Slicers offer somewhere between ten and twenty infill patterns and the internet offers a confident ranking of them. The ranking is mostly noise. Pattern choice does change a few things that matter, but strength at a given density is not reliably one of them, and the setting that dominates your print time and your material bill is the density slider sitting immediately above the pattern dropdown.
Three runs of the same lid, density only
A flat enclosure lid — 160 cm³ of enclosed volume, 25 mm tall, three walls, 0.2 mm layers, 100 mm/s, a 0.4 mm nozzle. Pattern unchanged; only the density moves.
| Infill | Time | Extruded volume |
|---|---|---|
| 10% | 2.784 h | 68,514 mm³ |
| 20% | 4.039 h | 99,790 mm³ |
| 40% | 6.548 h | 162,342 mm³ |
Each doubling costs more than the one before it: 10 to 20% adds roughly seventy-five minutes, 20 to 40% adds two and a half hours. Step the infill field of the print time estimator upward on a part of your own and the point where it stops paying for itself becomes obvious.
Nearly four extra hours and well over double the plastic, for one slider. No pattern change comes close to that, which is why the useful skill here is choosing a density deliberately and then picking a pattern for a specific reason.
The patterns that are genuinely different from each other
Ignore the marketing names and sort them by dimensionality — how many directions the structure actually connects in. That is the property that predicts behaviour.
- One-dimensional: lines, zigzag. Straight passes in one direction per layer, alternating. Fast, minimal travel, weak in every direction except along the lines. Fine for anything that just needs the top surface held up.
- Two-dimensional: grid, triangles, tri-hexagon, honeycomb. A pattern that repeats in the print plane and stacks identically layer on layer. Strong in the plane, no connection between layers beyond the plastic itself. Grid crosses over its own extrusions where lines meet, which is where those faint nozzle-strike ticking noises come from on some machines.
- Three-dimensional: gyroid, cubic, octet, adaptive cubic. The structure changes shape as it climbs, so it connects in all three axes. Gyroid in particular never crosses itself, prints as continuous curves, and is close to equally stiff in each direction.
- Special-purpose: lightning, concentric. Lightning is a support scaffold for the top skin and nothing else — it branches only where the top layers need holding up, so the interior is essentially empty. Concentric follows the perimeter, which keeps a flexible part flexible instead of stiffening it into a lattice.
Where the strength rankings come from, and how much to trust them
It is worth separating the claims by evidence quality, because they circulate as though they were equally solid.
Reasonably well supported. Three-dimensional patterns are more isotropic than two-dimensional ones. This follows from the geometry rather than from testing, and testing agrees.
Supported but usually overstated. Gyroid's advantage. It is a good general-purpose pattern with real properties — no self-intersection, near-isotropy, good behaviour in flexible materials — and the popular claim that it is dramatically stronger than the alternatives at the same density rests on a small number of widely-copied comparisons rather than a body of published work.
Genuinely contested. Any specific ranking at equal density. Published comparisons disagree with each other because they use different specimen geometries, wall counts, materials and load cases, and infill's contribution depends heavily on all four. A pattern that wins in compression can lose in three-point bending on the same machine.
Not supported. That changing pattern at a fixed density is a meaningful route to a stronger part. If a part is failing, wall count and orientation are where the leverage is; the pattern dropdown is not.
Choosing on grounds that are actually reliable
- Fastest, part is not structural: lines or lightning at low density. Lightning especially, if all you need is a supported top surface.
- General functional default: gyroid at 10 to 20%, or cubic if your slicer runs it faster.
- Flexible material: concentric or gyroid. Grid patterns make TPU parts feel oddly firm and hollow at once.
- Part will be crushed or bolted through: raise density before changing pattern, and consider a solid region rather than global infill.
- Very large part, mostly air: adaptive cubic or lightning, both of which vary density with proximity to a surface and can halve the interior work.
Two settings that live next to the pattern and matter more
Infill overlap. This is how far the infill is pushed into the inside face of the wall. Too little and the two never fuse, so a part with beautiful walls delaminates from its own interior under load. Too much and the pattern telegraphs through as ridges you can see and feel, which is the usual cause of infill showing through the walls. It is one of the few settings where the default is frequently wrong for a specific material.
Infill angle. Most slicers alternate the fill direction between layers by a fixed amount. On a long thin part it is worth setting the angle so the fill runs along the part's length rather than diagonally across it, which costs nothing and adds real stiffness in the direction the part is actually loaded.
The interaction nobody mentions
Infill and top-layer count are one decision, not two. Low infill needs more top layers to bridge the gaps without dimpling; high infill lets you get away with fewer. If your top surface is rough, adding infill and adding a top layer are alternative fixes with very different costs, and the top layer is almost always the cheaper one.
The same is true underneath: on a part printed on supports, the bottom skin is bridging over the same kind of gaps and behaves the same way.
What to change first, in order
- Set the density for the job — low for cosmetic, moderate for functional, high only for compression or machining.
- Set the pattern from the list above, for the stated reason.
- Set top and bottom layers to suit the density.
- Leave the pattern alone thereafter and adjust walls when the part needs to be stiffer.
Most people do this in reverse, spend an evening comparing patterns, and finish with a part that is a few percent different and several hours older.