MLPerf Inference v4.1 NVIDIA H200-SXM-141GBx8_TRT_MaxQ llama2-70b-99.9 Server LoadGen summary

AUDITED MLPerf LoadGen result for the paired Server power trace. Key fields verbatim: 'Completed samples per second : 79.36', 'Completed tokens per second: 23113.14', 'Result is : VALID', 'Performance constraints satisfied : Yes', scheduled samples/s 80.10, target_qps 80, ttft 2000000000 ns, tpot 200000000 ns, min_duration 2400000 ms, min_query_count 100. Latency percentiles are the point at which this run is latency-constrained rather than throughput-constrained: TPOT 50th 138.267 ms, 99th 158.802 ms, 99.9th 174.808 ms against a 200 ms constraint; TTFT 50th 204.7 ms, 99th 519.3 ms, 99.9th 658.8 ms against a 2000 ms constraint. Use 'Completed tokens per second' for this scenario, not 'Tokens per second' - the Offline and Server summaries do not use the same label for their headline figure. 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.