{ "article_key": "llm-235b-hybrid", "cases": [ { "case_wall_seconds": 1352.721352025983, "completed_response_resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 87432.48105413717, "reason": null, "response_union_seconds": 882.8749595810659, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 99.03155606046961, "trapezoid_pieces": 893, "watt_hours": 24.28680029281588 }, "gpu_sample_count": 883, "host_sample_count": 434, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86956191744, "mean": 86912404852.79263, "median": 86905057280.0, "minimum": 86897274880, "sample_count": 434 }, "gpu_board_watts": { "maximum": 137.35, "mean": 99.02436013590034, "median": 95.14, "minimum": 88.97, "sample_count": 883 }, "gpu_memory_used_bytes": { "maximum": 94157930496.0, "mean": 94157089735.17554, "median": 94157930496.0, "minimum": 94155833344.0, "sample_count": 883 }, "gpu_sm_clock_mhz": { "maximum": 2317.0, "mean": 2284.4246885617213, "median": 2280.0, "minimum": 1867.0, "sample_count": 883 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 34.90939977349943, "median": 10.0, "minimum": 2.0, "sample_count": 883 }, "host_mem_available_bytes": { "maximum": 95637037056, "mean": 95531510581.08755, "median": 95521652736.0, "minimum": 95466434560, "sample_count": 434 }, "process_pss_bytes": { "maximum": 55339509760, "mean": 55163167262.67281, "median": 55336781824.0, "minimum": 54836813824, "sample_count": 434 }, "process_rss_bytes": { "maximum": 55342489600, "mean": 55166208018.87558, "median": 55339761664.0, "minimum": 54839808000, "sample_count": 434 } }, "process_sample_count": 434, "response_union_seconds": 882.8749595810659, "sampled_counter_deltas": { "deltas": { "major_faults": 0, "read_syscall_bytes": 0, "storage_read_bytes": 0, "storage_write_bytes": 360448 }, "observed_endpoint_seconds": 859.7034017620608, "observed_intervals": 10, "requested_intervals": 10 } }, "groups": [ { "aggregate_output_tokens_per_second": 3.6192757941330425, "completed_output_tokens": 1280, "completed_request_count": 5, "context_tokens": 2048, "cpu_snapshot_intervals": { "host_capacity_percent": 38.159420998783055, "interval_count": 5, "interval_seconds": 353.6806808580877, "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.", "one_logical_cpu_percent": 1221.1014719610578, "process_cpu_seconds": 4318.8 }, "failed_request_count": 0, "measured_rounds": 5, "mode": "completion", "resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 33952.31959522913, "reason": null, "response_union_seconds": 353.66191271605203, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 96.00219411381389, "trapezoid_pieces": 359, "watt_hours": 9.431199887563647 }, "gpu_sample_count": 354, "host_sample_count": 167, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86956191744, "mean": 86928198306.49101, "median": 86925025280, "minimum": 86913802240, "sample_count": 167 }, "gpu_board_watts": { "maximum": 136.38, "mean": 95.9926836158192, "median": 94.245, "minimum": 88.97, "sample_count": 354 }, "gpu_memory_used_bytes": { "maximum": 94155833344.0, "mean": 94155833344.0, "median": 94155833344.0, "minimum": 94155833344.0, "sample_count": 354 }, "gpu_sm_clock_mhz": { "maximum": 2317.0, "mean": 2284.135593220339, "median": 2280.0, "minimum": 1957.0, "sample_count": 354 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 23.166666666666668, "median": 10.0, "minimum": 2.0, "sample_count": 354 }, "host_mem_available_bytes": { "maximum": 95637037056, "mean": 95563106218.15569, "median": 95547498496, "minimum": 95503069184, "sample_count": 167 }, "process_pss_bytes": { "maximum": 54901213184, "mean": 54883534688.57485, "median": 54899740672, "minimum": 54836813824, "sample_count": 167 }, "process_rss_bytes": { "maximum": 54904193024, "mean": 54886587704.71856, "median": 54903541760, "minimum": 54839808000, "sample_count": 167 } }, "process_sample_count": 167, "response_union_seconds": 353.66191271605203, "sampled_counter_deltas": { "deltas": { "major_faults": 0, "read_syscall_bytes": 0, "storage_read_bytes": 0, "storage_write_bytes": 143360 }, "observed_endpoint_seconds": 341.29245696699945, "observed_intervals": 5, "requested_intervals": 5 } }, "response_outer_span_seconds": 353.68840590899345, "response_seconds": { "maximum": 70.85927938303212, "mean": 70.7323825432104, "median": 70.71976610599086, "minimum": 70.65188393200515, "sample_count": 5 }, "response_window_union_seconds": 353.66191271605203, "server_decode_tokens_per_second": { "maximum": 4.72037786569281, "mean": 4.705007937571641, "median": 4.704372092144592, "minimum": 4.692452607700804, "sample_count": 5 }, "synthetic_first_token_seconds": { "maximum": 16.539735411002766, "mean": 16.512870826828294, "median": 16.516706649039406, "minimum": 16.472258550056722, "sample_count": 5 }, "warmup_completed_rounds": 1, "warmup_requests_per_round": 1 }, { "aggregate_output_tokens_per_second": 2.4186856457574994, "completed_output_tokens": 1280, "completed_request_count": 5, "context_tokens": 8192, "cpu_snapshot_intervals": { "host_capacity_percent": 26.53628150643171, "interval_count": 5, "interval_seconds": 529.2353224620456, "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.", "one_logical_cpu_percent": 849.1610082058147, "process_cpu_seconds": 4494.06 }, "failed_request_count": 0, "measured_rounds": 5, "mode": "completion", "resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 53480.16145890802, "reason": null, "response_union_seconds": 529.2130468650139, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 101.05601472926116, "trapezoid_pieces": 534, "watt_hours": 14.855600405252227 }, "gpu_sample_count": 529, "host_sample_count": 267, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86921846784, "mean": 86902526550.29213, "median": 86901604352, "minimum": 86897274880, "sample_count": 267 }, "gpu_board_watts": { "maximum": 137.35, "mean": 101.0531190926276, "median": 95.67, "minimum": 90.98, "sample_count": 529 }, "gpu_memory_used_bytes": { "maximum": 94157930496.0, "mean": 94157930496.0, "median": 94157930496.0, "minimum": 94157930496.0, "sample_count": 529 }, "gpu_sm_clock_mhz": { "maximum": 2317.0, "mean": 2284.618147448015, "median": 2280.0, "minimum": 1867.0, "sample_count": 529 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 42.76748582230624, "median": 10.0, "minimum": 2.0, "sample_count": 529 }, "host_mem_available_bytes": { "maximum": 95557570560, "mean": 95511748515.95506, "median": 95512559616, "minimum": 95466434560, "sample_count": 267 }, "process_pss_bytes": { "maximum": 55339509760, "mean": 55338068535.61049, "median": 55338133504, "minimum": 55335350272, "sample_count": 267 }, "process_rss_bytes": { "maximum": 55342489600, "mean": 55341101623.61049, "median": 55341113344, "minimum": 55339749376, "sample_count": 267 } }, "process_sample_count": 267, "response_union_seconds": 529.2130468650139, "sampled_counter_deltas": { "deltas": { "major_faults": 0, "read_syscall_bytes": 0, "storage_read_bytes": 0, "storage_write_bytes": 217088 }, "observed_endpoint_seconds": 518.4109447950614, "observed_intervals": 5, "requested_intervals": 5 } }, "response_outer_span_seconds": 529.246267170005, "response_seconds": { "maximum": 105.9342662889976, "mean": 105.84260937300277, "median": 105.85307992796879, "minimum": 105.70028573501622, "sample_count": 5 }, "response_window_union_seconds": 529.2130468650139, "server_decode_tokens_per_second": { "maximum": 4.717623238471859, "mean": 4.707775473637765, "median": 4.707951723075252, "minimum": 4.6967906663607675, "sample_count": 5 }, "synthetic_first_token_seconds": { "maximum": 51.79277144599473, "mean": 51.67663359480211, "median": 51.64751424099086, "minimum": 51.63793887099018, "sample_count": 5 }, "warmup_completed_rounds": 1, "warmup_requests_per_round": 1 } ], "id": "capacity-235b", "model": "qwen3-235b-q5", "module": "hybrid-capacity", "placement": { "buffers": [ { "device": "CPU_Mapped", "reported_mib": 27759.73 }, { "device": "CPU_Mapped", "reported_mib": 28323.86 }, { "device": "CPU_Mapped", "reported_mib": 18322.9 }, { "device": "CUDA0", "reported_mib": 86732.62 } ], "gpu_layers": 95, "meaning": "Runtime buffer allocation is not measured resident process RAM.", "requested_cpu_expert_layers": 44, "total_layers": 95 }, "record_sha256": "3a49e3e92a194f228457843e1a3264cd5a2af9e3e1302b9862ba46d0660bdd2a", "resources_by_phase": { "baseline": { "gpu_energy": { "available": false, "joules": null, "reason": "unbracketed_boundary" }, "gpu_sample_count": 2, "host_sample_count": 2, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86569291776, "mean": 86560978944, "median": 86560978944.0, "minimum": 86552666112, "sample_count": 2 }, "gpu_board_watts": { "maximum": 16.06, "mean": 16.005, "median": 16.005, "minimum": 15.95, "sample_count": 2 }, "gpu_memory_used_bytes": { "maximum": 2097152.0, "mean": 2097152.0, "median": 2097152.0, "minimum": 2097152.0, "sample_count": 2 }, "gpu_sm_clock_mhz": { "maximum": 180.0, "mean": 180.0, "median": 180.0, "minimum": 180.0, "sample_count": 2 }, "gpu_utilization_percent": { "maximum": 0.0, "mean": 0.0, "median": 0.0, "minimum": 0.0, "sample_count": 2 }, "host_mem_available_bytes": { "maximum": 96450252800, "mean": 96438243328, "median": 96438243328.0, "minimum": 96426233856, "sample_count": 2 }, "process_pss_bytes": null, "process_rss_bytes": null }, "process_sample_count": 0, "response_union_seconds": 1.100754403974861, "sampled_counter_deltas": { "deltas": null, "observed_endpoint_seconds": 0.0, "observed_intervals": 0, "requested_intervals": 1 } }, "cleanup": { "gpu_energy": { "available": false, "joules": null, "reason": "unbracketed_boundary" }, "gpu_sample_count": 1, "host_sample_count": 1, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86905700352, "mean": 86905700352, "median": 86905700352, "minimum": 86905700352, "sample_count": 1 }, "gpu_board_watts": { "maximum": 79.73, "mean": 79.73, "median": 79.73, "minimum": 79.73, "sample_count": 1 }, "gpu_memory_used_bytes": { "maximum": 94151639040.0, "mean": 94151639040.0, "median": 94151639040.0, "minimum": 94151639040.0, "sample_count": 1 }, "gpu_sm_clock_mhz": { "maximum": 2272.0, "mean": 2272.0, "median": 2272.0, "minimum": 2272.0, "sample_count": 1 }, "gpu_utilization_percent": { "maximum": 0.0, "mean": 0.0, "median": 0.0, "minimum": 0.0, "sample_count": 1 }, "host_mem_available_bytes": { "maximum": 95525027840, "mean": 95525027840, "median": 95525027840, "minimum": 95525027840, "sample_count": 1 }, "process_pss_bytes": null, "process_rss_bytes": null }, "process_sample_count": 0, "response_union_seconds": 1.1650557749671862, "sampled_counter_deltas": { "deltas": null, "observed_endpoint_seconds": 0.0, "observed_intervals": 0, "requested_intervals": 1 } }, "load": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 5680.773122758429, "reason": null, "response_union_seconds": 116.2166981890332, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 48.8808683371672, "trapezoid_pieces": 117, "watt_hours": 1.5779925340995635 }, "gpu_sample_count": 116, "host_sample_count": 89, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86973087744, "mean": 86957874831.82022, "median": 86961295360, "minimum": 86779006976, "sample_count": 89 }, "gpu_board_watts": { "maximum": 77.03, "mean": 49.04491379310345, "median": 71.38, "minimum": 15.7, "sample_count": 116 }, "gpu_memory_used_bytes": { "maximum": 94088724480.0, "mean": 39070574521.37931, "median": 587202560.0, "minimum": 587202560.0, "sample_count": 116 }, "gpu_sm_clock_mhz": { "maximum": 2280.0, "mean": 1387.0086206896551, "median": 2272.0, "minimum": 172.0, "sample_count": 116 }, "gpu_utilization_percent": { "maximum": 18.0, "mean": 4.905172413793103, "median": 0.0, "minimum": 0.0, "sample_count": 116 }, "host_mem_available_bytes": { "maximum": 96184197120, "mean": 95922319406.02248, "median": 95875031040, "minimum": 95732076544, "sample_count": 89 }, "process_pss_bytes": { "maximum": 86750541824, "mean": 66394780545.4382, "median": 85652535296, "minimum": 339699712, "sample_count": 89 }, "process_rss_bytes": { "maximum": 86753521664, "mean": 66397802449.97753, "median": 85655515136, "minimum": 342679552, "sample_count": 89 } }, "process_sample_count": 89, "response_union_seconds": 116.2166981890332, "sampled_counter_deltas": { "deltas": { "major_faults": 56045, "read_syscall_bytes": 13, "storage_read_bytes": 265319071744, "storage_write_bytes": 12288 }, "observed_endpoint_seconds": 115.25819665094605, "observed_intervals": 1, "requested_intervals": 1 } }, "ready": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 279.6970268935408, "reason": null, "response_union_seconds": 3.371804744063411, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 82.9517270790935, "trapezoid_pieces": 6, "watt_hours": 0.07769361858153911 }, "gpu_sample_count": 3, "host_sample_count": 2, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86954340352, "mean": 86929561600, "median": 86929561600.0, "minimum": 86904782848, "sample_count": 2 }, "gpu_board_watts": { "maximum": 88.89, "mean": 83.37, "median": 84.96, "minimum": 76.26, "sample_count": 3 }, "gpu_memory_used_bytes": { "maximum": 94157930496.0, "mean": 94135560874.66667, "median": 94155833344.0, "minimum": 94092918784.0, "sample_count": 3 }, "gpu_sm_clock_mhz": { "maximum": 2272.0, "mean": 2272.0, "median": 2272.0, "minimum": 2272.0, "sample_count": 3 }, "gpu_utilization_percent": { "maximum": 0.0, "mean": 0.0, "median": 0.0, "minimum": 0.0, "sample_count": 3 }, "host_mem_available_bytes": { "maximum": 95830491136, "mean": 95675967488, "median": 95675967488.0, "minimum": 95521443840, "sample_count": 2 }, "process_pss_bytes": { "maximum": 55339509760, "mean": 32317054976, "median": 32317054976.0, "minimum": 9294600192, "sample_count": 2 }, "process_rss_bytes": { "maximum": 55342489600, "mean": 32320034816, "median": 32320034816.0, "minimum": 9297580032, "sample_count": 2 } }, "process_sample_count": 2, "response_union_seconds": 3.371804744063411, "sampled_counter_deltas": { "deltas": null, "observed_endpoint_seconds": 0.0, "observed_intervals": 0, "requested_intervals": 3 } }, "responses": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 87439.59809728934, "reason": null, "response_union_seconds": 882.9507616019691, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 99.03111464408792, "trapezoid_pieces": 885, "watt_hours": 24.28877724924704 }, "gpu_sample_count": 883, "host_sample_count": 436, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86956191744, "mean": 86912462077.65138, "median": 86905081856.0, "minimum": 86897274880, "sample_count": 436 }, "gpu_board_watts": { "maximum": 137.35, "mean": 99.02436013590034, "median": 95.14, "minimum": 88.97, "sample_count": 883 }, "gpu_memory_used_bytes": { "maximum": 94157930496.0, "mean": 94157089735.17554, "median": 94157930496.0, "minimum": 94155833344.0, "sample_count": 883 }, "gpu_sm_clock_mhz": { "maximum": 2317.0, "mean": 2284.4246885617213, "median": 2280.0, "minimum": 1867.0, "sample_count": 883 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 34.90939977349943, "median": 10.0, "minimum": 2.0, "sample_count": 883 }, "host_mem_available_bytes": { "maximum": 95637037056, "mean": 95531524809.98166, "median": 95521929216.0, "minimum": 95466434560, "sample_count": 436 }, "process_pss_bytes": { "maximum": 55339509760, "mean": 55161964248.073395, "median": 55336781824.0, "minimum": 54836813824, "sample_count": 436 }, "process_rss_bytes": { "maximum": 55342489600, "mean": 55165004724.84404, "median": 55339761664.0, "minimum": 54839808000, "sample_count": 436 } }, "process_sample_count": 436, "response_union_seconds": 882.9507616019691, "sampled_counter_deltas": { "deltas": { "major_faults": 0, "read_syscall_bytes": 0, "storage_read_bytes": 0, "storage_write_bytes": 385024 }, "observed_endpoint_seconds": 878.285594719986, "observed_intervals": 2, "requested_intervals": 2 } }, "warmup": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 18356.05087851068, "reason": null, "response_union_seconds": 192.7558839409612, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 95.22952297597782, "trapezoid_pieces": 195, "watt_hours": 5.098903021808522 }, "gpu_sample_count": 193, "host_sample_count": 104, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86972977152, "mean": 86935542350.76923, "median": 86941855744.0, "minimum": 86844063744, "sample_count": 104 }, "gpu_board_watts": { "maximum": 133.51, "mean": 95.20875647668393, "median": 93.83, "minimum": 74.26, "sample_count": 193 }, "gpu_memory_used_bytes": { "maximum": 94157930496.0, "mean": 94156430977.98964, "median": 94155833344.0, "minimum": 94155833344.0, "sample_count": 193 }, "gpu_sm_clock_mhz": { "maximum": 2317.0, "mean": 2280.352331606218, "median": 2280.0, "minimum": 1942.0, "sample_count": 193 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 31.139896373056995, "median": 10.0, "minimum": 2.0, "sample_count": 193 }, "host_mem_available_bytes": { "maximum": 95645704192, "mean": 95576961142.15384, "median": 95548684288.0, "minimum": 95511699456, "sample_count": 104 }, "process_pss_bytes": { "maximum": 55336093696, "mean": 50895688615.38461, "median": 54901958656.0, "minimum": 11203971072, "sample_count": 104 }, "process_rss_bytes": { "maximum": 55339073536, "mean": 50898682210.46154, "median": 54904938496.0, "minimum": 11206950912, "sample_count": 104 } }, "process_sample_count": 104, "response_union_seconds": 192.7558839409612, "sampled_counter_deltas": { "deltas": { "major_faults": 5517, "read_syscall_bytes": 0, "storage_read_bytes": 44982132736, "storage_write_bytes": 73728 }, "observed_endpoint_seconds": 190.65362677600933, "observed_intervals": 2, "requested_intervals": 2 } } }, "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c", "settings": { "batch": 2048, "cache_prompt": false, "cache_ram_bytes": 0, "concurrency": 1, "context_shift": false, "contexts": [ 2048, 8192 ], "endpoint": "/completion", "fit": "off", "flash_attention": true, "ignore_eos": true, "kv_cache": "f16", "lazy_mode": "off", "load_mode": "mmap", "measured_rounds": 5, "mlock": false, "network_scope": "loopback", "output_tokens": 256, "p95_minimum_requests": 200, "placement": { "cpu_expert_layers": 44, "gpu_layers": "all" }, "reasoning": "not_applicable_raw_completion", "seed": 20260927, "server_slots": 1, "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187", "temperature": 0, "threads": 16, "tokenization": "model_tokenizer_add_special_false_repeat_then_truncate", "total_context_tokens": 8704, "ubatch": 512, "warmup_rounds": 1 }, "status": "completed", "telemetry": { "gpu_samples": 1198, "host_samples": 634, "maximum_gpu_gap_seconds": 1.4088406169903465, "maximum_host_gap_seconds": 2.6100394439999945, "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.", "thermal_slowdown_fully_available": true, "thermal_slowdown_observed_samples": 0, "thermal_slowdown_states": { "hardware": { "active": 0, "not_active": 1198, "unavailable": 0 }, "software": { "active": 0, "not_active": 1198, "unavailable": 0 } } } }, { "case_wall_seconds": 519.5668174049933, "completed_response_resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 19300.75655985953, "reason": null, "response_union_seconds": 214.645721802779, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 89.91912998663662, "trapezoid_pieces": 239, "watt_hours": 5.361321266627646 }, "gpu_sample_count": 215, "host_sample_count": 104, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86972829696, "mean": 86943777083.07692, "median": 86953607168.0, "minimum": 86751805440, "sample_count": 104 }, "gpu_board_watts": { "maximum": 118.27, "mean": 89.92758139534884, "median": 87.15, "minimum": 80.39, "sample_count": 215 }, "gpu_memory_used_bytes": { "maximum": 94176804864.0, "mean": 94171683911.44186, "median": 94172610560.0, "minimum": 94145347584.0, "sample_count": 215 }, "gpu_sm_clock_mhz": { "maximum": 2325.0, "mean": 2310.8418604651165, "median": 2317.0, "minimum": 2130.0, "sample_count": 215 }, "gpu_utilization_percent": { "maximum": 96.0, "mean": 58.62325581395349, "median": 78.0, "minimum": 0.0, "sample_count": 215 }, "host_mem_available_bytes": { "maximum": 95620997120, "mean": 95580679483.07692, "median": 95579918336.0, "minimum": 95542239232, "sample_count": 104 }, "process_pss_bytes": { "maximum": 73515068416, "mean": 68495140657.23077, "median": 71700847616.0, "minimum": 43000757248, "sample_count": 104 }, "process_rss_bytes": { "maximum": 73518055424, "mean": 68498155047.38461, "median": 71703834624.0, "minimum": 43003744256, "sample_count": 104 } }, "process_sample_count": 104, "response_union_seconds": 214.645721802779, "sampled_counter_deltas": { "deltas": { "major_faults": 10941, "read_syscall_bytes": 0, "storage_read_bytes": 26791116800, "storage_write_bytes": 32768 }, "observed_endpoint_seconds": 166.1631237110705, "observed_intervals": 24, "requested_intervals": 24 } }, "groups": [ { "completed_output_tokens": 262, "completed_request_count": 24, "context_tokens": null, "failed_request_count": 0, "meaning": "Distinct authored tasks, one attempt each; no pooled practical latency ranking or general quality claim.", "measured_rounds": 24, "mode": "chat", "resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 19300.75655985953, "reason": null, "response_union_seconds": 214.645721802779, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 89.91912998663662, "trapezoid_pieces": 239, "watt_hours": 5.361321266627646 }, "gpu_sample_count": 215, "host_sample_count": 104, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86972829696, "mean": 86943777083.07692, "median": 86953607168.0, "minimum": 86751805440, "sample_count": 104 }, "gpu_board_watts": { "maximum": 118.27, "mean": 89.92758139534884, "median": 87.15, "minimum": 80.39, "sample_count": 215 }, "gpu_memory_used_bytes": { "maximum": 94176804864.0, "mean": 94171683911.44186, "median": 94172610560.0, "minimum": 94145347584.0, "sample_count": 215 }, "gpu_sm_clock_mhz": { "maximum": 2325.0, "mean": 2310.8418604651165, "median": 2317.0, "minimum": 2130.0, "sample_count": 215 }, "gpu_utilization_percent": { "maximum": 96.0, "mean": 58.62325581395349, "median": 78.0, "minimum": 0.0, "sample_count": 215 }, "host_mem_available_bytes": { "maximum": 95620997120, "mean": 95580679483.07692, "median": 95579918336.0, "minimum": 95542239232, "sample_count": 104 }, "process_pss_bytes": { "maximum": 73515068416, "mean": 68495140657.23077, "median": 71700847616.0, "minimum": 43000757248, "sample_count": 104 }, "process_rss_bytes": { "maximum": 73518055424, "mean": 68498155047.38461, "median": 71703834624.0, "minimum": 43003744256, "sample_count": 104 } }, "process_sample_count": 104, "response_union_seconds": 214.645721802779, "sampled_counter_deltas": { "deltas": { "major_faults": 10941, "read_syscall_bytes": 0, "storage_read_bytes": 26791116800, "storage_write_bytes": 32768 }, "observed_endpoint_seconds": 166.1631237110705, "observed_intervals": 24, "requested_intervals": 24 } }, "response_outer_span_seconds": 214.70433678099653, "response_window_union_seconds": 214.645721802779, "score": { "by_family": { "arithmetic": { "expected_value_pass": 4, "strict_format_pass": 5, "task_count": 6 }, "classification": { "expected_value_pass": 6, "strict_format_pass": 6, "task_count": 6 }, "extraction": { "expected_value_pass": 5, "strict_format_pass": 6, "task_count": 6 }, "grounded-qa": { "expected_value_pass": 6, "strict_format_pass": 6, "task_count": 6 } }, "by_language": { "en": { "expected_value_pass": 9, "strict_format_pass": 11, "task_count": 12 }, "pl": { "expected_value_pass": 12, "strict_format_pass": 12, "task_count": 12 } }, "completed_output": 24, "completed_transport": 24, "expected_value_pass": 21, "strict_format_pass": 23, "task_count": 24, "truncated": 0 }, "tasks": [ { "completed_output": true, "connection_seconds": 13.583070762979332, "failure_reason": null, "family": "arithmetic", "finish_reason": "stop", "first_generated_event_seconds": 11.753905638004653, "first_visible_content_seconds": 11.753905638004653, "generated_tokens": 8, "input_tokens": 69, "language": "pl", "last_visible_content_seconds": 13.320375241979491, "reasoning_events_observed": false, "response_seconds": 13.58291348197963, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "arithmetic-pl-1", "truncated": false }, { "completed_output": true, "connection_seconds": 11.222968201967888, "failure_reason": null, "family": "arithmetic", "finish_reason": "stop", "first_generated_event_seconds": 8.15195504296571, "first_visible_content_seconds": 8.15195504296571, "generated_tokens": 14, "input_tokens": 69, "language": "pl", "last_visible_content_seconds": 10.913710985973012, "reasoning_events_observed": false, "response_seconds": 11.222829844977241, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "arithmetic-pl-2", "truncated": false }, { "completed_output": true, "connection_seconds": 9.111604891950265, "failure_reason": null, "family": "arithmetic", "finish_reason": "stop", "first_generated_event_seconds": 7.609585767961107, "first_visible_content_seconds": 7.609585767961107, "generated_tokens": 8, "input_tokens": 93, "language": "pl", "last_visible_content_seconds": 8.899494110955857, "reasoning_events_observed": false, "response_seconds": 9.111451656965073, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "arithmetic-pl-3", "truncated": false }, { "completed_output": true, "connection_seconds": 7.83672762895003, "failure_reason": null, "family": "arithmetic", "finish_reason": "stop", "first_generated_event_seconds": 6.289020258001983, "first_visible_content_seconds": 6.289020258001983, "generated_tokens": 8, "input_tokens": 49, "language": "en", "last_visible_content_seconds": 7.617563018982764, "reasoning_events_observed": false, "response_seconds": 7.836561900970992, "score": { "completed": true, "expected_value_pass": false, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "arithmetic-en-1", "truncated": false }, { "completed_output": true, "connection_seconds": 8.593278787040617, "failure_reason": null, "family": "arithmetic", "finish_reason": "stop", "first_generated_event_seconds": 6.61590343300486, "first_visible_content_seconds": 6.61590343300486, "generated_tokens": 10, "input_tokens": 54, "language": "en", "last_visible_content_seconds": 8.366578131041024, "reasoning_events_observed": false, "response_seconds": 8.59308045799844, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "arithmetic-en-2", "truncated": false }, { "completed_output": true, "connection_seconds": 8.821014977991581, "failure_reason": null, "family": "arithmetic", "finish_reason": "stop", "first_generated_event_seconds": 6.420646424987353, "first_visible_content_seconds": 6.420646424987353, "generated_tokens": 12, "input_tokens": 56, "language": "en", "last_visible_content_seconds": 8.583703558018897, "reasoning_events_observed": false, "response_seconds": 8.820840994012542, "score": { "completed": true, "expected_value_pass": false, "strict_format_pass": false }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "arithmetic-en-3", "truncated": false }, { "completed_output": true, "connection_seconds": 11.624865641002543, "failure_reason": null, "family": "extraction", "finish_reason": "stop", "first_generated_event_seconds": 7.236178473976906, "first_visible_content_seconds": 7.236178473976906, "generated_tokens": 21, "input_tokens": 84, "language": "pl", "last_visible_content_seconds": 11.386044683982618, "reasoning_events_observed": false, "response_seconds": 11.624679514963645, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "extraction-pl-1", "truncated": false }, { "completed_output": true, "connection_seconds": 11.437166781979613, "failure_reason": null, "family": "extraction", "finish_reason": "stop", "first_generated_event_seconds": 6.874179092003033, "first_visible_content_seconds": 6.874179092003033, "generated_tokens": 22, "input_tokens": 81, "language": "pl", "last_visible_content_seconds": 11.210961457982194, "reasoning_events_observed": false, "response_seconds": 11.43702499900246, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "extraction-pl-2", "truncated": false }, { "completed_output": true, "connection_seconds": 9.50213718204759, "failure_reason": null, "family": "extraction", "finish_reason": "stop", "first_generated_event_seconds": 6.080808499013074, "first_visible_content_seconds": 6.080808499013074, "generated_tokens": 17, "input_tokens": 72, "language": "pl", "last_visible_content_seconds": 9.297318112046923, "reasoning_events_observed": false, "response_seconds": 9.50197277701227, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "extraction-pl-3", "truncated": false }, { "completed_output": true, "connection_seconds": 10.387426305038389, "failure_reason": null, "family": "extraction", "finish_reason": "stop", "first_generated_event_seconds": 6.486654097039718, "first_visible_content_seconds": 6.486654097039718, "generated_tokens": 19, "input_tokens": 61, "language": "en", "last_visible_content_seconds": 10.170321426005103, "reasoning_events_observed": false, "response_seconds": 10.387277828995138, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "extraction-en-1", "truncated": false }, { "completed_output": true, "connection_seconds": 9.730554731970187, "failure_reason": null, "family": "extraction", "finish_reason": "stop", "first_generated_event_seconds": 6.0840944799711, "first_visible_content_seconds": 6.0840944799711, "generated_tokens": 18, "input_tokens": 51, "language": "en", "last_visible_content_seconds": 9.521599187981337, "reasoning_events_observed": false, "response_seconds": 9.730390105978586, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "extraction-en-2", "truncated": false }, { "completed_output": true, "connection_seconds": 9.24161141901277, "failure_reason": null, "family": "extraction", "finish_reason": "stop", "first_generated_event_seconds": 5.852490445016883, "first_visible_content_seconds": 5.852490445016883, "generated_tokens": 17, "input_tokens": 53, "language": "en", "last_visible_content_seconds": 9.026987073011696, "reasoning_events_observed": false, "response_seconds": 9.241449408000335, "score": { "completed": true, "expected_value_pass": false, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "extraction-en-3", "truncated": false }, { "completed_output": true, "connection_seconds": 7.9818881780374795, "failure_reason": null, "family": "classification", "finish_reason": "stop", "first_generated_event_seconds": 6.718955708027352, "first_visible_content_seconds": 6.718955708027352, "generated_tokens": 7, "input_tokens": 87, "language": "pl", "last_visible_content_seconds": 7.771279089036398, "reasoning_events_observed": false, "response_seconds": 7.981710618012585, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "classification-pl-1", "truncated": false }, { "completed_output": true, "connection_seconds": 8.087027255038265, "failure_reason": null, "family": "classification", "finish_reason": "stop", "first_generated_event_seconds": 6.827422423986718, "first_visible_content_seconds": 6.827422423986718, "generated_tokens": 7, "input_tokens": 87, "language": "pl", "last_visible_content_seconds": 7.865569265035447, "reasoning_events_observed": false, "response_seconds": 8.08685941301519, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "classification-pl-2", "truncated": false }, { "completed_output": true, "connection_seconds": 7.949002139037475, "failure_reason": null, "family": "classification", "finish_reason": "stop", "first_generated_event_seconds": 6.248805720009841, "first_visible_content_seconds": 6.248805720009841, "generated_tokens": 9, "input_tokens": 76, "language": "pl", "last_visible_content_seconds": 7.737396181037184, "reasoning_events_observed": false, "response_seconds": 7.9487804470118135, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "classification-pl-3", "truncated": false }, { "completed_output": true, "connection_seconds": 7.826057630998548, "failure_reason": null, "family": "classification", "finish_reason": "stop", "first_generated_event_seconds": 6.5673799850046635, "first_visible_content_seconds": 6.5673799850046635, "generated_tokens": 7, "input_tokens": 65, "language": "en", "last_visible_content_seconds": 7.621764203009661, "reasoning_events_observed": false, "response_seconds": 7.825891762971878, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "classification-en-1", "truncated": false }, { "completed_output": true, "connection_seconds": 7.492189083015546, "failure_reason": null, "family": "classification", "finish_reason": "stop", "first_generated_event_seconds": 6.23327365500154, "first_visible_content_seconds": 6.23327365500154, "generated_tokens": 7, "input_tokens": 64, "language": "en", "last_visible_content_seconds": 7.279204571968876, "reasoning_events_observed": false, "response_seconds": 7.492010992020369, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "classification-en-2", "truncated": false }, { "completed_output": true, "connection_seconds": 6.762584539013915, "failure_reason": null, "family": "classification", "finish_reason": "stop", "first_generated_event_seconds": 5.67244211898651, "first_visible_content_seconds": 5.67244211898651, "generated_tokens": 6, "input_tokens": 48, "language": "en", "last_visible_content_seconds": 6.542209693987388, "reasoning_events_observed": false, "response_seconds": 6.762443727988284, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "classification-en-3", "truncated": false }, { "completed_output": true, "connection_seconds": 7.668065410980489, "failure_reason": null, "family": "grounded-qa", "finish_reason": "stop", "first_generated_event_seconds": 6.403421438997611, "first_visible_content_seconds": 6.403421438997611, "generated_tokens": 7, "input_tokens": 78, "language": "pl", "last_visible_content_seconds": 7.461564408033155, "reasoning_events_observed": false, "response_seconds": 7.667886228009593, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "grounded-pl-1", "truncated": false }, { "completed_output": true, "connection_seconds": 7.881595277984161, "failure_reason": null, "family": "grounded-qa", "finish_reason": "stop", "first_generated_event_seconds": 6.181176976999268, "first_visible_content_seconds": 6.181176976999268, "generated_tokens": 9, "input_tokens": 89, "language": "pl", "last_visible_content_seconds": 7.651056489965413, "reasoning_events_observed": false, "response_seconds": 7.881407508975826, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "grounded-pl-2", "truncated": false }, { "completed_output": true, "connection_seconds": 8.305406797968317, "failure_reason": null, "family": "grounded-qa", "finish_reason": "stop", "first_generated_event_seconds": 7.048749969981145, "first_visible_content_seconds": 7.048749969981145, "generated_tokens": 7, "input_tokens": 108, "language": "pl", "last_visible_content_seconds": 8.094621035968885, "reasoning_events_observed": false, "response_seconds": 8.305239226960111, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "grounded-pl-3", "truncated": false }, { "completed_output": true, "connection_seconds": 7.758344300033059, "failure_reason": null, "family": "grounded-qa", "finish_reason": "stop", "first_generated_event_seconds": 6.494405768986326, "first_visible_content_seconds": 6.494405768986326, "generated_tokens": 7, "input_tokens": 52, "language": "en", "last_visible_content_seconds": 7.553320523991715, "reasoning_events_observed": false, "response_seconds": 7.7581756060244516, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "grounded-en-1", "truncated": false }, { "completed_output": true, "connection_seconds": 7.272968183970079, "failure_reason": null, "family": "grounded-qa", "finish_reason": "stop", "first_generated_event_seconds": 6.025387093948666, "first_visible_content_seconds": 6.025387093948666, "generated_tokens": 7, "input_tokens": 61, "language": "en", "last_visible_content_seconds": 7.066188130993396, "reasoning_events_observed": false, "response_seconds": 7.272804338950664, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "grounded-en-2", "truncated": false }, { "completed_output": true, "connection_seconds": 8.572214197949506, "failure_reason": null, "family": "grounded-qa", "finish_reason": "stop", "first_generated_event_seconds": 7.092147689952981, "first_visible_content_seconds": 7.092147689952981, "generated_tokens": 8, "input_tokens": 83, "language": "en", "last_visible_content_seconds": 8.351030136982445, "reasoning_events_observed": false, "response_seconds": 8.572038961981889, "score": { "completed": true, "expected_value_pass": true, "strict_format_pass": true }, "server_decode_tokens_per_second": null, "status": "completed", "task_id": "grounded-en-3", "truncated": false } ], "warmup_completed_rounds": 1, "warmup_requests_per_round": 1 } ], "id": "practical-235b", "model": "qwen3-235b-q5", "module": "practical", "placement": { "buffers": [ { "device": "CPU_Mapped", "reported_mib": 27759.73 }, { "device": "CPU_Mapped", "reported_mib": 28323.86 }, { "device": "CPU_Mapped", "reported_mib": 18322.9 }, { "device": "CUDA0", "reported_mib": 86732.62 } ], "gpu_layers": 95, "meaning": "Runtime buffer allocation is not measured resident process RAM.", "requested_cpu_expert_layers": 44, "total_layers": 95 }, "record_sha256": "e4b245570fb95c8646e8413e1491daee97364c0f61d1eb5d5f55d65ad0f7354c", "resources_by_phase": { "baseline": { "gpu_energy": { "available": false, "joules": null, "reason": "unbracketed_boundary" }, "gpu_sample_count": 2, "host_sample_count": 2, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86158008320, "mean": 86149492736, "median": 86149492736.0, "minimum": 86140977152, "sample_count": 2 }, "gpu_board_watts": { "maximum": 16.24, "mean": 16.18, "median": 16.18, "minimum": 16.12, "sample_count": 2 }, "gpu_memory_used_bytes": { "maximum": 2097152.0, "mean": 2097152.0, "median": 2097152.0, "minimum": 2097152.0, "sample_count": 2 }, "gpu_sm_clock_mhz": { "maximum": 180.0, "mean": 180.0, "median": 180.0, "minimum": 180.0, "sample_count": 2 }, "gpu_utilization_percent": { "maximum": 0.0, "mean": 0.0, "median": 0.0, "minimum": 0.0, "sample_count": 2 }, "host_mem_available_bytes": { "maximum": 96376918016, "mean": 96365586432, "median": 96365586432.0, "minimum": 96354254848, "sample_count": 2 }, "process_pss_bytes": null, "process_rss_bytes": null }, "process_sample_count": 0, "response_union_seconds": 1.1007638040464371, "sampled_counter_deltas": { "deltas": null, "observed_endpoint_seconds": 0.0, "observed_intervals": 0, "requested_intervals": 1 } }, "cleanup": { "gpu_energy": { "available": false, "joules": null, "reason": "unbracketed_boundary" }, "gpu_sample_count": 1, "host_sample_count": 1, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86669819904, "mean": 86669819904, "median": 86669819904, "minimum": 86669819904, "sample_count": 1 }, "gpu_board_watts": { "maximum": 78.13, "mean": 78.13, "median": 78.13, "minimum": 78.13, "sample_count": 1 }, "gpu_memory_used_bytes": { "maximum": 94176804864.0, "mean": 94176804864.0, "median": 94176804864.0, "minimum": 94176804864.0, "sample_count": 1 }, "gpu_sm_clock_mhz": { "maximum": 2272.0, "mean": 2272.0, "median": 2272.0, "minimum": 2272.0, "sample_count": 1 }, "gpu_utilization_percent": { "maximum": 0.0, "mean": 0.0, "median": 0.0, "minimum": 0.0, "sample_count": 1 }, "host_mem_available_bytes": { "maximum": 95835246592, "mean": 95835246592, "median": 95835246592, "minimum": 95835246592, "sample_count": 1 }, "process_pss_bytes": null, "process_rss_bytes": null }, "process_sample_count": 0, "response_union_seconds": 1.1610513600171544, "sampled_counter_deltas": { "deltas": null, "observed_endpoint_seconds": 0.0, "observed_intervals": 0, "requested_intervals": 1 } }, "load": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 4127.699400766261, "reason": null, "response_union_seconds": 79.19833586795721, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 52.11851177843098, "trapezoid_pieces": 80, "watt_hours": 1.1465831668795168 }, "gpu_sample_count": 79, "host_sample_count": 58, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86973087744, "mean": 86923087448.27586, "median": 86962081792.0, "minimum": 85281234944, "sample_count": 58 }, "gpu_board_watts": { "maximum": 77.81, "mean": 52.37987341772152, "median": 71.54, "minimum": 15.46, "sample_count": 79 }, "gpu_memory_used_bytes": { "maximum": 94088724480.0, "mean": 38674748066.025314, "median": 587202560.0, "minimum": 587202560.0, "sample_count": 79 }, "gpu_sm_clock_mhz": { "maximum": 2280.0, "mean": 1507.860759493671, "median": 2272.0, "minimum": 180.0, "sample_count": 79 }, "gpu_utilization_percent": { "maximum": 34.0, "mean": 7.0, "median": 0.0, "minimum": 0.0, "sample_count": 79 }, "host_mem_available_bytes": { "maximum": 96077516800, "mean": 95845226142.89655, "median": 95821776896.0, "minimum": 95730827264, "sample_count": 58 }, "process_pss_bytes": { "maximum": 86550830080, "mean": 72792922747.58621, "median": 85804739584.0, "minimum": 1283374080, "sample_count": 58 }, "process_rss_bytes": { "maximum": 86553817088, "mean": 72795910002.75862, "median": 85807726592.0, "minimum": 1286361088, "sample_count": 58 } }, "process_sample_count": 58, "response_union_seconds": 79.19833586795721, "sampled_counter_deltas": { "deltas": { "major_faults": 57576, "read_syscall_bytes": 13, "storage_read_bytes": 144988631040, "storage_write_bytes": 12288 }, "observed_endpoint_seconds": 77.89480808604276, "observed_intervals": 1, "requested_intervals": 1 } }, "ready": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 190.05088581740472, "reason": null, "response_union_seconds": 2.2796509800245985, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 83.36841362240196, "trapezoid_pieces": 4, "watt_hours": 0.05279191272705687 }, "gpu_sample_count": 2, "host_sample_count": 2, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86942056448, "mean": 86935918592, "median": 86935918592.0, "minimum": 86929780736, "sample_count": 2 }, "gpu_board_watts": { "maximum": 87.87, "mean": 82.33000000000001, "median": 82.33000000000001, "minimum": 76.79, "sample_count": 2 }, "gpu_memory_used_bytes": { "maximum": 94176804864.0, "mean": 94134861824.0, "median": 94134861824.0, "minimum": 94092918784.0, "sample_count": 2 }, "gpu_sm_clock_mhz": { "maximum": 2295.0, "mean": 2283.5, "median": 2283.5, "minimum": 2272.0, "sample_count": 2 }, "gpu_utilization_percent": { "maximum": 0.0, "mean": 0.0, "median": 0.0, "minimum": 0.0, "sample_count": 2 }, "host_mem_available_bytes": { "maximum": 95807406080, "mean": 95690938368, "median": 95690938368.0, "minimum": 95574470656, "sample_count": 2 }, "process_pss_bytes": { "maximum": 73541479424, "mean": 42100476928, "median": 42100476928.0, "minimum": 10659474432, "sample_count": 2 }, "process_rss_bytes": { "maximum": 73544466432, "mean": 42103463936, "median": 42103463936.0, "minimum": 10662461440, "sample_count": 2 } }, "process_sample_count": 2, "response_union_seconds": 2.2796509800245985, "sampled_counter_deltas": { "deltas": null, "observed_endpoint_seconds": 0.0, "observed_intervals": 0, "requested_intervals": 2 } }, "responses": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 19306.502753033474, "reason": null, "response_union_seconds": 214.70691833097953, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 89.92026387930224, "trapezoid_pieces": 216, "watt_hours": 5.362917431398187 }, "gpu_sample_count": 215, "host_sample_count": 104, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86972829696, "mean": 86943777083.07692, "median": 86953607168.0, "minimum": 86751805440, "sample_count": 104 }, "gpu_board_watts": { "maximum": 118.27, "mean": 89.92758139534884, "median": 87.15, "minimum": 80.39, "sample_count": 215 }, "gpu_memory_used_bytes": { "maximum": 94176804864.0, "mean": 94171683911.44186, "median": 94172610560.0, "minimum": 94145347584.0, "sample_count": 215 }, "gpu_sm_clock_mhz": { "maximum": 2325.0, "mean": 2310.8418604651165, "median": 2317.0, "minimum": 2130.0, "sample_count": 215 }, "gpu_utilization_percent": { "maximum": 96.0, "mean": 58.62325581395349, "median": 78.0, "minimum": 0.0, "sample_count": 215 }, "host_mem_available_bytes": { "maximum": 95620997120, "mean": 95580679483.07692, "median": 95579918336.0, "minimum": 95542239232, "sample_count": 104 }, "process_pss_bytes": { "maximum": 73515068416, "mean": 68495140657.23077, "median": 71700847616.0, "minimum": 43000757248, "sample_count": 104 }, "process_rss_bytes": { "maximum": 73518055424, "mean": 68498155047.38461, "median": 71703834624.0, "minimum": 43003744256, "sample_count": 104 } }, "process_sample_count": 104, "response_union_seconds": 214.70691833097953, "sampled_counter_deltas": { "deltas": { "major_faults": 23997, "read_syscall_bytes": 0, "storage_read_bytes": 29071040512, "storage_write_bytes": 147456 }, "observed_endpoint_seconds": 212.04975535703124, "observed_intervals": 1, "requested_intervals": 1 } }, "warmup": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 995.0526985724146, "reason": null, "response_union_seconds": 13.300178417004645, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 74.81498874482858, "trapezoid_pieces": 14, "watt_hours": 0.27640352738122625 }, "gpu_sample_count": 13, "host_sample_count": 11, "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.", "metrics": { "cgroup_memory_current_bytes": { "maximum": 86970953728, "mean": 86944415744, "median": 86959644672, "minimum": 86869540864, "sample_count": 11 }, "gpu_board_watts": { "maximum": 90.66, "mean": 74.63384615384615, "median": 73.3, "minimum": 73.06, "sample_count": 13 }, "gpu_memory_used_bytes": { "maximum": 94141153280.0, "mean": 94141153280.0, "median": 94141153280.0, "minimum": 94141153280.0, "sample_count": 13 }, "gpu_sm_clock_mhz": { "maximum": 2272.0, "mean": 2272.0, "median": 2272.0, "minimum": 2272.0, "sample_count": 13 }, "gpu_utilization_percent": { "maximum": 1.0, "mean": 0.07692307692307693, "median": 0.0, "minimum": 0.0, "sample_count": 13 }, "host_mem_available_bytes": { "maximum": 95628480512, "mean": 95611195019.63637, "median": 95612923904, "minimum": 95555829760, "sample_count": 11 }, "process_pss_bytes": { "maximum": 41014807552, "mean": 26555915264, "median": 26415000576, "minimum": 12266339328, "sample_count": 11 }, "process_rss_bytes": { "maximum": 41017794560, "mean": 26558902272, "median": 26417987584, "minimum": 12269326336, "sample_count": 11 } }, "process_sample_count": 11, "response_union_seconds": 13.300178417004645, "sampled_counter_deltas": { "deltas": { "major_faults": 224642, "read_syscall_bytes": 0, "storage_read_bytes": 31496257536, "storage_write_bytes": 0 }, "observed_endpoint_seconds": 12.271783648000564, "observed_intervals": 1, "requested_intervals": 1 } } }, "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c", "settings": { "batch": 2048, "cache_prompt": false, "cache_ram_bytes": 0, "concurrency": 1, "context_shift": false, "contexts": [], "endpoint": "/v1/chat/completions", "fit": "off", "flash_attention": true, "ignore_eos": false, "kv_cache": "f16", "lazy_mode": "off", "load_mode": "mmap", "measured_rounds": 24, "mlock": false, "network_scope": "loopback", "output_tokens": 1024, "p95_minimum_requests": 200, "placement": { "cpu_expert_layers": 44, "gpu_layers": "all" }, "reasoning": "disabled", "seed": 20260927, "server_slots": 1, "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187", "temperature": 0, "threads": 16, "tokenization": "natural_chat_template_no_padding", "total_context_tokens": 8704, "ubatch": 512, "warmup": "one_disjoint_transport_request", "warmup_rounds": 1 }, "status": "completed", "telemetry": { "gpu_samples": 312, "host_samples": 178, "maximum_gpu_gap_seconds": 1.1819236329756677, "maximum_host_gap_seconds": 2.7705111099639907, "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.", "thermal_slowdown_fully_available": true, "thermal_slowdown_observed_samples": 0, "thermal_slowdown_states": { "hardware": { "active": 0, "not_active": 312, "unavailable": 0 }, "software": { "active": 0, "not_active": 312, "unavailable": 0 } } } } ], "comparisons": [], "environment": { "kernel": "6.12.0-211.56.1.el10_2.0.1.x86_64", "nvidia_driver": "595.91.07", "os": { "id": "rocky", "name": "Rocky Linux", "version_id": "10.2" } }, "hardware": { "cpu": "AMD EPYC 4565P 16-Core Processor", "fan_profile": "GPU300WAuto", "gpu": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition", "gpu_nominal_vram_gb": 96, "gpu_vram_bytes": 102641958912, "installed_ram_bytes": 103079215104, "kernel_mem_total_bytes": 100271353856, "logical_cpu_count": 32, "physical_cpu_count": 16 }, "historical_results_pooled": false, "limitations": [ "Externally pinned capture integrity is not remote hardware attestation.", "Five repeated rounds share a process and are not five independent installations.", "Synthetic token latency is not practical first-visible answer latency.", "GPU board power is not whole-server AC power; energy is unavailable without bracketing samples.", "Sampled memory peaks may not coincide and do not establish minimum required system RAM.", "The practical corpus contains 24 authored tasks, not a broad model-quality benchmark.", "Engine telemetry includes process loading and internal warmup; exact per-sample resource windows are unavailable.", "No historical result is relabeled or pooled into this versioned series." ], "models": [ { "architecture": "moe", "chat_template_sha256": "57f1fd00f0013a2be96aa79b857391f27e23df5b5f847072b524c897e24d0361", "effective_chat_template_rule": "llama.cpp-7fe450e19-chat-source-v1", "effective_chat_template_sha256": "57f1fd00f0013a2be96aa79b857391f27e23df5b5f847072b524c897e24d0361", "files": [ { "bytes": 29673975136, "sha256": "706382f26d0d6f72cd89c9ac1bfeb40d4b4afaee986dad2ea229ea24f9168fab" }, { "bytes": 29748770496, "sha256": "9ab54360c3809c580439b75fe7e16f3af24adcc701c352e39b06ad557aca0ca3" }, { "bytes": 29749049024, "sha256": "93c863b35e810b97605dc973c2881e608751a269e773b3772c42e5e25cb01d79" }, { "bytes": 29642094272, "sha256": "c0322fa3f66ca392a9d7f95bb2543b6f6213ac04c16e613a2ae3a82429452071" }, { "bytes": 29856282304, "sha256": "1dce96f79ba9c2d86d36bbd856101f7a2e4f3c4cff4ef4a2a970c0ba336edb15" }, { "bytes": 18143985760, "sha256": "15a2192eadc2cc2da6fc0df6d25cb6d39626f292716526cfee8baba9043fbe62" } ], "id": "qwen3-235b-q5", "name": "Qwen3-235B-A22B", "quantization": "Q5_K_M", "repository": "Qwen/Qwen3-235B-A22B-GGUF", "revision": "211e807ecd6d37446a747409108783e39a04e80f", "tokenizer_sha256": "2629971f294a1a43ef545d689a2090a820e0e1a15e5371e1c88d1c21159ef849" } ], "protocol": "gsh-llm-core/1.1", "runtime": { "archives_sha256": [ "c2ab9e19838513ff69d1af8d999ad717dd3c7ee4714ac04c7ed5ab9077c50e4e", "1466daea60aad1144819e151b2bae19d54556cf1da6c129c4f55a5ded2637c25" ], "binary_sha256": "376626f953efd7bbf16dee112747196d2b1de26717bcd1d37e203340dc5c043c", "commit": "7fe450e19305b828c199d602c23a8337aaa1f03b", "name": "llama.cpp", "release": "b11146" }, "schema": "gsh-llm-core-article/1", "sources": [ { "exporter_sha256": "4f4046b91d99ba35175c3c6c04473a0021f54c6baa4892625cb66435f4c7e2ec", "finished_at": "2026-09-27T16:13:37.029489+00:00", "plan_id": "core-capacity-20260927t151737-capacityb1f78281", "plan_sha256": "20b8022e204c32d4826a7ef00445ccd3e6fc8b0fe2bbccc222714e0fb4fbe2d3", "protocol_sha256": "3ce8ce9f2b17e66a655c07fd7e7aa3f1762790ad9be38f87d29e5e1279dec3ac", "record_sha256": "b80b2a18f3d821f8c7929b5d1d418771b14e88b3e73c0059056a1a8c31d33416", "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c", "runner_sha256": "1bac2436c5aa2eca4f4b42cc9b8f0942b5c66256eb1a41a2169b88c12b044f4e", "started_at": "2026-09-27T15:24:32.393663+00:00", "task_corpus_sha256": "0642db889565ce25b980e24afa5048eb83e4d15c36754457474a65b2c1f71c21" } ], "status": "completed", "test_date_timezone": "Europe/Warsaw", "tested_at": "2026-09-27" }