NeoSCADResults from other machines
Official NeoSCAD releases, timed by their users with
neoscad bench, most alongside OpenSCAD on the same
machine. Each result is checked automatically against its release
before it appears here.
Results
Loading results…
By platform
Every run
Speedup is OpenSCAD's time over NeoSCAD's: the
geometric mean over the models both finished, as
neoscad bench prints it (above 1× means NeoSCAD is
faster). Geometric mean is NeoSCAD's own time over
the models it finished. Quick runs (--quick) time a subset of
the models once each, so they are only comparable with other quick runs.
| CPU | OS | OpenSCAD | Models | Geometric mean | Speedup | Details |
|---|
Release baseline
Every release times itself on GitHub's shared runners, without OpenSCAD: the same machine types for every version, so these compare releases with each other rather than showing any machine's best.
| Target | CPU | Models | Geometric mean | Cold start | Details |
|---|
How to add yours
Install an official release, v0.2.0 or later (the
downloads, Homebrew, Scoop, the
.deb and .rpm packages or Docker), and run
the command below. Earlier releases don't include
neoscad bench.
neoscad bench --submit
It downloads the release's bench kit, times its models (a few minutes), and shows you the exact result before asking to submit it as an issue on neoscad/benchmarks. If OpenSCAD is installed it offers to time that too; please say yes. A result without OpenSCAD is accepted, but the comparison on the same machine is what makes it most useful.
What's checked, and what's public
- Only official release binaries: the executable's SHA-256 must be the one the release publishes for its platform, and the kit must be the release's own.
- No host name, user name, file path or network address is collected. The result, and your GitHub account as the issue's author, are public.
- The rules, the checks and every result are in the repository; the data is CC0.
This page loads the results' summary from
GitHub (raw.githubusercontent.com) when it opens.