People spend a remarkable amount of energy choosing between gyroid and cubic, and almost none choosing between two walls and four. The second decision matters far more, and it is cheaper.
The four variants, measured
One housing: 4,900 mm² of footprint, 55 mm tall, 0.2 mm layers, 120 mm/s, 0.4 mm nozzle. PLA at $20 a kilogram, machine at 30 cents an hour, 8% failure rate. Four ways of making it stiffer, each one input changed.
| Setting | Time | Plastic | Mass | Cost |
|---|---|---|---|---|
| 2 walls, 20% infill | 2.446 h | 70,868 mm³ | 88 g | $2.76 |
| 4 walls, 20% infill | 2.794 h | 81,288 mm³ | 101 g | $3.16 |
| 6 walls, 20% infill | 3.133 h | 91,430 mm³ | 113 g | $3.54 |
| 2 walls, 40% infill | 4.106 h | 120,526 mm³ | 149 g | $4.66 |
Going from two walls to six takes the print from $2.76 to $3.54 and adds forty-one minutes. Doubling the infill and leaving the walls alone takes it to $4.66, adds an hour and forty minutes, and lays down nearly two and a half times as much extra plastic as tripling the wall count did.
You can re-run the geometry with your own part; the ratios hold for anything with a reasonable amount of surface relative to its volume, which is most functional prints.
Why the walls do the work
The cost table would be an argument for walls even if the two were structurally equivalent. They are not equivalent, and the reason is geometric rather than experimental.
A wall is a continuous extrusion running the full length of the perimeter, fused along its entire boundary to the wall inside it and to the layers above and below. Load applied in the plane of that wall is carried by unbroken material.
Infill is a sparse lattice that touches the shell only where a cell wall happens to meet it. Load has to cross that contact to get into the lattice at all, and the lattice is mostly air. It resists the shell caving inward, it stops the top skin sagging, and in compression it does real work — but in bending, which is how most brackets and housings actually fail, the outer fibres carry nearly everything and the outer fibres are the walls.
That is why wall count is the first setting to reach for and how many walls to use is a more useful question than any pattern comparison.
What the testing actually supports, and what it does not
Independent community bench testing on this subject is plentiful and broadly consistent in one direction: increasing wall count produces large, repeatable gains, increasing infill percentage produces smaller ones, and swapping between the mainstream patterns at a fixed percentage produces differences that are small and depend heavily on which way the load is applied.
Three honest qualifications.
Absolute numbers do not transfer. These are bench tests on particular geometries, materials and machines. The ordering has held up; any specific percentage improvement you have seen quoted is a property of that test, not of 3D printing.
Most pattern comparisons do not control for mass. This is the important one. Two patterns set to the same infill percentage do not lay down the same amount of plastic — the percentage is a density target the slicer approximates differently for each geometry. So a test showing pattern A beating pattern B at 20% is often showing that pattern A is heavier at 20%. The comparison worth making is at equal grams, and it is rarely the comparison that gets made.
Load case decides everything. A pattern that wins in three-point bending can lose in compression normal to the plate. There is no scalar "strongest infill", and any article that gives you one has picked a test without telling you.
Where the pattern genuinely does matter
Not never. Just not for the reason it usually gets discussed.
- Gyroid and cubic distribute material in all three axes, so they behave similarly regardless of load direction. Good default when you do not know how the part will be loaded — and gyroid has no self-intersections, so it prints without the crossings that cause trouble below.
- Grid and lines are strongly directional and cheap in time. Fine when the load is in one known axis.
- Honeycomb is efficient in compression along the cell axis, which is exactly the load case in a spacer or a standoff.
- Concentric is the one to use with flexible filaments, because it follows the perimeter rather than crossing it, and crossing extrusions in a soft material is where flexible prints go wrong.
- Lightning is a support structure for the top skin and nothing else. It is the correct choice for a display model and the wrong one for anything loaded.
The patterns that create problems
Grid lays a line across a line at every crossing, which means a double thickness of plastic in the same layer. On a well-tuned machine it is fine; on a machine already running a little hot on flow it produces a raised ridge the nozzle then drags through, and that is one of the ways a print ends up with a rough top surface or worse.
Too little infill under a solid top produces pillowing, because the skin has nothing to bridge to. That is a spacing problem, not a strength problem, and raising the percentage is only one of two fixes — adding a top solid layer is usually cheaper.
And if you see gaps between the infill and the walls, no pattern change will help. That is an overlap setting.
What to actually set
One genuine exception before the settings. Anywhere a fastener enters the part — a heat-set insert, a tapped hole, a self-tapping screw into a boss — the load is carried by whatever material happens to be within a few millimetres of the hole, and on a sparse lattice that may be almost nothing. Those regions want local solid material rather than a global percentage rise, which most slicers will give you as a modifier volume around the feature. Raising the whole part's infill to serve four screw holes is the expensive way to solve a local problem.
For functional parts: four walls, gyroid, and whatever infill percentage the top skin needs to bridge properly — typically somewhere in the teens or low twenties. Add walls before percentage, every time. What infill to use for functional parts has the number; the pattern guide covers the geometry of each option properly.
Then look at what those settings cost. Infill percentage moves both print cost and print time more than almost any other slicer setting, in the same direction, at the same time — which makes it an expensive habit to carry at a number you picked once and never revisited. It is the same shape of mistake as defaulting to fine layers: a setting chosen for a reason that no longer applies, quietly charged to every print since. Put your own part through the cost calculator at both settings and the decision makes itself.