Home › Evidence › Records › eng-0065
⚠ This notebook has stopped taking notes: newest record is 23 days old (eng-0280), against an expected cadence of 14 days.

eng-0065

window recorded
citable URL: https://halobench.com/records/eng-0065/ — this address never moves; the anchor /records/#eng-0065 keeps resolving
5.56 Whmean 120.7 W − 10.1 W idle floor (box-idle-all-empty) = 110.6 W attributable
window 2026-08-14T20:29:53Z → 2026-08-14T20:32:39Z · provenance recorded
method wall-meter · aihydra energy sensor via Home Assistant (counter-difference on sensor.hardware_ai_hydra_energy) · no contention
covers run-0245 run-0246
aihydra run-meta.jsonl entries rocm-q8-d0-rep1/2/3 (qwen38-screen), suite qwen38-screen, cfg-0062, depth 0. 3 fresh-process llama-bench reps merged into one window (page cache dropped between reps, per run-0245/run-0246 comment). Retroactive join computed 2026-08-15 — the qwen38-27b bench published 2026-08-14 night with no energy join run against it; this closes that gap.

Wh, never joules — read as a cumulative-counter difference across the window; delta over the idle baseline is the only figure that means anything

Cited by — computed at build time, never stored