eng-0119
window recorded
citable URL: https://halobench.com/records/eng-0119/ — this address never moves; the anchor /records/#eng-0119 keeps resolving
41.36 Whmean 165.44 W − 10.1 W idle floor (box-idle-all-empty) = 155.34 W attributable
window 2026-08-16T07:39:00Z → 2026-08-16T07:54:00Z · provenance recorded
method wall-meter · aihydra energy sensor via Home Assistant (counter-difference on sensor.hardware_ai_hydra_energy, raw ~10s-resolution history samples, linearly interpolated to the run-meta window) · no contention
aihydra laguna-s21-fullbench throughput matrix cell rocm-d131072 (N=1 rep, pp1024/tg256 at depth 131072 -- the deepest cell either backend completed this job). Counter interpolated from raw ~10s-resolution HA history samples pulled fresh this session: start 21.675783821154234 kWh, end 21.717143484197067 kWh. Recomputed independently from HA raw history; matches the MODEL-PAGE.md draft to within rounding (41.36 Wh / 165.4 W). Highest mean power in the matrix -- this cell's 900s window is dominated by prefill at depth on a 118B/~8B-active hybrid model.
Wh, never joules — read as a cumulative-counter difference across the window; delta over the idle baseline is the only figure that means anything