How Thick Should 3D-Printed Walls Be?

A printed wall should be a whole multiple of the extrusion width, and at least two of them. On a 0.4 mm nozzle that means 0.8 mm for anything decorative and 1.2 to 1.6 mm for anything that takes load. The number is not a property of the design. It is a property of the nozzle that will print it, which is why the same model wants a different wall on a 0.6 mm machine.

A slicer does not fill a wall with material. It draws perimeter lines of a set extrusion width — usually close to the nozzle diameter, often a little wider — and stacks them layer by layer. A 1.2 mm wall on a 0.4 mm nozzle is three lines side by side. A 1.0 mm wall is two lines and a 0.2 mm remainder that has nowhere to go.

For years that remainder was the whole story: the slicer either left a void down the middle of the wall or over-extruded to close it, and both were visible in the part. Modern slicers handle it better. Variable-width extrusion, which most current slicers implement in some form, widens or narrows perimeters to absorb an intermediate wall rather than leaving a gap.

That makes the whole-multiple rule softer than it used to be, not obsolete. A wall sized to the extrusion width still prints as clean stacked lines with predictable dimensions. A wall sized arbitrarily prints as lines the slicer has quietly reshaped, and the dimension you get back is less predictable than the one you asked for. If a wall has a tolerance, size it to the nozzle.

The practical rule: decide the extrusion width first, then make every structural wall an integer multiple of it. Two lines is the floor. One-line walls exist — vase mode is the extreme case — but a single perimeter has no neighbour to bond to sideways, so it splits along the layer lines under almost any load.

The common instinct when a part breaks is to raise the infill percentage. It is usually the wrong lever. For bending and for most real loads, the material far from the neutral axis does the work — and that material is the perimeter. Adding a fourth perimeter moves more material to the outside of the section; raising infill from 20 to 40 percent adds material near the middle, where it contributes least and costs the most print time.

The failure mechanism supports this. Extruded parts are not solid: the strands are laid down as rounded beads, and the gaps between adjacent beads and between layers are voids. Those voids concentrate stress and are where cracks start, and a review of void formation in fused filament fabrication traces most of the strength shortfall in printed parts to exactly this.^[1] Perimeters bond along their whole length to their neighbours. Infill touches the wall at discrete points.

A second consequence: perimeter count and wall thickness are the same decision. If a design calls for a 2.4 mm wall on a 0.4 mm nozzle, it is asking for six perimeters, and the slicer will print six perimeters whether or not anyone meant to specify that. Making the intent explicit in the model — wall thickness as a named dimension derived from extrusion width — is what keeps it true when the part is scaled or the nozzle changes.

A printed part is not an isotropic solid. Within a layer, material is continuous along the extrusion path. Between layers, it is held by a bond formed while the previous layer was still hot enough to fuse. That bond is reliably the weaker one, and reviews of process, structure and properties in polymer material extrusion report the gap between in-plane and through-thickness strength as one of the defining characteristics of the process.^[2]

So wall thickness and print orientation are a single decision. A 1.6 mm wall standing vertically is strong in the direction the lines run. The same wall lying flat, loaded across its layers, is weaker than its thickness suggests. Thickening a wall that fails in layer separation buys less than reorienting it.

Two design responses follow. Put the long axis of any feature that takes tension along the layers rather than across them. And when orientation cannot change — a printed clip that has to bend somewhere specific, a boss that has to stand up — treat the through-layer direction as the design case and size for it, rather than sizing for the direction that happens to look strong in the viewport.

Process parameter studies show the same thing from the other side: raster orientation and infill pattern move measured tensile properties substantially, independent of how much material is present.^[3][4]

These are starting points for a first print, not specifications. Every one of them should be replaced by a measured value from the machine that will do the work — see the section below on why that matters more than usual here.

Table: Starting wall thickness by process and nozzle. Calibrate before relying on any of it.

Three adjustments to make before treating any row as settled.

Height matters as much as thickness. A tall thin wall fails by buckling and by wobbling during printing, not by breaking. A 0.8 mm wall 10 mm tall is fine; the same wall 80 mm tall is a problem the slicer cannot fix. Add a rib, a taper, or a flange rather than thickness.

Hollow resin parts need drain holes. A sealed hollow print traps uncured resin and creates a suction cup that can pull a part off the plate. Two holes, so air can enter as resin leaves.

The first layer is not the wall you modelled. Squish spreads the bottom layer outward — elephant's foot — so a wall measures wider at the base than in the middle. It matters when the bottom of a wall is a mating surface.

A slicer will warn about a wall it cannot print. It will not warn about a wall that prints badly, and it certainly will not warn that the wall was thin because a variable in the model went somewhere unexpected.

That is the failure worth designing against. A hollowed box with a wall parameter is fine at its default and fine at the sizes anyone tested. Scale the box up without scaling the wall and it is under-built; scale it down and the wall consumes the interior. Neither state announces itself, and in a parametric model that other people drive, neither state is hypothetical.

The fix is to carry wall thickness as a named value rather than a typed number, derive it from extrusion width rather than from taste, and put a check on it that fires when it drops below two lines. In a parametric tool that means one validator expression; in a CAD file that means a note and a habit. Either way the point is the same — the constraint belongs where the geometry is, not in a slicer preview that happens once per print.

The general form of this is design for additive manufacturing: treating the manufacturing constraint as an input to the model rather than a check applied afterwards.^[5]

The numbers in the table above deserve a caveat that most articles on this subject do not give.

There is a large, rigorous literature on dimensional accuracy and tolerancing in additive manufacturing. Almost none of it is about desktop machines. The careful work on geometric capability and assemblability targets metal powder-bed fusion; the careful work on process-structure-property relationships characterises materials rather than issuing design rules. The specific figures that circulate for desktop FDM — 0.8 mm minimum walls, two perimeters, 1.2 mm for structure — trace back to printer manufacturers and service-bureau design guides, not to published studies.

This does not make them wrong. Vendor guidance is usually derived from a great deal of real printing, and the physical reasoning behind the whole-multiple rule is sound on its own terms. But it does mean the confident tone in which these numbers are usually repeated is not earned, and that anyone who calibrates their own machine has better data for their machine than the internet does.

Which is the practical conclusion: print a test part with walls stepping from one to six lines, load it until it breaks, and write down what happened. That measurement outranks every table in this article, including the one above.

The other two design rules in this series take the same shape — a familiar number, and the thing it is actually measuring. The 45° Overhang Rule, and When It Is Wrong is about how far a layer can lean before the one below stops holding it up. Clearances for 3D-Printed Parts That Have to Move is about the gap two parts need between them before either will move. Both come with a coupon to measure it on your own machine.

For the modelling side — hollowing a solid two different ways, and sizing the wall that results — see Shell, Hollowing, and Wall Thickness for Printing.

For carrying a wall thickness as a checked, named value rather than a typed number, so that a model cannot be scaled into an unprintable state without saying so, see Beyond Sliders: Safe Parameter Spaces for Printable Products.