[ { "50.00 percentile latency (ns)": 510458, "90.00 percentile latency (ns)": 517067, "90th percentile latency (ns)": 517067, "95.00 percentile latency (ns)": 519805, "97.00 percentile latency (ns)": 521886, "99.00 percentile latency (ns)": 525876, "99.90 percentile latency (ns)": 731785, "Max latency (ns)": 4980275, "Mean latency (ns)": 512157, "Min duration satisfied": "Yes", "Min latency (ns)": 477628, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 1903.2, "QPS w/o loadgen overhead": 1952.53, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "", "accelerator_model_name": "NVIDIA A100-PCIe", "accelerator_on-chip_memories": "", "accelerators_per_node": 2, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.90th_percentile_latency_ms": 0.517067, "characteristics.90th_percentile_latency_ns": 517067.0, "characteristics.90th_percentile_latency_s": 0.000517067, "characteristics.90th_percentile_latency_us": 517.067, "characteristics.accuracy": 75.892, "characteristics.good": 37946, "characteristics.power": 0.20108790458175768, "characteristics.power.normalized_per_core": 0.10054395229087884, "characteristics.power.normalized_per_processor": 0.10054395229087884, "characteristics.total": 50000, "ck_system": "NE5260M5_A100-PCIex2_TRT_MaxQ", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.1, CUDA 11.3", "host_memory_capacity": "1 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 28, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Gold 6258R", "host_processors_per_node": 2, "host_storage_capacity": "2.9 TB * 2", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 2, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/results/NE5260M5_A100-PCIex2_TRT_MaxQ", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.1, 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": 2048, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/NE5260M5_A100-PCIex2_TRT_MaxQ", "system_name": "NE5260M5 (2x A100-PCIe, TensorRT)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1515.15, "task": "image classification", "task2": "image classification", "total_cores": 56, "uid": "4a9fcf6cb2ce11e5", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 426523, "90.00 percentile latency (ns)": 428528, "90th percentile latency (ns)": 428528, "95.00 percentile latency (ns)": 429135, "97.00 percentile latency (ns)": 429557, "99.00 percentile latency (ns)": 431705, "99.90 percentile latency (ns)": 455208, "Max latency (ns)": 4987091, "Mean latency (ns)": 426839, "Min duration satisfied": "Yes", "Min latency (ns)": 418679, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 2320.54, "QPS w/o loadgen overhead": 2342.8, "Result is": "VALID", "SUT name": "Triton_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80 GB", "accelerator_memory_configuration": "", "accelerator_model_name": "NVIDIA A100-SXM-80GB", "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": "", "characteristics.90th_percentile_latency_ms": 0.428528, "characteristics.90th_percentile_latency_ns": 428528.0, "characteristics.90th_percentile_latency_s": 0.000428528, "characteristics.90th_percentile_latency_us": 428.528, "characteristics.accuracy": 75.894, "characteristics.good": 37947, "characteristics.total": 50000, "ck_system": "NF5688M6_A100-SXM-80GBx1_TRT_Triton", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.1, CUDA 11.3", "host_memory_capacity": "2 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 32, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8358", "host_processors_per_node": 2, "host_storage_capacity": "28 TB + 448GB*2", "host_storage_type": "NVMe SSD + M.2", "hw_notes": "NVIDIA A100-SXM-80GB-500W", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/results/NF5688M6_A100-SXM-80GBx1_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.1, CUDA 11.3, cuDNN 8.2.1, Driver 470.57.02, DALI 1.0.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 2048, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/NF5688M6_A100-SXM-80GBx1_TRT_Triton", "system_name": "NF5688M6 (8x A100-SXM-80GB, TensorRT, Triton)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1515.15, "task": "image classification", "task2": "image classification", "total_cores": 64, "uid": "f874f65837e713a6", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 461317, "90.00 percentile latency (ns)": 467102, "90th percentile latency (ns)": 467102, "95.00 percentile latency (ns)": 469010, "97.00 percentile latency (ns)": 470038, "99.00 percentile latency (ns)": 472375, "99.90 percentile latency (ns)": 526146, "Max latency (ns)": 6443382, "Mean latency (ns)": 462755, "Min duration satisfied": "Yes", "Min latency (ns)": 442260, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 2140.09, "QPS w/o loadgen overhead": 2160.97, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "", "accelerator_model_name": "NVIDIA A100-PCIe", "accelerator_on-chip_memories": "", "accelerators_per_node": 2, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.90th_percentile_latency_ms": 0.467102, "characteristics.90th_percentile_latency_ns": 467102.0, "characteristics.90th_percentile_latency_s": 0.000467102, "characteristics.90th_percentile_latency_us": 467.102, "characteristics.accuracy": 75.886, "characteristics.good": 37943, "characteristics.total": 50000, "ck_system": "NE5260M5_A100-PCIex2_TRT", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.1, CUDA 11.3", "host_memory_capacity": "1 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 28, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Gold 6258R", "host_processors_per_node": 2, "host_storage_capacity": "2.9 TB * 2", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 2, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/results/NE5260M5_A100-PCIex2_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.1, 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": 2048, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/NE5260M5_A100-PCIex2_TRT", "system_name": "NE5260M5 (2x A100-PCIe, TensorRT)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1515.15, "task": "image classification", "task2": "image classification", "total_cores": 56, "uid": "660ee7354735b73d", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 441285, "90.00 percentile latency (ns)": 444291, "90th percentile latency (ns)": 444291, "95.00 percentile latency (ns)": 445473, "97.00 percentile latency (ns)": 446518, "99.00 percentile latency (ns)": 448736, "99.90 percentile latency (ns)": 471303, "Max latency (ns)": 5175042, "Mean latency (ns)": 441437, "Min duration satisfied": "Yes", "Min latency (ns)": 432205, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 2241.12, "QPS w/o loadgen overhead": 2265.33, "Result is": "VALID", "SUT name": "Triton_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "", "accelerator_model_name": "NVIDIA A100-PCIe", "accelerator_on-chip_memories": "", "accelerators_per_node": 2, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.90th_percentile_latency_ms": 0.444291, "characteristics.90th_percentile_latency_ns": 444291.0, "characteristics.90th_percentile_latency_s": 0.000444291, "characteristics.90th_percentile_latency_us": 444.291, "characteristics.accuracy": 75.888, "characteristics.good": 37944, "characteristics.total": 50000, "ck_system": "NE5260M5_A100-PCIex2_TRT_Triton", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.1, CUDA 11.3", "host_memory_capacity": "1 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 28, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Gold 6258R", "host_processors_per_node": 2, "host_storage_capacity": "2.9 TB * 2", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 2, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/results/NE5260M5_A100-PCIex2_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.1, 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": 2048, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/NE5260M5_A100-PCIex2_TRT_Triton", "system_name": "NE5260M5 (2x A100-PCIe, TensorRT)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1515.15, "task": "image classification", "task2": "image classification", "total_cores": 56, "uid": "ff8874d3dad8216a", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 435654, "90.00 percentile latency (ns)": 440059, "90th percentile latency (ns)": 440059, "95.00 percentile latency (ns)": 441388, "97.00 percentile latency (ns)": 442572, "99.00 percentile latency (ns)": 448945, "99.90 percentile latency (ns)": 486654, "Max latency (ns)": 5620929, "Mean latency (ns)": 436113, "Min duration satisfied": "Yes", "Min latency (ns)": 423104, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 2214.42, "QPS w/o loadgen overhead": 2292.98, "Result is": "VALID", "SUT name": "Triton_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80 GB", "accelerator_memory_configuration": "", "accelerator_model_name": "NVIDIA A100-SXM-80GB", "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": "", "characteristics.90th_percentile_latency_ms": 0.440059, "characteristics.90th_percentile_latency_ns": 440059.0, "characteristics.90th_percentile_latency_s": 0.000440059, "characteristics.90th_percentile_latency_us": 440.059, "characteristics.accuracy": 75.888, "characteristics.good": 37944, "characteristics.total": 50000, "ck_system": "NF5488A5_A100-SXM-80GBx1_TRT_Triton", "ck_used": false, "cooling": "Liquid", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.1, CUDA 11.3", "host_memory_capacity": "2 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 64, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7742", "host_processors_per_node": 2, "host_storage_capacity": "11.7 TB + 448GB*2", "host_storage_type": "NVMe SSD + M.2", "hw_notes": "NVIDIA A100-SXM-80GB-500W", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Inspur/results/NF5488A5_A100-SXM-80GBx1_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.1, CUDA 11.3, cuDNN 8.2.1, Driver 470.57.02, DALI 0.31.0, Triton 21.07", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 2048, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/NF5488A5_A100-SXM-80GBx1_TRT_Triton", "system_name": "NF5488A5 (1x A100-SXM-80GB, TensorRT, Triton)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1515.15, "task": "image classification", "task2": "image classification", "total_cores": 128, "uid": "db994c83e7b73818", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 375968, "90.00 percentile latency (ns)": 380502, "90th percentile latency (ns)": 380502, "95.00 percentile latency (ns)": 381954, "97.00 percentile latency (ns)": 383551, "99.00 percentile latency (ns)": 386264, "99.90 percentile latency (ns)": 397130, "Max latency (ns)": 2837297, "Mean latency (ns)": 375905, "Min duration satisfied": "Yes", "Min latency (ns)": 357017, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 2628.93, "QPS w/o loadgen overhead": 2660.25, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-PCIe", "accelerator_on-chip_memories": "", "accelerators_per_node": 2, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 0.380502, "characteristics.90th_percentile_latency_ns": 380502.0, "characteristics.90th_percentile_latency_s": 0.000380502, "characteristics.90th_percentile_latency_us": 380.502, "characteristics.accuracy": 75.84, "characteristics.good": 37920, "characteristics.total": 50000, "ck_system": "NE5260M5_A100-PCIex2_TRT", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "resnet50", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2, CUDA Version: 11.2", "host_memory_capacity": "1 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 28, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Gold 6258R CPU @ 2.70GHz", "host_processors_per_node": 2, "host_storage_capacity": "5 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.0, "normalize_cores": 2, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/open/Inspur/results/NE5260M5_A100-PCIex2_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.2", "other_software_stack": "TensorRT 7.2, CUDA 11.2, cuDNN 8.0.2, DALI 0.25.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 2048, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 1, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/NE5260M5_A100-PCIex2_TRT", "system_name": "NE5260M5", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1515.15, "task": "image classification", "task2": "image classification", "total_cores": 56, "uid": "b6fcf64d3a8eafe4", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 466391, "90.00 percentile latency (ns)": 471511, "90th percentile latency (ns)": 471511, "95.00 percentile latency (ns)": 473241, "97.00 percentile latency (ns)": 474556, "99.00 percentile latency (ns)": 477054, "99.90 percentile latency (ns)": 490148, "Max latency (ns)": 5089450, "Mean latency (ns)": 466782, "Min duration satisfied": "Yes", "Min latency (ns)": 444855, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 2121.07, "QPS w/o loadgen overhead": 2142.33, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-PCIe", "accelerator_on-chip_memories": "", "accelerators_per_node": 2, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 0.471511, "characteristics.90th_percentile_latency_ns": 471511.0, "characteristics.90th_percentile_latency_s": 0.000471511, "characteristics.90th_percentile_latency_us": 471.511, "characteristics.accuracy": 75.894, "characteristics.good": 37947, "characteristics.total": 50000, "ck_system": "NE5260M5_A100-PCIex2_TRT", "ck_used": false, "cooling": "", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2, CUDA Version: 11.2", "host_memory_capacity": "1 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 28, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Gold 6258R CPU @ 2.70GHz", "host_processors_per_node": 2, "host_storage_capacity": "5 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1024, "mlperf_version": 1.0, "normalize_cores": 2, "normalize_processors": 2, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Inspur/results/NE5260M5_A100-PCIex2_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.2", "other_software_stack": "TensorRT 7.2, CUDA 11.2, cuDNN 8.0.2, DALI 0.25.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 2048, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 1, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/NE5260M5_A100-PCIex2_TRT", "system_name": "NE5260M5", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1515.15, "task": "image classification", "task2": "image classification", "total_cores": 56, "uid": "cda848bd4153a026", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 399579, "90.00 percentile latency (ns)": 408267, "90th percentile latency (ns)": 408267, "95.00 percentile latency (ns)": 411583, "97.00 percentile latency (ns)": 414539, "99.00 percentile latency (ns)": 448645, "99.90 percentile latency (ns)": 1275588, "Max latency (ns)": 28615123, "Mean latency (ns)": 405170, "Min duration satisfied": "Yes", "Min latency (ns)": 362687, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 2442.72, "QPS w/o loadgen overhead": 2468.1, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "1410MHz", "accelerator_host_interconnect": "PCIe Gen4 64 GB/s", "accelerator_interconnect": "NVLink 600 GB/s", "accelerator_interconnect_topology": "Mesh", "accelerator_memory_capacity": "40GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-SXM4", "accelerator_on-chip_memories": "L1: 108x192KB, L2: 40MB per chip", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 0.408267, "characteristics.90th_percentile_latency_ns": 408267.0, "characteristics.90th_percentile_latency_s": 0.000408267, "characteristics.90th_percentile_latency_us": 408.267, "characteristics.accuracy": 75.858, "characteristics.good": 37929, "characteristics.total": 50000, "ck_system": "DGX-A100_A100-SXM4x1_TRT", "ck_used": true, "cooling": "Air-cooled", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "resnet50", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2, CUDA V11.0.221", "host_memory_capacity": "1 TB", "host_memory_configuration": "16 x 64GB DDR4 2933 MT/s", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "L1: 32KB I + 32KB D per core, L2: 512KB per core, L3: 256MB per chip", "host_processor_core_count": 120, "host_processor_frequency": "2.25GHz", "host_processor_interconnect": "4x 16GT/s xGMI", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "11.7TB + 445.7GB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/Inspur/results/DGX-A100_A100-SXM4x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)", "other_software_stack": "TensorRT 7.2, CUDA 11.0.221, cuDNN 8.0.2, DALI 0.25.0; GCC 7.5.0; Python 3.7.10", "performance_issue_same": true, "performance_issue_same_index": 0, "performance_issue_unique": true, "performance_sample_count": 2048, "print_timestamps": true, "problem": false, "qsl_rng_seed": 12786827339337101903, "retraining": "N", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "resnet50_v1.onnx", "status": "on-premise", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/DGX-A100_A100-SXM4x1_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1515.15, "task": "image classification", "task2": "image classification", "total_cores": 240, "uid": "954feebc5775375e", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 657651, "90.00 percentile latency (ns)": 660157, "90th percentile latency (ns)": 660157, "95.00 percentile latency (ns)": 661119, "97.00 percentile latency (ns)": 661899, "99.00 percentile latency (ns)": 663269, "99.90 percentile latency (ns)": 670418, "Max latency (ns)": 2386336, "Mean latency (ns)": 658059, "Min duration satisfied": "Yes", "Min latency (ns)": 641394, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 1508.88, "QPS w/o loadgen overhead": 1519.62, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "1590MHz", "accelerator_host_interconnect": "PCIE", "accelerator_interconnect": "PCIE", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 0.660157, "characteristics.90th_percentile_latency_ns": 660157.0, "characteristics.90th_percentile_latency_s": 0.000660157, "characteristics.90th_percentile_latency_us": 660.157, "characteristics.accuracy": 75.856, "characteristics.good": 37928, "characteristics.total": 50000, "ck_system": "T4x1_TRT", "ck_used": false, "cooling": "Air-cooled", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "open", "formal_model": "resnet50", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2, CUDA V11.0.221", "host_memory_capacity": "1.5T", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "L1d cache: 32K, L1i cache: 32K, L2 cache: 1024K, L3 cache: 36608K", "host_processor_core_count": 24, "host_processor_frequency": "2.90GHz", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8268 CPU @ 2.90GHz", "host_processors_per_node": 2, "host_storage_capacity": "2.9 TB", "host_storage_type": "SSD", "hw_notes": "ECC off", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/open/Inspur/results/T4x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.4", "other_software_stack": "TensorRT 7.2, CUDA V11.0.221, cuDNN 8.0.2, DALI 0.25.0", "performance_issue_same": true, "performance_issue_same_index": 0, "performance_issue_unique": true, "performance_sample_count": 2048, "print_timestamps": true, "problem": false, "qsl_rng_seed": 12786827339337101903, "retraining": "N", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/T4x1_TRT", "system_name": "NF5468M5", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1003.36, "task": "image classification", "task2": "image classification", "total_cores": 48, "uid": "2776874025850b2d", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 872938, "90.00 percentile latency (ns)": 875821, "90th percentile latency (ns)": 875821, "95.00 percentile latency (ns)": 876892, "97.00 percentile latency (ns)": 877692, "99.00 percentile latency (ns)": 879587, "99.90 percentile latency (ns)": 903419, "Max latency (ns)": 7069690, "Mean latency (ns)": 873423, "Min duration satisfied": "Yes", "Min latency (ns)": 865246, "Min queries satisfied": "Yes", "Mode": "Performance", "QPS w/ loadgen overhead": 1136.0, "QPS w/o loadgen overhead": 1144.92, "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "singlestream", "accelerator_frequency": "1590MHz", "accelerator_host_interconnect": "PCIE", "accelerator_interconnect": "PCIE", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "GDDR6", "accelerator_model_name": "NVIDIA T4", "accelerator_on-chip_memories": "", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 0.875821, "characteristics.90th_percentile_latency_ns": 875821.0, "characteristics.90th_percentile_latency_s": 0.000875821, "characteristics.90th_percentile_latency_us": 875.821, "characteristics.accuracy": 75.92, "characteristics.good": 37960, "characteristics.total": 50000, "ck_system": "T4x1_TRT", "ck_used": false, "cooling": "Air-cooled", "dataset": "ImageNet 2012", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/imagenet2012.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 7.2, CUDA V11.0.221", "host_memory_capacity": "1.5T", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "L1d cache: 32K, L1i cache: 32K, L2 cache: 1024K, L3 cache: 36608K", "host_processor_core_count": 24, "host_processor_frequency": "2.90GHz", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8268 CPU @ 2.90GHz", "host_processors_per_node": 2, "host_storage_capacity": "2.9 TB", "host_storage_type": "SSD", "hw_notes": "ECC off", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 60000, "min_query_count": 1024, "mlperf_version": 0.7, "normalize_cores": 1, "normalize_processors": 1, "note_code": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Inspur/code", "note_details": "https://github.com/mlcommons/inference_results_v0.7/tree/master/closed/Inspur/results/T4x1_TRT", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.4", "other_software_stack": "TensorRT 7.2, CUDA V11.0.221, cuDNN 8.0.2, DALI 0.25.0", "performance_issue_same": true, "performance_issue_same_index": 0, "performance_issue_unique": true, "performance_sample_count": 2048, "print_timestamps": true, "problem": false, "qsl_rng_seed": 12786827339337101903, "retraining": "N", "sample_index_rng_seed": 12640797754436136668, "samples_per_query": 1, "schedule_rng_seed": 3135815929913719677, "starting_weights_filename": "resnet50_v1.onnx", "status": "available", "submitter": "Inspur", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Inspur", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/T4x1_TRT", "system_name": "NF5468M5", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 1003.36, "task": "image classification", "task2": "image classification", "total_cores": 48, "uid": "9fe2f144debbdc57", "use_accelerator": true, "weight_data_types": "int8,fp16", "weight_transformations": "quantization, affine fusion" } ]