running along a layer line. Reprint in PETG, flat. Also check the temperature it is actually living at — a thermometer in the rack often reads 10°C above what you assumed.
Prints warp off the bed on a long flat part. Long thin parts are the worst case for warping because they have maximum shrinkage across the longest dimension. A brim of 8 mm, a clean bed, and a first layer at 0.28 mm fixes almost all of it. PETG additionally likes a slightly less hot bed than the internet suggests; over-adhesion at 90°C can pull chunks out of a PEI sheet.
The machine runs hotter than it did on the desk. Airflow. Compare the printed mount against the original feet: the machine had 8 mm of standoff and now has 0. Add 6 mm bumps to the pocket floor and cut more of the plate away. A mount that costs you 15°C has undone the reason you bought a mini PC.
It rattles. Printed plastic against a rack is a resonator, and the fans will find its frequency. Thin adhesive foam between the ear and the rack rail, and grommets under the machine, kills it — the same decoupling logic as in building a silent rack for a flat, applied at a smaller scale.
The verdict
Print the mounts. On a homelab with three or more mini PCs, this is the highest-satisfaction-per-euro modification available: about €4 and four hours of printer time per rack unit against €40 for a shelf that does the job worse. The result is a rack where you can read every label, trace every cable, and pull one machine without disturbing its neighbours.
The honest case against: it takes an evening in CAD you may not want to spend, the tolerance stack means your first print will be wrong, and if you own two mini PCs the shelf is fine and you should buy the shelf. There is also a real maintenance tax — buy a different mini PC next year and you redesign, whereas a shelf holds anything. I regard that as the price of a rack that looks deliberate, and I accept that this is partly vanity.
What surprised me is the second-order benefit. Once mounting a machine costs €4 and an evening rather than €40 and a compromise, the rack stops accreting. Every machine has a place, a label and a reason, and the pile of bricks-on-a-shelf that used to be my “temporary” arrangement — three years temporary — finally went away.
Related Content
