Slicer software puts infill front and center: a slider from zero to one hundred percent that looks like the main strength dial on a printed part. It is not. On a typical FDM part the perimeter walls carry most of the load, and infill past a certain point buys weight, print time, and filament far faster than it buys stiffness. Knowing where that crossover sits is the difference between a bracket that survives a drop test at 90 grams and one that survives at 210 grams for twice the machine time.

What Infill Actually Does

An FDM part is not solid plastic. The slicer builds a shell — perimeter loops on every layer plus solid top and bottom skins — and fills the interior with a sparse lattice whose density the infill percentage sets. At 15 percent, roughly one seventh of the internal volume is plastic and the rest is air.

That lattice has three real jobs. It holds up the top solid layers so they do not sag into the void as they bridge. It resists crushing when something presses on a flat face. And it ties opposing walls together so a thin section does not buckle inward. Notice what is missing: carrying a bending load, which is the failure mode most functional prototypes actually see.

Wall Count Beats Infill Percentage

Bending stiffness depends on how far material sits from the neutral axis of the cross-section. In a printed part the material farthest from the center is the perimeter, which is why every additional wall loop does more structural work than another ten points of infill. Test bars printed at 20 percent infill with four perimeters routinely out-perform bars at 50 percent infill with two perimeters — lighter, faster, and stiffer at once.

The practical rule for a 0.4 mm nozzle: set three to five perimeters, which is 1.2 to 2.0 mm of genuinely solid wall, before you touch the infill slider at all. For a load-carrying bracket, five perimeters with 20 percent gyroid beats two perimeters with 60 percent grid on every metric that matters. And when a part does break, the crack usually runs along a layer boundary rather than through the lattice, which is why print orientation and layer adhesion deserve attention before the fill density does.

Patterns, and What Each One Is Good For

  • Grid and rectilinear — fastest to print, weak in shear, and the crossing extrusions invite nozzle collisions at high density. Fine for cosmetic parts.
  • Gyroid — a smooth self-supporting curve, close to isotropic, never crossing itself within a layer. The best default for functional parts, at roughly a 10 percent time penalty over grid.
  • Cubic and triangular — cubic subdivision skips the deep interior of large parts, cutting time on anything bigger than a fist.
  • Honeycomb — excellent compressive strength perpendicular to the face, slow to print. Worth it for pads and impact absorbers.
  • Concentric — the right choice for elastomers, where a rigid lattice fights the part. Pair it with the settings in TPU printing for flexible parts.
  • Lightning — a branching scaffold that supports only the top skin. Near-zero mechanical value, enormous time savings, perfect for display models.

Choosing a Percentage by Use Case

  1. Appearance and form models: 8 to 12 percent, lightning or grid, three walls. Nobody is going to load it.
  2. Fit and assembly checks: 15 percent, three walls. Enough rigidity that the part does not flex while you are checking clearances, which is the whole point of fit checks before tooling.
  3. Functional brackets and handheld housings: 20 to 25 percent gyroid, four to five walls. This is where the majority of engineering prototypes should live.
  4. Drop-tested and load-bearing parts: 35 to 40 percent, five to six walls, with fillets everywhere a wall meets a floor.
  5. Screw bosses and insert locations: a modifier volume at 100 percent around the feature, not a higher global setting. Hoop stress from a heat-set insert is intensely local, and solid plastic near the hole is what stops the boss splitting.

Modifier volumes are the highest-leverage trick here: most parts have two or three regions needing solid plastic and a large volume needing almost none.

What Each Step Costs in Time and Money

Take an enclosure lid roughly 100 by 60 by 25 mm, PLA on a 0.4 mm nozzle at 0.2 mm layers. At 15 percent infill it prints in about four hours and consumes near 40 grams. At 50 percent, seven and a half hours and 78 grams. At 100 percent, thirteen hours and 150 grams. Stiffness in three-point bending climbs perhaps 20 to 25 percent across the first step, and the jump to solid adds less than people expect.

Filament is the cheap half: commodity PLA runs about $22 per kilogram, so the material delta across that whole range is under $2.50. Machine time is the expensive half — a print service bills roughly $4 to $9 per machine hour, so tripling the duration roughly triples the invoice. Across eight iterations, the gap between a considered profile and a reflexive 60 percent becomes a real line item in what a prototype actually costs.

Where Infill Cannot Help You

No infill setting fixes a part loaded along the Z axis, because the weak plane is the bond between layers and the lattice does not meaningfully cross it. None fixes a wall thinner than two extrusion widths, where there is no interior to fill. And none fixes a thermal problem — PLA softens near 60 degrees Celsius regardless of density, the first argument for a different material, as weighed in ABS versus PLA for prototypes.

Geometry That Beats Any Slicer Setting

Before raising density, add ribs. A 1.6 mm rib standing 8 mm off a flat panel adds more stiffness than filling the entire part solid, at a fraction of the mass. Fillet internal corners to spread stress rather than concentrate it on a sharp layer line, thicken the wall locally where a fastener or hinge lands, and keep overhangs shallow so the printer is not laying material onto air. These habits, collected in our design for 3D printing guidelines, compound in a way slider settings never do.

Getting the Settings Right for Your Part

Projects House prints functional prototypes for US product ventures and tunes wall count, orientation, and local density against the actual load case rather than a default profile. Send your CAD file and how the part gets loaded through our contact form and we will come back with a build recommendation and a price.