Walls vs. infill: where should you add material?

Know which part of the structure you are changing

Walls, often called perimeters, follow the outline of each layer. Top and bottom solid layers close the shell. Sparse infill occupies the space inside it. A setting of 20% infill therefore does not mean the complete object is 20% plastic: its shell remains separate from that percentage.

Prusa’s infill documentation identifies support for the top layers as a central job of infill, alongside effects on stiffness, weight, and material consumption. A hollow decorative shape that closes gradually is a different problem from a wide, flat lid whose roof must span the interior.

Slice first and hide the upper layers. Look at what remains inside. A thin rib may consist entirely of walls, so changing global infill may barely affect that rib.

Match the change to the symptom

A broken layer bond will not necessarily improve because the middle becomes denser. Similarly, a bent part can need a different shape rather than more plastic in the same outline. The table organizes possible experiments; it does not predict strength gains.

Why wall count is only part of wall thickness

Prusa recommends examining perimeters when seeking stronger parts, but count alone does not define the physical shell. Extrusion width, overlapping paths, and the available model thickness matter. A model that is too thin cannot acquire unlimited extra walls without changing its geometry.

As a rough planning illustration, three nominal 0.45 mm paths suggest 1.35 mm before accounting for how the slicer spaces and overlaps them. That multiplication is not a CAD tolerance or a measurement of the finished wall.

PrusaSlicer’s Arachne generator can vary extrusion width around geometry. Check the preview's actual paths instead of assuming every perimeter is exactly one nozzle diameter wide. This is especially useful around text, thin partitions, and small holes.

Check top thickness after changing layer height

Layer count and physical thickness are different quantities. In a simplified constant-height example, five 0.20 mm top layers make 1.00 mm, while five 0.12 mm layers make 0.60 mm. Those are geometric totals, not a recommendation for either roof.

If you reduce layer height and leave a top-layer count unchanged, review the resulting top shell. Some profiles specify minimum thickness as well as layer count; the exact control depends on the slicer. Examine where the roof starts and how much support is below it before changing infill and top layers together.

Our layer-height guide explains the separate detail and time decision. Keep that choice stable while evaluating the structure of a part.

Run a small comparison with a clear question

Suppose a lightweight desk tray flexes too much at its rim. Save a baseline and make one copy with an additional wall, keeping infill unchanged. Make a second copy with denser infill, keeping the original walls. These are proposed experiments, not tested settings.

1. Slice all three and record grams, estimated time, and the visible structure near the rim.
2. Reject a variant that does not change the region you care about.
3. Print small representative sections if they preserve the behavior you need to judge.
4. Compare fit and flex under the same ordinary use conditions, without treating this as a load certification.

Use the cost calculator with each complete job estimate. Choosing by infill percentage alone can hide the extra material in walls and supports.

Revisit the design when settings stop helping

UltiMaker’s discussion of strength distinguishes properties such as stiffness and impact behavior and explains why orientation matters. A part that must resist bending does not necessarily need the same material or structure as a part that must survive a knock.

For a hook or bracket, sketch where the force enters and where the part is restrained. Check whether the intended use tends to separate layers. Our orientation guide gives a way to compare candidates before adding more material.

Keep your successful project, including material and orientation, alongside the result. “Four walls” is not enough information to reproduce it. If a failure could injure someone or damage valuable equipment, a successful hobby test is not evidence of an adequate engineering design.