{ "model_tag": "nomic-embed-text:latest", "arm": "one-dynboost", "box_class": "gpu-5090-laptop-24g", "mode": "embed", "target_uuid": "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff", "base_url": "http://127.0.0.1:11470", "live_seat_arm": false, "prompt_sha256": "bfb6771b30e1740a52b012e047d697629dee6a1b023a048c17c3fb103a591248", "prompt_sha256_note": "the digest of the deterministically generated batch this arm embedded -- not the series' frozen prompt, which this arm does not use", "card_labels": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": "soldered (mobile; no socket \u2014 link width is traced under load, never assumed) at 00000000:01:00.0 - vendor unread - GPU-edff232c" }, "card_identity": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": { "index": null, "bus_id": "00000000:01:00.0", "pci_sub_device_id": null, "vendor": null, "board_per_operator": null, "vbios": null, "serial": null, "pcie_link_width_max": null, "pcie_link_width_negotiated": null, "memory_total_mib": null, "memory_bus_width_bits": null, "clocks_max_sm_mhz": null, "clocks_max_memory_mhz": null, "power_default_limit_w": null, "power_min_limit_w": null, "power_max_limit_w": null, "power_limit_at_bench_start_w": null, "note": "the single NVIDIA GeForce RTX 5090 Laptop GPU 24 GB SOLDERED to this machine's board at 00000000:01:00.0. There is no seat, no partner and no card to swap. \u26a0 THE TRAP ON THIS BOX IS NOT A STALE BOOT UNIT, IT IS A RUNNING DAEMON: nvidia-powerd.service (NVIDIA Dynamic Boost) floats this board's power limit between its default and its maximum against the CPU's draw, continuously and without asking, and it has been running since 2026-09-08. The lesson every leg of this ladder carries \u2014 a cap read once at the start of a night is not a cap \u2014 is true here in its strongest form: such a reading is a sample of a moving signal. Every arm reads the limit back in the same call as its own data, every stage reads it again at its close, and a stage that finds it moved STOPS rather than relabelling its file. The second trap is a CLOCK LOCK: ai-perf.service runs `nvidia-smi -lgc 1200,2550` at boot on this box and the cards that produced the rows this bench compares against ran unlocked, so a floored clock changes what a power cap means and every record carries clocks.sm min and max so the confound is visible rather than implied." } }, "memory_temp_support": { "checked_utc": "2026-09-21T20:54:23Z", "paths": { "nvidia-smi --query-gpu=temperature.memory": { "raw": "0, GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff, N/A", "readable": false }, "nvidia-smi -q -d TEMPERATURE": { "lines": [ "GPU Target Temperature : 87 C", "Memory Current Temp : N/A", "Memory Max Operating T.Limit Temp : N/A" ], "readable": false }, "NVML NVML_FI_DEV_MEMORY_TEMP": { "available": true, "field_id": 82, "cards": { "0": { "call_rc": 0, "field_rc": 3, "supported": false, "not_supported": true, "value_c": null } } } }, "readable": false, "verdict": "the memory die's temperature is NOT readable on these cards through any of the three paths asked, so the memory-temperature stop condition could not arm and NO memory temperature is reported anywhere in this bench. The core temperature stop and the driver's own thermal-slowdown reasons are the thermal instrument instead." }, "started_utc": "2026-09-21T20:54:23Z", "batch_size": 64, "words_per_text": 150, "corpus_seed": 20260916, "corpus_note": "The corpus is generated in bench_embed.py from a fixed 100-word vocabulary and seed 20260916: 64 texts of 150 words each, identical on every box and every run. No estate, user or document text is used, so no such text can end up in a result file.", "batches_scored": 5, "batches": [ { "batch_size": 64, "embeddings_returned": 64, "dimensions": 768, "wall_s": 0.2352, "ms_per_batch": 235.17, "texts_per_s": 272.147, "prompt_eval_count": 14448, "total_duration_ns": 227567391, "load_duration_ns": 11279055, "batch": 1, "power": { "idle_baseline": { "seconds": 2.0, "samples": 4, "power_w": { "min": 22.56, "median": 22.56, "max": 22.57, "mean": 22.56, "n": 4 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 4 }, "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 4 }, "power_w_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 22.56 }, "memory_used_mib_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 1284.0 }, "power_w_all_cards": 22.56, "ups_whole_box": { "load_pct": null, "realpower_w": null, "realpower_nominal_w": null, "status": null, "note": "the UPS's reading of EVERYTHING it feeds -- the whole box, not the cards. The card figures beside it are board power from nvidia-smi. Neither is the other." } }, "mean_w": 22.54, "max_w": 22.54, "over_idle_w": -0.02, "samples": 1, "per_gpu": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": { "index": "0", "name": "NVIDIA GeForce RTX 5090 Laptop GPU", "power_w": { "min": 22.54, "median": 22.54, "max": 22.54, "mean": 22.54, "n": 1 }, "power_limit_w": 150.0, "power_limit_field": "enforced.power.limit", "power_limit_enforced_w": 150.0, "cap": "dynamic-boost", "power_posture": "dynamic-boost", "power_limit_enforced_stats_w": { "min": 150.0, "median": 150.0, "max": 150.0, "mean": 150.0, "n": 1 }, "power_posture_note": "dynamic-boost -- nvidia-powerd running and no cap set: the board floats between its power.default_limit and its power.max_limit against the CPU's draw, DURING the scored run. `power_limit_enforced_stats_w` is that limit reduced over this run's own 2 Hz trace and is the only honest cap cell; `power_limit_enforced_w` is the FIRST sample and is kept only so the two can be compared. A single wattage in a cap column for this posture is a fabrication.", "cap_cell": "150 W, flat", "cap_cell_note": "the enforced limit read 150 W on every one of 1 samples of this run's 2 Hz trace. Under dynamic-boost that is a FINDING, not a range.", "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 1 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 1 }, "temperature_c": { "min": 35.0, "median": 35.0, "max": 35.0, "mean": 35.0, "n": 1 }, "memory_temperature_c": null, "fan_pct": null, "pcie_width_current": { "min": 8.0, "median": 8.0, "max": 8.0, "mean": 8.0, "n": 1 }, "clocks_sm_mhz": { "min": 1590.0, "median": 1590.0, "max": 1590.0, "mean": 1590.0, "n": 1 }, "clocks_mem_mhz": { "min": 14001.0, "median": 14001.0, "max": 14001.0, "mean": 14001.0, "n": 1 }, "throttle_sw_power_cap_fraction": 0.0, "throttle_hw_slowdown_fraction": 0.0, "throttle_sw_thermal_fraction": 0.0, "throttle_hw_thermal_fraction": 0.0, "n": 1, "busy": null, "busy_samples": 0 } } }, "energy_j_per_1k_texts": 82.83, "thermal": { "rule_version": 2, "measured_over": "the whole sampling window -- too few busy samples to separate decoding from idle, so this reading includes idle time", "busy_samples": null, "temp_max_c": 35.0, "temp_min_c": 35.0, "temp_rise_c": 0.0, "memory_temp_max_c": null, "memory_temp_median_c": null, "memory_temp_readable": false, "pcie_width_min": 8.0, "pcie_width_max": 8.0, "pcie_width_median": 8.0, "fan_mean_pct": null, "fan_max_pct": null, "clock_floor_mhz": 1590.0, "clock_max_mhz": 1590.0, "clock_median_mhz": 1590.0, "mem_clock_floor_mhz": 14001.0, "mem_clock_median_mhz": 14001.0, "mem_clock_max_mhz": 14001.0, "mem_clock_held": true, "memory_bus_width_bits": 256, "memory_technology": "GDDR7", "memory_bits_per_clock": null, "peak_mem_bandwidth_gbs_at_median_clock": null, "peak_mem_bandwidth_gbs_at_floor_clock": null, "peak_mem_bandwidth_gbs_at_max_clock": null, "peak_mem_bandwidth_withheld_because": "this harness's peak-bandwidth formula assumes two bits per memory clock per pin, which is exact for GDDR6 and GDDR6X and is checked against three published bandwidths. It has no receipt for the bits-per-clock of this board's memory, so the DERIVED bytes-per-second ceiling is withheld rather than computed from an unverified constant. The memory clock itself is a reading and is published unchanged.", "peak_mem_bandwidth_formula": "GB/s = clocks.mem MHz x 1e6 x x (bus bits / 8) / 1e9; a CEILING at the clock that was observed, never an achieved rate. The bits-per-clock is registered per board and is NOT defaulted.", "clock_dropped": false, "power_limit_w": 150.0, "power_mean_w": 22.54, "power_pct_of_limit": 15.0, "power_pinned_at_limit": false, "sw_power_cap_active_fraction": 0.0, "hw_slowdown_active_fraction": 0.0, "sw_thermal_active_fraction": 0.0, "hw_thermal_active_fraction": 0.0, "at_throttling_temperature": false, "stop_core_temp_c": 83.0, "stop_memory_temp_c": 100.0, "hit_core_temp_stop": false, "hit_memory_temp_stop": false, "throttle_temperature_c": 80.0, "temp_rising_within_run": false, "verdict": "no clock drop during decode" } }, { "batch_size": 64, "embeddings_returned": 64, "dimensions": 768, "wall_s": 0.2282, "ms_per_batch": 228.16, "texts_per_s": 280.504, "prompt_eval_count": 14448, "total_duration_ns": 220954514, "load_duration_ns": 13371512, "batch": 2, "power": { "idle_baseline": { "seconds": 2.0, "samples": 4, "power_w": { "min": 24.59, "median": 30.04, "max": 39.07, "mean": 30.94, "n": 4 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 4 }, "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 4 }, "power_w_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 30.94 }, "memory_used_mib_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 1284.0 }, "power_w_all_cards": 30.94, "ups_whole_box": { "load_pct": null, "realpower_w": null, "realpower_nominal_w": null, "status": null, "note": "the UPS's reading of EVERYTHING it feeds -- the whole box, not the cards. The card figures beside it are board power from nvidia-smi. Neither is the other." } }, "mean_w": 22.65, "max_w": 22.65, "over_idle_w": -8.29, "samples": 1, "per_gpu": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": { "index": "0", "name": "NVIDIA GeForce RTX 5090 Laptop GPU", "power_w": { "min": 22.65, "median": 22.65, "max": 22.65, "mean": 22.65, "n": 1 }, "power_limit_w": 150.0, "power_limit_field": "enforced.power.limit", "power_limit_enforced_w": 150.0, "cap": "dynamic-boost", "power_posture": "dynamic-boost", "power_limit_enforced_stats_w": { "min": 150.0, "median": 150.0, "max": 150.0, "mean": 150.0, "n": 1 }, "power_posture_note": "dynamic-boost -- nvidia-powerd running and no cap set: the board floats between its power.default_limit and its power.max_limit against the CPU's draw, DURING the scored run. `power_limit_enforced_stats_w` is that limit reduced over this run's own 2 Hz trace and is the only honest cap cell; `power_limit_enforced_w` is the FIRST sample and is kept only so the two can be compared. A single wattage in a cap column for this posture is a fabrication.", "cap_cell": "150 W, flat", "cap_cell_note": "the enforced limit read 150 W on every one of 1 samples of this run's 2 Hz trace. Under dynamic-boost that is a FINDING, not a range.", "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 1 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 1 }, "temperature_c": { "min": 35.0, "median": 35.0, "max": 35.0, "mean": 35.0, "n": 1 }, "memory_temperature_c": null, "fan_pct": null, "pcie_width_current": { "min": 8.0, "median": 8.0, "max": 8.0, "mean": 8.0, "n": 1 }, "clocks_sm_mhz": { "min": 1590.0, "median": 1590.0, "max": 1590.0, "mean": 1590.0, "n": 1 }, "clocks_mem_mhz": { "min": 14001.0, "median": 14001.0, "max": 14001.0, "mean": 14001.0, "n": 1 }, "throttle_sw_power_cap_fraction": 0.0, "throttle_hw_slowdown_fraction": 0.0, "throttle_sw_thermal_fraction": 0.0, "throttle_hw_thermal_fraction": 0.0, "n": 1, "busy": null, "busy_samples": 0 } } }, "energy_j_per_1k_texts": 80.76, "thermal": { "rule_version": 2, "measured_over": "the whole sampling window -- too few busy samples to separate decoding from idle, so this reading includes idle time", "busy_samples": null, "temp_max_c": 35.0, "temp_min_c": 35.0, "temp_rise_c": 0.0, "memory_temp_max_c": null, "memory_temp_median_c": null, "memory_temp_readable": false, "pcie_width_min": 8.0, "pcie_width_max": 8.0, "pcie_width_median": 8.0, "fan_mean_pct": null, "fan_max_pct": null, "clock_floor_mhz": 1590.0, "clock_max_mhz": 1590.0, "clock_median_mhz": 1590.0, "mem_clock_floor_mhz": 14001.0, "mem_clock_median_mhz": 14001.0, "mem_clock_max_mhz": 14001.0, "mem_clock_held": true, "memory_bus_width_bits": 256, "memory_technology": "GDDR7", "memory_bits_per_clock": null, "peak_mem_bandwidth_gbs_at_median_clock": null, "peak_mem_bandwidth_gbs_at_floor_clock": null, "peak_mem_bandwidth_gbs_at_max_clock": null, "peak_mem_bandwidth_withheld_because": "this harness's peak-bandwidth formula assumes two bits per memory clock per pin, which is exact for GDDR6 and GDDR6X and is checked against three published bandwidths. It has no receipt for the bits-per-clock of this board's memory, so the DERIVED bytes-per-second ceiling is withheld rather than computed from an unverified constant. The memory clock itself is a reading and is published unchanged.", "peak_mem_bandwidth_formula": "GB/s = clocks.mem MHz x 1e6 x x (bus bits / 8) / 1e9; a CEILING at the clock that was observed, never an achieved rate. The bits-per-clock is registered per board and is NOT defaulted.", "clock_dropped": false, "power_limit_w": 150.0, "power_mean_w": 22.65, "power_pct_of_limit": 15.1, "power_pinned_at_limit": false, "sw_power_cap_active_fraction": 0.0, "hw_slowdown_active_fraction": 0.0, "sw_thermal_active_fraction": 0.0, "hw_thermal_active_fraction": 0.0, "at_throttling_temperature": false, "stop_core_temp_c": 83.0, "stop_memory_temp_c": 100.0, "hit_core_temp_stop": false, "hit_memory_temp_stop": false, "throttle_temperature_c": 80.0, "temp_rising_within_run": false, "verdict": "no clock drop during decode" } }, { "batch_size": 64, "embeddings_returned": 64, "dimensions": 768, "wall_s": 0.233, "ms_per_batch": 233.01, "texts_per_s": 274.663, "prompt_eval_count": 14448, "total_duration_ns": 226555043, "load_duration_ns": 11169201, "batch": 3, "power": { "idle_baseline": { "seconds": 2.0, "samples": 4, "power_w": { "min": 28.86, "median": 32.17, "max": 36.78, "mean": 32.49, "n": 4 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 4 }, "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 4 }, "power_w_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 32.49 }, "memory_used_mib_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 1284.0 }, "power_w_all_cards": 32.49, "ups_whole_box": { "load_pct": null, "realpower_w": null, "realpower_nominal_w": null, "status": null, "note": "the UPS's reading of EVERYTHING it feeds -- the whole box, not the cards. The card figures beside it are board power from nvidia-smi. Neither is the other." } }, "mean_w": 22.63, "max_w": 22.63, "over_idle_w": -9.86, "samples": 1, "per_gpu": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": { "index": "0", "name": "NVIDIA GeForce RTX 5090 Laptop GPU", "power_w": { "min": 22.63, "median": 22.63, "max": 22.63, "mean": 22.63, "n": 1 }, "power_limit_w": 150.0, "power_limit_field": "enforced.power.limit", "power_limit_enforced_w": 150.0, "cap": "dynamic-boost", "power_posture": "dynamic-boost", "power_limit_enforced_stats_w": { "min": 150.0, "median": 150.0, "max": 150.0, "mean": 150.0, "n": 1 }, "power_posture_note": "dynamic-boost -- nvidia-powerd running and no cap set: the board floats between its power.default_limit and its power.max_limit against the CPU's draw, DURING the scored run. `power_limit_enforced_stats_w` is that limit reduced over this run's own 2 Hz trace and is the only honest cap cell; `power_limit_enforced_w` is the FIRST sample and is kept only so the two can be compared. A single wattage in a cap column for this posture is a fabrication.", "cap_cell": "150 W, flat", "cap_cell_note": "the enforced limit read 150 W on every one of 1 samples of this run's 2 Hz trace. Under dynamic-boost that is a FINDING, not a range.", "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 1 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 1 }, "temperature_c": { "min": 36.0, "median": 36.0, "max": 36.0, "mean": 36.0, "n": 1 }, "memory_temperature_c": null, "fan_pct": null, "pcie_width_current": { "min": 8.0, "median": 8.0, "max": 8.0, "mean": 8.0, "n": 1 }, "clocks_sm_mhz": { "min": 1590.0, "median": 1590.0, "max": 1590.0, "mean": 1590.0, "n": 1 }, "clocks_mem_mhz": { "min": 14001.0, "median": 14001.0, "max": 14001.0, "mean": 14001.0, "n": 1 }, "throttle_sw_power_cap_fraction": 0.0, "throttle_hw_slowdown_fraction": 0.0, "throttle_sw_thermal_fraction": 0.0, "throttle_hw_thermal_fraction": 0.0, "n": 1, "busy": null, "busy_samples": 0 } } }, "energy_j_per_1k_texts": 82.39, "thermal": { "rule_version": 2, "measured_over": "the whole sampling window -- too few busy samples to separate decoding from idle, so this reading includes idle time", "busy_samples": null, "temp_max_c": 36.0, "temp_min_c": 36.0, "temp_rise_c": 0.0, "memory_temp_max_c": null, "memory_temp_median_c": null, "memory_temp_readable": false, "pcie_width_min": 8.0, "pcie_width_max": 8.0, "pcie_width_median": 8.0, "fan_mean_pct": null, "fan_max_pct": null, "clock_floor_mhz": 1590.0, "clock_max_mhz": 1590.0, "clock_median_mhz": 1590.0, "mem_clock_floor_mhz": 14001.0, "mem_clock_median_mhz": 14001.0, "mem_clock_max_mhz": 14001.0, "mem_clock_held": true, "memory_bus_width_bits": 256, "memory_technology": "GDDR7", "memory_bits_per_clock": null, "peak_mem_bandwidth_gbs_at_median_clock": null, "peak_mem_bandwidth_gbs_at_floor_clock": null, "peak_mem_bandwidth_gbs_at_max_clock": null, "peak_mem_bandwidth_withheld_because": "this harness's peak-bandwidth formula assumes two bits per memory clock per pin, which is exact for GDDR6 and GDDR6X and is checked against three published bandwidths. It has no receipt for the bits-per-clock of this board's memory, so the DERIVED bytes-per-second ceiling is withheld rather than computed from an unverified constant. The memory clock itself is a reading and is published unchanged.", "peak_mem_bandwidth_formula": "GB/s = clocks.mem MHz x 1e6 x x (bus bits / 8) / 1e9; a CEILING at the clock that was observed, never an achieved rate. The bits-per-clock is registered per board and is NOT defaulted.", "clock_dropped": false, "power_limit_w": 150.0, "power_mean_w": 22.63, "power_pct_of_limit": 15.1, "power_pinned_at_limit": false, "sw_power_cap_active_fraction": 0.0, "hw_slowdown_active_fraction": 0.0, "sw_thermal_active_fraction": 0.0, "hw_thermal_active_fraction": 0.0, "at_throttling_temperature": false, "stop_core_temp_c": 83.0, "stop_memory_temp_c": 100.0, "hit_core_temp_stop": false, "hit_memory_temp_stop": false, "throttle_temperature_c": 80.0, "temp_rising_within_run": false, "verdict": "no clock drop during decode" } }, { "batch_size": 64, "embeddings_returned": 64, "dimensions": 768, "wall_s": 0.2371, "ms_per_batch": 237.08, "texts_per_s": 269.952, "prompt_eval_count": 14448, "total_duration_ns": 229298904, "load_duration_ns": 10986477, "batch": 4, "power": { "idle_baseline": { "seconds": 2.0, "samples": 4, "power_w": { "min": 35.73, "median": 41.2, "max": 41.83, "mean": 39.99, "n": 4 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 4 }, "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 4 }, "power_w_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 39.99 }, "memory_used_mib_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 1284.0 }, "power_w_all_cards": 39.99, "ups_whole_box": { "load_pct": null, "realpower_w": null, "realpower_nominal_w": null, "status": null, "note": "the UPS's reading of EVERYTHING it feeds -- the whole box, not the cards. The card figures beside it are board power from nvidia-smi. Neither is the other." } }, "mean_w": 22.62, "max_w": 22.62, "over_idle_w": -17.37, "samples": 1, "per_gpu": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": { "index": "0", "name": "NVIDIA GeForce RTX 5090 Laptop GPU", "power_w": { "min": 22.62, "median": 22.62, "max": 22.62, "mean": 22.62, "n": 1 }, "power_limit_w": 150.0, "power_limit_field": "enforced.power.limit", "power_limit_enforced_w": 150.0, "cap": "dynamic-boost", "power_posture": "dynamic-boost", "power_limit_enforced_stats_w": { "min": 150.0, "median": 150.0, "max": 150.0, "mean": 150.0, "n": 1 }, "power_posture_note": "dynamic-boost -- nvidia-powerd running and no cap set: the board floats between its power.default_limit and its power.max_limit against the CPU's draw, DURING the scored run. `power_limit_enforced_stats_w` is that limit reduced over this run's own 2 Hz trace and is the only honest cap cell; `power_limit_enforced_w` is the FIRST sample and is kept only so the two can be compared. A single wattage in a cap column for this posture is a fabrication.", "cap_cell": "150 W, flat", "cap_cell_note": "the enforced limit read 150 W on every one of 1 samples of this run's 2 Hz trace. Under dynamic-boost that is a FINDING, not a range.", "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 1 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 1 }, "temperature_c": { "min": 36.0, "median": 36.0, "max": 36.0, "mean": 36.0, "n": 1 }, "memory_temperature_c": null, "fan_pct": null, "pcie_width_current": { "min": 8.0, "median": 8.0, "max": 8.0, "mean": 8.0, "n": 1 }, "clocks_sm_mhz": { "min": 1590.0, "median": 1590.0, "max": 1590.0, "mean": 1590.0, "n": 1 }, "clocks_mem_mhz": { "min": 14001.0, "median": 14001.0, "max": 14001.0, "mean": 14001.0, "n": 1 }, "throttle_sw_power_cap_fraction": 0.0, "throttle_hw_slowdown_fraction": 0.0, "throttle_sw_thermal_fraction": 0.0, "throttle_hw_thermal_fraction": 0.0, "n": 1, "busy": null, "busy_samples": 0 } } }, "energy_j_per_1k_texts": 83.8, "thermal": { "rule_version": 2, "measured_over": "the whole sampling window -- too few busy samples to separate decoding from idle, so this reading includes idle time", "busy_samples": null, "temp_max_c": 36.0, "temp_min_c": 36.0, "temp_rise_c": 0.0, "memory_temp_max_c": null, "memory_temp_median_c": null, "memory_temp_readable": false, "pcie_width_min": 8.0, "pcie_width_max": 8.0, "pcie_width_median": 8.0, "fan_mean_pct": null, "fan_max_pct": null, "clock_floor_mhz": 1590.0, "clock_max_mhz": 1590.0, "clock_median_mhz": 1590.0, "mem_clock_floor_mhz": 14001.0, "mem_clock_median_mhz": 14001.0, "mem_clock_max_mhz": 14001.0, "mem_clock_held": true, "memory_bus_width_bits": 256, "memory_technology": "GDDR7", "memory_bits_per_clock": null, "peak_mem_bandwidth_gbs_at_median_clock": null, "peak_mem_bandwidth_gbs_at_floor_clock": null, "peak_mem_bandwidth_gbs_at_max_clock": null, "peak_mem_bandwidth_withheld_because": "this harness's peak-bandwidth formula assumes two bits per memory clock per pin, which is exact for GDDR6 and GDDR6X and is checked against three published bandwidths. It has no receipt for the bits-per-clock of this board's memory, so the DERIVED bytes-per-second ceiling is withheld rather than computed from an unverified constant. The memory clock itself is a reading and is published unchanged.", "peak_mem_bandwidth_formula": "GB/s = clocks.mem MHz x 1e6 x x (bus bits / 8) / 1e9; a CEILING at the clock that was observed, never an achieved rate. The bits-per-clock is registered per board and is NOT defaulted.", "clock_dropped": false, "power_limit_w": 150.0, "power_mean_w": 22.62, "power_pct_of_limit": 15.1, "power_pinned_at_limit": false, "sw_power_cap_active_fraction": 0.0, "hw_slowdown_active_fraction": 0.0, "sw_thermal_active_fraction": 0.0, "hw_thermal_active_fraction": 0.0, "at_throttling_temperature": false, "stop_core_temp_c": 83.0, "stop_memory_temp_c": 100.0, "hit_core_temp_stop": false, "hit_memory_temp_stop": false, "throttle_temperature_c": 80.0, "temp_rising_within_run": false, "verdict": "no clock drop during decode" } }, { "batch_size": 64, "embeddings_returned": 64, "dimensions": 768, "wall_s": 0.2375, "ms_per_batch": 237.55, "texts_per_s": 269.421, "prompt_eval_count": 14448, "total_duration_ns": 229594375, "load_duration_ns": 12708267, "batch": 5, "power": { "idle_baseline": { "seconds": 2.0, "samples": 4, "power_w": { "min": 35.39, "median": 35.51, "max": 38.3, "mean": 36.18, "n": 4 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 4 }, "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 4 }, "power_w_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 36.18 }, "memory_used_mib_per_card": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": 1284.0 }, "power_w_all_cards": 36.18, "ups_whole_box": { "load_pct": null, "realpower_w": null, "realpower_nominal_w": null, "status": null, "note": "the UPS's reading of EVERYTHING it feeds -- the whole box, not the cards. The card figures beside it are board power from nvidia-smi. Neither is the other." } }, "mean_w": 22.47, "max_w": 22.47, "over_idle_w": -13.71, "samples": 1, "per_gpu": { "GPU-edff232c-7dbf-2bac-07fb-921a7c9eecff": { "index": "0", "name": "NVIDIA GeForce RTX 5090 Laptop GPU", "power_w": { "min": 22.47, "median": 22.47, "max": 22.47, "mean": 22.47, "n": 1 }, "power_limit_w": 150.0, "power_limit_field": "enforced.power.limit", "power_limit_enforced_w": 150.0, "cap": "dynamic-boost", "power_posture": "dynamic-boost", "power_limit_enforced_stats_w": { "min": 150.0, "median": 150.0, "max": 150.0, "mean": 150.0, "n": 1 }, "power_posture_note": "dynamic-boost -- nvidia-powerd running and no cap set: the board floats between its power.default_limit and its power.max_limit against the CPU's draw, DURING the scored run. `power_limit_enforced_stats_w` is that limit reduced over this run's own 2 Hz trace and is the only honest cap cell; `power_limit_enforced_w` is the FIRST sample and is kept only so the two can be compared. A single wattage in a cap column for this posture is a fabrication.", "cap_cell": "150 W, flat", "cap_cell_note": "the enforced limit read 150 W on every one of 1 samples of this run's 2 Hz trace. Under dynamic-boost that is a FINDING, not a range.", "memory_used_mib": { "min": 1284.0, "median": 1284.0, "max": 1284.0, "mean": 1284.0, "n": 1 }, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 1 }, "temperature_c": { "min": 36.0, "median": 36.0, "max": 36.0, "mean": 36.0, "n": 1 }, "memory_temperature_c": null, "fan_pct": null, "pcie_width_current": { "min": 8.0, "median": 8.0, "max": 8.0, "mean": 8.0, "n": 1 }, "clocks_sm_mhz": { "min": 1590.0, "median": 1590.0, "max": 1590.0, "mean": 1590.0, "n": 1 }, "clocks_mem_mhz": { "min": 14001.0, "median": 14001.0, "max": 14001.0, "mean": 14001.0, "n": 1 }, "throttle_sw_power_cap_fraction": 0.0, "throttle_hw_slowdown_fraction": 0.0, "throttle_sw_thermal_fraction": 0.0, "throttle_hw_thermal_fraction": 0.0, "n": 1, "busy": null, "busy_samples": 0 } } }, "energy_j_per_1k_texts": 83.38, "thermal": { "rule_version": 2, "measured_over": "the whole sampling window -- too few busy samples to separate decoding from idle, so this reading includes idle time", "busy_samples": null, "temp_max_c": 36.0, "temp_min_c": 36.0, "temp_rise_c": 0.0, "memory_temp_max_c": null, "memory_temp_median_c": null, "memory_temp_readable": false, "pcie_width_min": 8.0, "pcie_width_max": 8.0, "pcie_width_median": 8.0, "fan_mean_pct": null, "fan_max_pct": null, "clock_floor_mhz": 1590.0, "clock_max_mhz": 1590.0, "clock_median_mhz": 1590.0, "mem_clock_floor_mhz": 14001.0, "mem_clock_median_mhz": 14001.0, "mem_clock_max_mhz": 14001.0, "mem_clock_held": true, "memory_bus_width_bits": 256, "memory_technology": "GDDR7", "memory_bits_per_clock": null, "peak_mem_bandwidth_gbs_at_median_clock": null, "peak_mem_bandwidth_gbs_at_floor_clock": null, "peak_mem_bandwidth_gbs_at_max_clock": null, "peak_mem_bandwidth_withheld_because": "this harness's peak-bandwidth formula assumes two bits per memory clock per pin, which is exact for GDDR6 and GDDR6X and is checked against three published bandwidths. It has no receipt for the bits-per-clock of this board's memory, so the DERIVED bytes-per-second ceiling is withheld rather than computed from an unverified constant. The memory clock itself is a reading and is published unchanged.", "peak_mem_bandwidth_formula": "GB/s = clocks.mem MHz x 1e6 x x (bus bits / 8) / 1e9; a CEILING at the clock that was observed, never an achieved rate. The bits-per-clock is registered per board and is NOT defaulted.", "clock_dropped": false, "power_limit_w": 150.0, "power_mean_w": 22.47, "power_pct_of_limit": 15.0, "power_pinned_at_limit": false, "sw_power_cap_active_fraction": 0.0, "hw_slowdown_active_fraction": 0.0, "sw_thermal_active_fraction": 0.0, "hw_thermal_active_fraction": 0.0, "at_throttling_temperature": false, "stop_core_temp_c": 83.0, "stop_memory_temp_c": 100.0, "hit_core_temp_stop": false, "hit_memory_temp_stop": false, "throttle_temperature_c": 80.0, "temp_rising_within_run": false, "verdict": "no clock drop during decode" } } ], "errors": [], "ollama_version": "0.32.13", "resident_before": false, "expires_at_before": null, "pinned_before": false, "keep_alive_policy": "bench instance: held for the arm's duration", "keep_alive_field_omitted": false, "warmup_discarded": { "batch_size": 64, "embeddings_returned": 64, "dimensions": 768, "wall_s": 1.7483, "ms_per_batch": 1748.3, "texts_per_s": 36.607, "prompt_eval_count": 14448, "total_duration_ns": 1740477226, "load_duration_ns": 1328945690 }, "ps_after_load": [ { "name": "nomic-embed-text:latest", "model": "nomic-embed-text:latest", "size": 323150151, "digest": "0a109f422b47e3a30ba2b10eca18548e944e8a23073ee3f3e947efcf3c45e59f", "details": { "parent_model": "", "format": "gguf", "family": "nomic-bert", "families": [ "nomic-bert" ], "parameter_size": "137M", "quantization_level": "F16" }, "expires_at": "2026-09-21T21:04:25.070933768Z", "size_vram": 323150151, "context_length": 2048 } ], "size_total": 323150151, "size_vram": 323150151, "fit_verdict": "fits: fully resident on the card", "contention_gate": { "attempts": [ { "window_s": 10.0, "samples": 20, "utilization_pct": { "min": 0.0, "median": 0.0, "max": 0.0, "mean": 0.0, "n": 20 }, "power_w": { "min": 22.56, "median": 22.6, "max": 46.56, "mean": 26.24, "n": 20 }, "card_seen": true, "attempt": 1, "mean_util_pct": 0.0, "verdict": "quiet: mean 0.0% <= 5.0%" } ], "passed": true }, "spread_gate": { "limit": 0.15, "values": [ 235.17, 228.16, 233.01, 237.08, 237.55 ], "spread_fraction": 0.0399, "verdict": "spread 4.0% of the median within the 15% gate" }, "voided_by_gate": false, "derived": { "ms_per_batch": { "min": 228.16, "median": 235.17, "max": 237.55, "mean": 234.19, "n": 5 }, "texts_per_s": { "min": 269.421, "median": 272.147, "max": 280.504, "mean": 273.337, "n": 5 }, "power_mean_w": { "min": 22.47, "median": 22.62, "max": 22.65, "mean": 22.58, "n": 5 }, "power_over_idle_w": { "min": -17.37, "median": -9.86, "max": -0.02, "mean": -9.85, "n": 5 }, "energy_j_per_1k_texts": { "min": 80.76, "median": 82.83, "max": 83.8, "mean": 82.63, "n": 5 }, "temp_max_c": { "min": 35.0, "median": 36.0, "max": 36.0, "mean": 35.6, "n": 5 }, "fan_mean_pct": null, "dimensions": 768 }, "ps_final": [], "expires_at_after": null, "pinned_after": false, "finished_utc": "2026-09-21T20:54:46Z", "headline": "one-dynboost: 272.1 texts/s on a batch of 64 (235.17 ms per batch)" }