{"service":"echosizer-web","source":"engine-snapshot","note":"v0.1 web shell serves golden CLI snapshots. Live engine runs via `echosizer run`.","record":{"id":"interactive-8b-400","as_of":"2026-08-27","winner":{"gpu_id":"h200-sxm","tp":1,"batch_vram":227,"batch_tpot":6,"batch_sla":6,"tok_s_user":35.15,"ttft_s":0.0038,"gpus":1,"topology":"colocated","binding_constraint":"NONE"},"server":{"pattern_id":"pcie-8gpu","pattern_name":"Generic 8× GPU PCIe Node","nodes":1,"gpus_installed":8,"gpus_per_node":8,"n_plus_1":false,"ram_gib_per_node":1536,"nvme_tb_per_node":7.68,"network_gbps":200,"nic_class_gbps":200,"t0_gib_per_gpu":113.9,"t1_gib_per_node":455.6,"t2_tb_per_node":3.559375,"t3":false,"node_kw":6.6,"it_kw":6.6,"facility_kw":9.2,"rack_u":4,"warnings":[]},"concurrency":{"lambda_rps":0.3333,"service_time_s":10.3,"mean":3.43,"design":6,"iterated":false},"memory":{"weight_gib":7.6,"kv_bytes_per_token":65536,"seq_reserved":8192,"kv_per_request_gib":0.5},"assumptions":["decode_eff=0.50 ESTIMATE","prefill_eff=0.55 ESTIMATE"],"rejected":[],"sensitivity":[{"knob":"C*","low":"h200-sxm","base":"h200-sxm","high":"h200-sxm","sku_family_changed":false},{"knob":"S","low":"h200-sxm","base":"h200-sxm","high":"h200-sxm","sku_family_changed":false},{"knob":"h","low":"h200-sxm","base":"h200-sxm","high":"h200-sxm","sku_family_changed":false},{"knob":"R","low":"h200-sxm","base":"h200-sxm","high":"h200-sxm","sku_family_changed":false},{"knob":"precision","low":"h200-sxm","base":"h200-sxm","high":"h200-sxm","sku_family_changed":false}]}}