Articles › 3D printing

Strong 3D prints: walls before infill, layer direction and temperature

Why extra walls beat extra infill, with the maths worked out; which infill pattern to pick, temperatures that make layers bond, material choice and a water-bottle test.

Intermediate 7 min read October 21, 2026 3D printing, from choosing a printer to smooth parts · lesson 17/20

What you will learn
  • Parts that bend fail at the outer skin. Going from 2 to 4 walls does more than pushing infill from 15% to 40%, and uses less plastic.
  • The joint between layers is the weak spot. A part printed standing up keeps only about 60-70% of the material's strength across the layers.
  • A hotter nozzle, within the maker's range, bonds layers noticeably better: PETG layer strength almost doubled between 200 and 245 °C in a published test.
  • Gyroid and cubic infill are close to equally strong in every direction; grid and rectilinear print fast but are stronger along one axis.

The hook printed beautifully, held a bag for two days, then snapped at the root. Problems like this rarely come from too little infill, the first thing most people reach for. They come from thin walls, the wrong orientation, or a nozzle run so cold the layers never bonded. This lesson goes through what actually makes an FDM part strong, with numbers you can check.

Figures assume a 0.4 mm nozzle, 0.45 mm line width and 0.2 mm layers, the defaults most slicers ship with.

Where parts break, and what carries the load

Most real parts see bending: hooks, brackets, small beams. When a part flexes, the outer skin is stretched and squeezed hardest while the middle carries almost nothing. It is why a steel beam is an I-shape, not a solid bar.

In an FDM print the walls (perimeters) are that outer skin and infill is the middle. Adding walls puts plastic exactly where the load is. Infill mainly holds up the top surface and stops the walls caving inward.

2 walls · 60% infillheavier, slower, not stronger5 walls · 20% infillstronger with less filament
Most of the load runs in the shell; add walls before adding infill

Worked example: 2 walls vs 4 walls vs dense infill

Take a bracket arm with a 20 × 20 mm cross-section (400 mm²) printed with 0.45 mm lines. Count only the plastic in one layer’s cross-section.

SettingWall thickness per sideWall areaInfill areaTotal plastic
2 walls + 15% infill0.9 mm400 − 18.2² ≈ 69 mm²331 × 0.15 ≈ 50 mm²≈ 119 mm² (30%)
2 walls + 40% infill0.9 mm≈ 69 mm²331 × 0.40 ≈ 132 mm²≈ 201 mm² (50%)
4 walls + 15% infill1.8 mm400 − 16.4² ≈ 131 mm²269 × 0.15 ≈ 40 mm²≈ 171 mm² (43%)

Reading the table:

  1. Going from 2 to 4 walls raises the plastic from 119 to 171 mm², about 13% of the cross-section, and every bit of it sits at the outer skin that resists bending.
  2. Pushing infill from 15% to 40% raises it to 201 mm², about 20% of the cross-section, spread through the middle where the load is small.
  3. So the 4-wall version uses less material, prints faster and, in general, takes more bending load.

This is pure geometry, independent of printer brand. The OrcaSlicer documentation says the same thing: more wall loops improve layer adhesion, strength and rigidity at the cost of print time.

For load-bearing parts, start at 4 walls (about 1.8 mm) before you think about infill. If that is not enough, go to 5 or 6 walls, then look at infill.

Exceptions: parts only 3-4 mm thick are already solid at 4 walls, so infill is moot. And parts loaded in compression through the whole body, such as feet or spacers, do gain from dense infill because the middle carries load too.

Wall settings in the slicer

  • Wall loops / Perimeters: the number of wall lines per layer. Defaults are usually 2-3; use 4-6 for structural parts.
  • Alternate extra wall (OrcaSlicer): adds one more wall on every other layer, so the infill is wedged vertically between walls. The docs say not to combine it with lightning infill or with ensure vertical shell thickness set to All.
  • Infill/wall overlap: lets the infill push slightly into the walls. OrcaSlicer suggests about 10-15%. Too little and the infill and walls separate as a crack line.
  • Wall thickness that divides evenly by the line width, for example 1.8 mm for 0.45 mm lines. Design a 2.0 mm wall and the slicer is left with a thin gap it fills with tiny lines (gap fill), which are weaker than real walls.

Choosing an infill pattern

Once the walls are right, infill gets a say. This table is condensed from the PrusaSlicer documentation.

PatternStrengthTime / materialGood for
GyroidNear-equal in every direction, never crosses itself in a layerFast, moderate materialGeneral structural parts, a good default
Cubic / Adaptive cubicStrong in 3D; adaptive saves about 1/4 of the materialModerateLarge solid parts
GridDense, good layer bonding, but crossings build up bumpsFastEveryday parts that are not critical
RectilinearBasic; supports top layers wellFastest100% infill and decorative parts
HoneycombGood mechanical resistanceUp to twice as slow, about 25% more materialWhen stiffness matters and time does not
LightningWeak, only there to hold up the top surfaceLeastDecorative parts with no load

On our bench, gyroid or cubic at 15-25% with 4-5 walls covers most load-bearing work. Watch the one-direction patterns (rectilinear, line): they are strong along the lines and weak across them, so if you use them you need to know where the load comes from.

Lines within one layer fuse reasonably well. The joint between layers is hot plastic laid onto plastic that has already cooled, so fusion is incomplete, and a part tears along a layer line far more easily than across it.

CNC Kitchen’s test puts numbers on it. PLA specimens printed lying down (pull parallel to the layers) failed around 60 MPa; printed standing up (pull separating the layers) they failed at 37-40 MPa. PETG: 55 MPa lying, 32 MPa standing at best, only about 60% left.

The simple rule: orient the part so the main pull runs parallel to the bed. Print a hook lying on its side, not standing up. If a part must stand, thicken the root that takes the load, or split it and assemble. More on this in the orientation lesson.

Temperature and layer bonding

The hotter the plastic, the better a new layer melts into the one below. CNC Kitchen’s figures (standing specimens, so layer bond strength):

MaterialLow end of rangeHigh end of rangeDifference
PLA200 °C → 37 MPa230 °C → 40 MPaabout +8%
PETG200 °C → 18 MPa245 °C → 32 MPaabout +78%

Past the peak, strength falls again: PLA at 270 °C dropped to 32 MPa and PETG at 260 °C to 24 MPa. Overheating also brings stringing, bubbly surfaces and more fumes. So:

  • For show surfaces, use the low end of the maker’s range. For strength, use the high end.
  • Prusa’s knowledge base suggests, when layers split, raising the nozzle temperature in 5 °C steps and trying the extrusion multiplier in 5% steps, since more material pressed out bonds the layers better.
  • The part cooling fan chills each layer faster and weakens the bond. PLA needs the fan for shape, but ABS should print with the fan off inside an enclosure or it cracks along the layers.
  • Wet filament leaves bubbles at the layer boundary, and the hotter you print, the drier it needs to be (lesson 8).

Pick the material for the load

MaterialTensile strength (typical published tests)StrengthsWeaknesses
PLAabout 50 MPaStiff, easy to print, accurateBrittle, softens at 55-60 °C
PETGabout 40-50 MPaTough, good layer bonding, takes about 80 °CScratches, strings
ABS / ASAabout 34-36 MPaImpact resistant, takes about 100 °CShrinks, splits without an enclosure, fumes

Numbers vary by brand and test method; compare roughly and check the datasheet of the spool you run. The point is that “stiff” and “tough” differ. PLA is stiff but snaps; PETG bends before it breaks. Anything that gets knocked or yanked usually wants PETG over the highest tensile number.

A kitchen-table strength test

You do not need a test machine to compare settings.

  1. Print three identical hooks or small beams, changing one variable, say 2 / 4 / 6 walls, everything else the same.
  2. Clamp one end to a table edge and hang a bag from the other. Add 500 ml water bottles (about 0.5 kg each) until it breaks. Note the count.
  3. Look at the break. Clean along a layer line means layer bonding (orientation, temperature, fan). Ragged across layers means not enough material (walls).

Wear safety glasses, stand back, and keep anything valuable out from under the test piece. The counts only compare settings on the same printer and spool; they are not universal values.

Common mistakes

  • Cranking infill to 60-80% and hoping while still using 2 walls. Print time doubles for a small gain in strength.
  • Printing a hook standing up to avoid supports, then watching it snap along a layer at the root.
  • Lowering the temperature to cure stringing until the layers stop bonding. Fix stringing with retraction and dry filament instead.
  • A 2.0 mm wall with 0.45 mm lines, leaving a gap for gap fill. Make it 1.8 or 2.25 mm.
  • Lightning infill on a structural part. It exists only to hold up the top skin.
  • Printing ABS with the fan on full outside an enclosure, and getting a part full of layer cracks.

The next lesson is about speed, which hits layer bonding directly: faster means the plastic has less time to melt together.

Sources

FAQ

Should I just use 100% infill?

Usually not worth it. Print time and material climb a lot while strength gains less than adding walls would give. If you really need solid, for a small part in compression, use rectilinear, which the slicer switches to automatically at 100%.

Why does my part always break along a layer line?

Each new layer lands on one that has already cooled, so the plastic never fully fuses. That joint is weaker than a continuous line. Orient the part so the main pull runs parallel to the layers, print hotter, and cut cooling where the material allows.

Which is strongest, PLA, PETG or ABS?

PLA has the highest tensile figure but is brittle and softens at 55-60 °C. PETG is tougher with better layer bonding. ABS tests lower on tensile strength but takes heat and impact. The answer depends on the job, not on one number.

Circuits and code were tested on sample boards. Boards differ in pin layout, so check your board's pinout before you power anything.