eng-0084
window recorded
citable URL: https://halobench.com/records/eng-0084/ — this address never moves; the anchor /records/#eng-0084 keeps resolving
76.06 Whmean 157.63 W − 10.1 W idle floor (box-idle-all-empty) = 147.53 W attributable
window 2026-08-15T10:48:12Z → 2026-08-15T11:17:09Z · provenance recorded
method wall-meter · aihydra energy sensor via Home Assistant (counter-difference on sensor.hardware_ai_hydra_energy) · no contention
per task 19.0149 Wh · 0.5761 p at 30.3 p/kWh
covers run-0274
Nemotron-3-Super-120B-A12B Q4_K_M SMOKE arm (run-0274, cfg-0072), the fair-quant re-screen this candidate had been waiting on since 2026-08-09. wh_per_task is Wh-per-CORRECT-ANSWER: wh_total / tasks_passed (4 of 5, run-0274.metrics). Counter samples nearest the window bounds: start 19.1516851633787076 kWh (2026-08-15T10:48:09Z, 3s before window open — nearest recorded tick), end 19.2277447432279476 kWh (2026-08-15T11:17:09Z, exact). At 157.6 W mean active draw this is a notably heavier arm than the 35B-class models in the corpus — consistent with Nemotron-3-Super's own throughput being the slowest measured on this box (clm-0035). Retroactive join computed 2026-08-15, same day as the run.
Wh, never joules — read as a cumulative-counter difference across the window; delta over the idle baseline is the only figure that means anything