eng-0231
window recorded
citable URL: https://halobench.com/records/eng-0231/ — this address never moves; the anchor /records/#eng-0231 keeps resolving
49.65 Whmean 161.03 W − 10.1 W idle floor (box-idle-all-empty) = 150.93 W attributable
window 2026-08-22T23:05:23Z → 2026-08-22T23:24:06Z · provenance recorded
method wall-meter · aihydra energy sensor via Home Assistant (counter-difference on sensor.hardware_ai_hydra_energy, raw ~10s samples linearly interpolated to the recorded window) · peak 180.58 W · no contention
per task 1.9096 Wh · — p at 30.3 p/kWh
covers run-0571
HO-009-AB n2/MTP arm tau2 full-26 (reasoning OFF). Counter interpolated from 33.202848 to 33.252787 kWh across the n2 tau2 window 23:05:23-23:24:06Z on 2026-08-22 (first task start to last task end, from tau2-full26-summary.json). 0/26 tasks passed, so pence_per_correct_answer is null (no correct-task denominator; a zero-count capability cell). wh_per_task uses 26. Power profile shows a single isolated active band (idle ~11.5W before 23:05Z and after 23:24Z) under the HO-009-AB claim, no contention. Build 7077abb, ROCm0/gfx1151, f16/f16 KV, --spec-type draft-mtp --spec-draft-n-max 2, -rea off. Plain-vs-n2 capability comparison only; the higher mean_w_active reflects the same idle baseline subtracted and a slightly shorter wall window, not a throughput claim.
Wh, never joules — read as a cumulative-counter difference across the window; delta over the idle baseline is the only figure that means anything