⏳ Sandglass v3 — test bench

512 KiB chained fill + 4 interleaved dependent walks (2M steps). Compare browser tabs vs GPUs vs native CPUs — the number that matters is the ratio between devices (and per-joule, if you enter watts).

This device:
self-test digest: computing… (must match bench_gpu.py output)

Tip: use the max button — v3 workers only hold 512 KiB each, so all cores is safe on any device. The Argon2 baseline uses the same openpgpjs/argon2id WASM as the real browserCoin miner (one fixed 65 MB per worker, warmed up sequentially) — if memory runs out it continues with as many workers as fit; the status line shows the count used. Results auto-submit to the shared list below.

Shared results
timelabeldevicepathalgoworkersH/sWH/J