[
  {
    "50.00 percentile latency (ns)": 19083568,
    "90.00 percentile latency (ns)": 33987728,
    "90th percentile latency (ns)": 33987728,
    "95.00 percentile latency (ns)": 38361254,
    "97.00 percentile latency (ns)": 41246106,
    "99.00 percentile latency (ns)": 43869085,
    "99.90 percentile latency (ns)": 46047177,
    "Max latency (ns)": 49652659,
    "Mean latency (ns)": 20854850,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 5762758,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "QPS w/ loadgen overhead": 47.79,
    "QPS w/o loadgen overhead": 47.95,
    "Result is": "VALID",
    "SUT name": "RNNT SERVER",
    "Scenario": "singlestream",
    "accelerator_cooling_type": "passive",
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "24GB",
    "accelerator_memory_configuration": "GDDR6",
    "accelerator_model_name": "NVIDIA A10",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 1,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "2.12.3",
    "characteristics.90th_percentile_latency_ms": 33.987728,
    "characteristics.90th_percentile_latency_ns": 33987728.0,
    "characteristics.90th_percentile_latency_s": 0.033987728,
    "characteristics.90th_percentile_latency_us": 33987.728,
    "characteristics.accuracy": 92.57029470788518,
    "characteristics.word error rate": 7.429705292114816,
    "ck_system": "XE2420_A10x1_TRT",
    "ck_used": false,
    "cooling": "",
    "dataset": "LibriSpeech",
    "dataset_link": "",
    "dim_x_default": "characteristics.90th_percentile_latency_ms",
    "dim_y_default": "characteristics.accuracy",
    "dim_y_maximize": true,
    "division": "closed",
    "filesystem": "ext3/ext4",
    "formal_model": "rnn-t",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "",
    "framework": "TensorRT 8.0.2, CUDA 11.3",
    "host_cooling_type": "air",
    "host_memory_capacity": "512 GB",
    "host_memory_configuration": "",
    "host_networking": "",
    "host_networking_topology": "",
    "host_processor_caches": "",
    "host_processor_core_count": 24,
    "host_processor_frequency": "",
    "host_processor_interconnect": "",
    "host_processor_model_name": "Intel(R) Xeon(R) Gold 6252N CPU @ 2.30GHz",
    "host_processors_per_node": 2,
    "host_storage_capacity": "4 TB",
    "host_storage_type": "NVMe SSD",
    "hw_notes": "ECC on",
    "informal_model": "rnnt",
    "input_data_types": "fp16",
    "key.accuracy": "characteristics.accuracy",
    "management_firmware_version": "5.00.00.00",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 1024,
    "mlperf_version": 1.1,
    "normalize_cores": 1,
    "normalize_processors": 1,
    "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Dell/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Dell/results/XE2420_A10x1_TRT",
    "number_of_nodes": 1,
    "operating_system": "Ubuntu 20.04.2",
    "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.57.02, DALI 0.31.0",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 2513,
    "power_settings": "closed/Dell/power/XE2420_A10x1_power_settings.md",
    "print_timestamps": 0,
    "problem": false,
    "psu_details": "2x2000W",
    "qsl_rng_seed": 1624344308455410291,
    "retraining": "N",
    "sample_index_rng_seed": 517984244576520566,
    "samples_per_query": 1,
    "schedule_rng_seed": 10051496985653635065,
    "starting_weights_filename": "DistributedDataParallel_1576581068.9962234-epoch-100.pt",
    "status": "available",
    "submitter": "Dell",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell",
    "sw_notes": "",
    "system_cooling_type": "air",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/XE2420_A10x1_TRT",
    "system_name": "Dell EMC PowerEdge XE2420 (1x A10, TensorRT)",
    "system_type": "edge",
    "target_latency (ns)": 0,
    "target_qps": 40,
    "task": "speech recognition",
    "task2": "speech recognition",
    "total_cores": 48,
    "uid": "31ec548df8a58854",
    "use_accelerator": true,
    "weight_data_types": "fp16",
    "weight_transformations": "quantization, affine fusion"
  },
  {
    "50.00 percentile latency (ns)": 19192478,
    "90.00 percentile latency (ns)": 34132335,
    "90th percentile latency (ns)": 34132335,
    "95.00 percentile latency (ns)": 38557600,
    "97.00 percentile latency (ns)": 41439362,
    "99.00 percentile latency (ns)": 44129425,
    "99.90 percentile latency (ns)": 46342277,
    "Max latency (ns)": 49937555,
    "Mean latency (ns)": 20974035,
    "Min duration satisfied": "Yes",
    "Min latency (ns)": 5826969,
    "Min queries satisfied": "Yes",
    "Mode": "PerformanceOnly",
    "QPS w/ loadgen overhead": 47.62,
    "QPS w/o loadgen overhead": 47.68,
    "Result is": "VALID",
    "SUT name": "RNNT SERVER",
    "Scenario": "singlestream",
    "accelerator_cooling_type": "passive",
    "accelerator_frequency": "",
    "accelerator_host_interconnect": "",
    "accelerator_interconnect": "",
    "accelerator_interconnect_topology": "",
    "accelerator_memory_capacity": "24GB",
    "accelerator_memory_configuration": "GDDR6",
    "accelerator_model_name": "NVIDIA A10",
    "accelerator_on-chip_memories": "",
    "accelerators_per_node": 1,
    "accuracy_log_probability": 0,
    "accuracy_log_rng_seed": 0,
    "accuracy_log_sampling_target": 0,
    "boot_firmware_version": "2.12.3",
    "characteristics.90th_percentile_latency_ms": 34.132335,
    "characteristics.90th_percentile_latency_ns": 34132335.0,
    "characteristics.90th_percentile_latency_s": 0.034132335,
    "characteristics.90th_percentile_latency_us": 34132.335,
    "characteristics.accuracy": 92.56578502333763,
    "characteristics.power": 8.577218416295999,
    "characteristics.power.normalized_per_core": 8.577218416295999,
    "characteristics.power.normalized_per_processor": 8.577218416295999,
    "characteristics.word error rate": 7.434214976662383,
    "ck_system": "XE2420_A10x1_TRT_MaxQ",
    "ck_used": false,
    "cooling": "",
    "dataset": "LibriSpeech",
    "dataset_link": "",
    "dim_x_default": "characteristics.90th_percentile_latency_ms",
    "dim_y_default": "characteristics.accuracy",
    "dim_y_maximize": true,
    "division": "closed",
    "filesystem": "ext3/ext4",
    "formal_model": "rnn-t",
    "formal_model_accuracy": 99.0,
    "formal_model_link": "",
    "framework": "TensorRT 8.0.2, CUDA 11.3",
    "host_cooling_type": "air",
    "host_memory_capacity": "512 GB",
    "host_memory_configuration": "",
    "host_networking": "",
    "host_networking_topology": "",
    "host_processor_caches": "",
    "host_processor_core_count": 24,
    "host_processor_frequency": "",
    "host_processor_interconnect": "",
    "host_processor_model_name": "Intel(R) Xeon(R) Gold 6252N CPU @ 2.30GHz",
    "host_processors_per_node": 2,
    "host_storage_capacity": "4 TB",
    "host_storage_type": "NVMe SSD",
    "hw_notes": "ECC on",
    "informal_model": "rnnt",
    "input_data_types": "fp16",
    "key.accuracy": "characteristics.accuracy",
    "management_firmware_version": "5.00.00.00",
    "max_async_queries": 1,
    "max_duration (ms)": 0,
    "max_query_count": 0,
    "min_duration (ms)": 600000,
    "min_query_count": 1024,
    "mlperf_version": 1.1,
    "normalize_cores": 1,
    "normalize_processors": 1,
    "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Dell/code",
    "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Dell/results/XE2420_A10x1_TRT_MaxQ",
    "number_of_nodes": 1,
    "operating_system": "Ubuntu 20.04.2",
    "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.57.02, DALI 0.31.0",
    "performance_issue_same": 0,
    "performance_issue_same_index": 0,
    "performance_issue_unique": 0,
    "performance_sample_count": 2513,
    "power_settings": "closed/Dell/power/XE2420_A10x1_power_settings.md",
    "print_timestamps": 0,
    "problem": false,
    "psu_details": "2x2000W",
    "qsl_rng_seed": 1624344308455410291,
    "retraining": "N",
    "sample_index_rng_seed": 517984244576520566,
    "samples_per_query": 1,
    "schedule_rng_seed": 10051496985653635065,
    "starting_weights_filename": "DistributedDataParallel_1576581068.9962234-epoch-100.pt",
    "status": "available",
    "submitter": "Dell",
    "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell",
    "sw_notes": "",
    "system_cooling_type": "air",
    "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/XE2420_A10x1_TRT_MaxQ",
    "system_name": "Dell EMC PowerEdge XE2420 (1x A10, MaxQ, TensorRT)",
    "system_type": "edge",
    "target_latency (ns)": 0,
    "target_qps": 40,
    "task": "speech recognition",
    "task2": "speech recognition",
    "total_cores": 48,
    "uid": "69150f9db281c334",
    "use_accelerator": true,
    "weight_data_types": "fp16",
    "weight_transformations": "quantization, affine fusion"
  }
]