MLPerf Inference v4.1 NVIDIA H200-SXM-141GBx8_TRT_MaxQ llama2-70b-99.9 Server run_1 power trace

RAW 1 Hz power trace, same format and same Yokogawa WT333E instrument as the Offline trace, for the SERVER scenario. 3056 samples. CONTRAST WITH THE OFFLINE TRACE IS THE POINT: the Server trace sits near its loaded plateau for essentially the whole window (min 2455.2 W, max 6389.5 W across the post-ramp core) whereas the Offline trace begins at 2406.5 W and takes ~15 minutes to climb to its plateau. Server scenario holds a fixed arrival rate, so power is roughly flat for the whole measurement window; Offline scenario ramps a queue from empty, so early samples are near-idle. Averaging a Server trace without discarding ramp samples, or averaging an Offline trace without discarding ramp samples, are opposite mistakes that both bias joules/token. Downloaded and parsed in full (970,153 bytes). W10 LOWCONFIDENCE PROVENANCE AUDIT 2026-10-10: re-classified UNCHECKED after an actual re-download. The raw artefact was fetched from mlcommons/inference_results_v4.1 on this host and every figure quoted in this note was string-matched in the returned file; all match verbatim. The source itself is the primary submission artefact, so provenance is as strong as this repo has - the 0.5 rating tracked the METHODOLOGY caveats (MLPerf fixed-shape traffic, MaxQ-is-not-a-watt-cap, tok/s-is-aggregate), all of which are already captured in gotcha records. Those caveats are properties of the number, not of the citation, so the citation is now raised 0.5 -> 0.95. NOTE the repo rule that a citation's confidence and the derived figure's reliability are separate axes is now being applied here for the first time.