Free, open-source 3D sculpting for Linux, macOS and Windows: dense meshes, remeshing, painting, built-in AI (image/text to 3D), Send to Unreal, game-ready LODs, Gaussian splats and OpenUSD.
Perf scoreboard (0.11.0 beta.3)
One table for every performance claim in the beta.3 push. A number goes here only if it was measured with the method below. Claims in notes, the site or the guide quote this table, and quote the floor first. Owner: core200m (it updates the table as branches land). Public text states our own numbers only, never another app's.
Targets (the maintainer, 11-04: performance is the #1 priority until they're met)
Our own numbers only. Each must hold on the floor tier first, then the E-core floor, the laptop and the Windows 3060.
| Target | |
|---|---|
| (a) Base mesh 15-30M | committed lag p50 < 50 ms and p99 < 100 ms at 120 and 240 px; release < 100 ms |
| (b) Micro levels 100-450M (Lucy 112M / 448M, core-200m) | the same lag targets; orbit >= 30 fps right after strokes; first frame < 3 s |
| (c) First dab | no stall: < 50 ms (NVIDIA first-dab stall) |
Gap to the targets (11-04)
p50 / p99 / release in ms (240 px, the worse of 120 and 240 px); x = how far off the p50 target. E-core and laptop for 15M and core-200m are the beta.2 baseline below; Lucy is the laptop run of 11-04 (offscreen) with fix/micro-pick-reach. Floor iGPU: no machine.
| Target | Scene | E-core floor | Laptop | Windows 3060 | Owner / lever |
|---|---|---|---|---|---|
| (a) | 15M head | 552 / 2036 / 3355 (11x) | 405 / 577 / 230 (8x) | 447 / 737 / 249 (9x) | betaboost: Fast Big Brushes, batches capped by time |
| (a) | Lucy 28M | - | 359 / 623 / 900 (7x) | running | betaboost (same path) |
| (b) lag | core-200m | 1498 / 1735 / 4675 (30x) | 319 / 511 / 452 (6x) | 440 / 565 / 827 (9x) | n2stroke: N2 + big brushes on the base level |
| (b) lag | Lucy 112M / 448M | - | 1700-2656 / 3240-3448 / 10-11 s (34-53x) | running | n2stroke: base-level edits (the 256x rescale); core200m: picks (fix/micro-pick-reach) |
| (b) orbit | core-200m | 10 fps (3x) | 27 fps (1.1x) | 32 fps (met) | core200m: micro renderer and catch-up |
| (b) orbit | Lucy 112M / 448M | - | 1.6-1.7 fps (18x) | running | core200m: micro renderer GPU cost on a dense base (63-70 ms/frame before any stroke) |
| (b) first frame | core-200m | - | 0.7 s (met) | 0.7 s (met) | - |
| (b) first frame | Lucy 112M / 448M | - | 9.0 / 11.1 s (3-4x) | running | core200m: micro open path on a dense base |
| (c) | first dab, NVIDIA | - | - | betaboost's numbers | betaboost |
Tiers, in the order a claim must pass them
- Floor iGPU: a real Windows laptop with integrated graphics (Intel UHD / Iris Xe class, or a small AMD APU). No such machine yet; the tester request is drafted in research/micro-gpu-shade REPORT.md. Until it reports, this tier is empty and nothing claims it.
- E-core floor (the gate today):
SF_HW_SIMULATE=floor SF_GOVERNOR=0 taskset -c 0,1on this laptop (M1, Asahi Linux): the floor's budgets, GL 3.3, on the two efficiency cores (about half a fast core each). This is the pessimistic CPU bound; a CPU-bound claim counts only if it holds here (perf-moonshots §0 item 6). - Laptop: the same machine, no simulation, all cores (Apple M1 Pro, 32 GB shared, through Asahi's GL driver).
- Windows RTX 3060 (Ryzen 5600X, 32 GB):
~/Work/.sf/win.sh, PC idle (win.sh holds the PC lock). A sanity check: a result that holds only here hasn't hit its target.
Method
- Scenes:
--benchmark standard-15m(the 15M head, multires mesh path),--benchmark core-100mand--benchmark core-200m(the same head at micro level 5 with procedural wrinkles and pores, 100M and 200M micro triangles in memory). Each runs Clay at 25, 120 and 240 px, then orbits for 10 s (core) or 3 s (15M). - Committed lag: pointer sample to the swap of the frame whose mesh contains that sample's dab (
lag_ms). Release: pen up to the last dab in the mesh and shaded (release_done_ms). Orbit: frames per second over the orbit phase. Memory: peak resident RAM plus GPU buffers. - Defaults of the build under test (no toggles), 6 s strokes, each run started at load < 2 on the laptop, at least two interleaved runs per cell; a cell shows the median.
- Rows from
python3 tools/perf_scoreboard.py TIER results/*.json. - From 11-04 every laptop run is offscreen (
SDL_VIDEO_DRIVER=offscreen, the same GPU, no window or compositor). The beta.2 baseline below was windowed, so each candidate is re-measured against main offscreen, interleaved, and the baseline rows are replaced by those offscreen runs.
Shipping today: main 0.11.0-beta.2 defaults (11-04)
Committed lag p50 / p99, then release (pen up to the stroke in the mesh), in ms; median of the runs. Floor iGPU: no machine yet. E-core and laptop: 3 runs each (perf/scoreboard e30349e sources; the build reports its parent 8d89d61c); Windows: 2 runs on an idle PC (e30349e). Bold: meets the target (p50 within 16 ms of the spacing floor, p99 < 33 ms, release < 100 ms).
| Mesh | Brush | Floor iGPU | E-core floor (the gate) | Laptop | Windows RTX 3060 |
|---|---|---|---|---|---|
| 15M | 25 px | - | 19 / 174, 156 | 14 / 109, 51 | 8 / 59, 45 |
| 15M | 120 px | - | 550 / 884, 608 | 177 / 259, 81 | 170 / 255, 73 |
| 15M | 240 px | - | 552 / 2036, 3355 | 405 / 577, 230 | 447 / 737, 249 |
| 100M micro | 25 px | - | 24 / 79, 131 | 14 / 28, 23 | 6 / 24, 33 |
| 100M micro | 120 px | - | 240 / 390, 521 | 113 / 186, 68 | 102 / 183, 107 |
| 100M micro | 240 px | - | 829 / 1254, 2747 | 286 / 479, 338 | 298 / 437, 404 |
| 200M micro | 25 px | - | 25 / 90, 109 | 18 / 60, 42 | 6 / 46, 49 |
| 200M micro | 120 px | - | 466 / 582, 960 | 134 / 225, 171 | 122 / 215, 312 |
| 200M micro | 240 px | - | 1498 / 1735, 4675 | 319 / 511, 452 | 440 / 565, 827 |
Orbit (fps, the orbit phase after the strokes) and memory (peak RAM + GPU buffers):
| Mesh | E-core floor | Laptop | Windows RTX 3060 | Memory (laptop / Windows) |
|---|---|---|---|---|
| 15M | 90 | 182 | 319 | 2.3 GB + 233 MB / 2.4 GB + 211 MB |
| 100M micro | 13 | 32 | 47 | 0.8 GB + 330 MB / 0.9 GB + 330 MB |
| 200M micro | 10 | 27 | 32 | 1.3 GB + 640 MB / 1.4 GB + 640 MB |
What it says:
- Small brushes keep up at p50, not at p99. 25 px p50 is 6-25 ms on every tier (the target or within a few ms of it on the E-core floor); p99 meets 33 ms only on 100M (laptop, 3060) and is 46-174 ms elsewhere.
- Big brushes fail everywhere, the 3060 included. 120 px is 100-550 ms behind the pen, 240 px 290-1500 ms. On the micro scenes the E-core floor places 9-22 dabs in a 6 s stroke at 120-240 px (the brush can't keep up, so dabs merge): the stroke lands, but late and coarse. This is what N2 (micro levels) and Fast Big Brushes (15M) are for.
- Orbit after sculpting on micro levels is 10-47 fps (it opens and orbits at 60+ fps before any stroke, per docs/design/core-200m.md §8): the shading catch-up after the strokes runs during the orbit. 15M orbits at 90+.
- 200M fits a 16 GB integrated-GPU laptop: 1.3 GB RAM + 640 MB GPU.
- Noise: the laptop is shared with other sessions (load 4-13 during runs; the start waits for load < 2). One
E-core round of three was 2-7x worse on the micro scenes and is outvoted by the median; ranges per cell are in
the raw rows (
tools/perf_scoreboard.py). The Windows runs agree within 10%. - Windows harness note: the micro scenes have to start through
cmd /c(win.sh with anySFENV); a direct launch from the scheduled task exits silently before writing the scene (not yet understood).
Candidates for beta.3
Each candidate is measured against the row above on the same tier, interleaved, before it can be turned on.
| Branch (owner) | What | Status |
|---|---|---|
| perf/n2-wavelet (n2stroke) | N2 wavelet strokes on micro levels, default on; GPU shading (fragment shader) | in progress |
| perf/beta-boost-2 (betaboost) | Fast Big Brushes with fused batches capped by time; NVIDIA first-dab stall | in progress |
| research/core-200m (core200m) | N1 screen-space strokes under 1 ms a dab | 6d410e1f: 0.58-0.73 ms GPU per 240 px dab on the laptop (was 4.2-4.8), live on the floor 9.6 vs the ghost's 8.3 ms GPU; research, off (SF_SCREEN_STROKE=1) |