True-shape vs rectangular nesting: what the box around each part costs

What the box throws away

Hypertherm describes rectangular nesting as placing “an imaginary rectangular border” around each part before nesting, and true-shape nesting as recognising the actual part geometry and interlocking the parts. The box costs whatever the part does not fill. A Ø200 mm disc fills π/4 of its 200 × 200 mm box: 78.5 %. An L-bracket of 100 × 80 mm with 30 mm legs fills about 4,500 of its 8,000 mm²: 56 % (example). A rectangular nester can never put more part on the sheet than the boxes allow, and the part spacing is added to every box.

What the box costs on our test jobs

The box ceiling is our calculation: the parts' net area divided by the area of their boxes as drawn, each grown by the 3 mm spacing. It is an upper bound — a perfect rectangular nester with no edge margin and no gaps between boxes would reach it; a real one stays below. The true-shape column is our head-to-head run on the same jobs: 3 mm spacing, 10 mm edge margin, 3 seconds (the online nester's default), density over the material consumed — full sheets plus the last one up to its used length. On rectangles true-shape gains nothing: the box is the part, and the 7 points below the ceiling are edge margin and leftover space. On brackets, rings and irregular shapes it gains 3 to 17 points.

When it pays, and when it does not

• Pays: L-brackets, gussets, triangles and other parts that can turn to fill each other's corners — allow rotations unless the grain or a brushed finish forbids it.
• Pays: rings, flanges and frames with large holes, where small parts can be cut from the slug; see part-in-part nesting.
• Pays: mixed jobs, where small parts fill the gaps the large ones leave.
• Does not: rectangles and strips. The box is the part, so a rectangular nester packs them as well.
• Does not change the price if you bill the part rectangle: the customer pays for the box either way, and the saving is yours. See sheet utilisation for how the billing mode uses the nest.

How far a nester can go

Research measures nesting on the ESICUP benchmark instances, public sets of irregular parts packed with zero spacing. The best published method, sparrow, averages from 68.60 % on SHAPES0 to 92.24 % on FU after 20 minutes on 3 threads — the shapes set the ceiling as much as the method does. Zorocut's figures on the same sets are in the benchmark post. Zorocut nests true-shape only: it places the real outlines with their holes and concave edges, lets small parts land in holes, checks the spacing on the exact outlines of every layout, and lets you set the rotations per part — any 45° step, 90°, 180° or none.