[ { "50.00 percentile latency (ns)": 10407410, "90.00 percentile latency (ns)": 11859670, "95.00 percentile latency (ns)": 12306543, "97.00 percentile latency (ns)": 12589252, "99.00 percentile latency (ns)": 13110841, "99.90 percentile latency (ns)": 15637069, "Completed samples per second": 91262.68, "Max latency (ns)": 22484065, "Mean latency (ns)": 10460106, "Min duration satisfied": "Yes", "Min latency (ns)": 6774748, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 91264.18, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40 GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-PCIE-40GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 3, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.044, "characteristics.good": 38022, "characteristics.scheduled_queries_per_second": 91264.18, "characteristics.scheduled_queries_per_second.normalized_per_core": 30421.39333333333, "characteristics.scheduled_queries_per_second.normalized_per_processor": 30421.39333333333, "characteristics.total": 50000, "ck_system": "R7525_A100-PCIE-40GBx3_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_memory_capacity": "512 GB", "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": "AMD EPYC 7502", "host_processors_per_node": 2, "host_storage_capacity": "1.8 TB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 3, "normalize_processors": 3, "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/R7525_A100-PCIE-40GBx3_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "CentOS 8.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, 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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R7525_A100-PCIE-40GBx3_TRT", "system_name": "Dell EMC PowerEdge R7525 (3x A100-PCIE-40GB, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 91250, "task": "image classification", "task2": "image classification", "total_cores": 64, "uid": "180389c204439b35", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 11404338, "90.00 percentile latency (ns)": 12493005, "95.00 percentile latency (ns)": 12818205, "97.00 percentile latency (ns)": 13025255, "99.00 percentile latency (ns)": 13397984, "99.90 percentile latency (ns)": 13991008, "Completed samples per second": 135023.37, "Max latency (ns)": 18483852, "Mean latency (ns)": 11419629, "Min duration satisfied": "Yes", "Min latency (ns)": 4845475, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 135024.86, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80 GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-PCIE-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.034, "characteristics.good": 38017, "characteristics.scheduled_queries_per_second": 135024.86, "characteristics.scheduled_queries_per_second.normalized_per_core": 33756.215, "characteristics.scheduled_queries_per_second.normalized_per_processor": 33756.215, "characteristics.total": 50000, "ck_system": "R750xa_A100-PCIE-80GBx4_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_memory_capacity": "1 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": "Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz", "host_processors_per_node": 2, "host_storage_capacity": "3.5 TB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 4, "normalize_processors": 4, "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/R750xa_A100-PCIE-80GBx4_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "CentOS 8.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, 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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R750xa_A100-PCIE-80GBx4_TRT", "system_name": "Dell EMC PowerEdge R750xa (4x A100-PCIE-80GB, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 135000, "task": "image classification", "task2": "image classification", "total_cores": 128, "uid": "0ecf36203249c8a9", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5876134, "90.00 percentile latency (ns)": 8032388, "95.00 percentile latency (ns)": 8471031, "97.00 percentile latency (ns)": 9281925, "99.00 percentile latency (ns)": 10253549, "99.90 percentile latency (ns)": 18699609, "Completed samples per second": 40002.64, "Max latency (ns)": 141590231, "Mean latency (ns)": 5623839, "Min duration satisfied": "Yes", "Min latency (ns)": 971514, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "Triton_MultiMigServer", "Scenario": "server", "Scheduled samples per second": 40002.77, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80 GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-PCIE-80GB (7x1g.10gb MIG)", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.046, "characteristics.good": 38023, "characteristics.scheduled_queries_per_second": 40002.77, "characteristics.scheduled_queries_per_second.normalized_per_core": 10000.6925, "characteristics.scheduled_queries_per_second.normalized_per_processor": 10000.6925, "characteristics.total": 50000, "ck_system": "R750xa_A100-PCIE-80GB-MIG_28x1g.10gb_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_memory_capacity": "1 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": "Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz", "host_processors_per_node": 2, "host_storage_capacity": "3.5 TB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 4, "normalize_processors": 4, "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/R750xa_A100-PCIE-80GB-MIG_28x1g.10gb_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "CentOS 8.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, 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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R750xa_A100-PCIE-80GB-MIG_28x1g.10gb_TRT_Triton", "system_name": "Dell EMC PowerEdge R750xa (4x A100-PCIE-80GB-MIG-7x1g.10gb, TensorRT, Triton)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 40000, "task": "image classification", "task2": "image classification", "total_cores": 128, "uid": "84664a67065d3b3a", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 7974734, "90.00 percentile latency (ns)": 9988809, "95.00 percentile latency (ns)": 10612484, "97.00 percentile latency (ns)": 11175639, "99.00 percentile latency (ns)": 12673657, "99.90 percentile latency (ns)": 14758205, "Completed samples per second": 24094.0, "Max latency (ns)": 22207034, "Mean latency (ns)": 7865105, "Min duration satisfied": "Yes", "Min latency (ns)": 3200868, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 24094.36, "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": 2, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "2.12.3", "characteristics.accuracy": 76.03, "characteristics.good": 38015, "characteristics.power": 521.9498500000003, "characteristics.power.normalized_per_core": 260.97492500000016, "characteristics.power.normalized_per_processor": 260.97492500000016, "characteristics.scheduled_queries_per_second": 24094.36, "characteristics.scheduled_queries_per_second.normalized_per_core": 12047.18, "characteristics.scheduled_queries_per_second.normalized_per_processor": 12047.18, "characteristics.total": 50000, "ck_system": "XE2420_A10x2_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.scheduled_queries_per_second", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "filesystem": "ext3/ext4", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.2, CUDA 11.3", "host_cooling_type": "air", "host_memory_capacity": "384 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 6252 CPU @ 2.10GHz", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "5.00.00.00", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "normalize_cores": 2, "normalize_processors": 2, "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_A10x2_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": 2048, "power_settings": "closed/Dell/power/XE2420_A10x2_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": "resnet50_v1.onnx", "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_A10x2_TRT_MaxQ", "system_name": "Dell EMC PowerEdge XE2420 (2x A10, MaxQ, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 24100, "task": "image classification", "task2": "image classification", "total_cores": 48, "uid": "66735dc9dd49e397", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 11830572, "90.00 percentile latency (ns)": 13081416, "95.00 percentile latency (ns)": 13290682, "97.00 percentile latency (ns)": 13558146, "99.00 percentile latency (ns)": 14122238, "99.90 percentile latency (ns)": 15321267, "Completed samples per second": 272295.49, "Max latency (ns)": 24811709, "Mean latency (ns)": 10544167, "Min duration satisfied": "Yes", "Min latency (ns)": 2873234, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 272300.99, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80 GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-PCIE-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 10, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.038, "characteristics.good": 38019, "characteristics.scheduled_queries_per_second": 272300.99, "characteristics.scheduled_queries_per_second.normalized_per_core": 27230.099, "characteristics.scheduled_queries_per_second.normalized_per_processor": 27230.099, "characteristics.total": 50000, "ck_system": "DSS8440_A100-PCIE-80GBx10_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_memory_capacity": " 768GB", "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 6248R CPU @ 3.00GHz", "host_processors_per_node": 2, "host_storage_capacity": "1.8 TB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 10, "normalize_processors": 10, "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/DSS8440_A100-PCIE-80GBx10_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "CentOS 8.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, 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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/DSS8440_A100-PCIE-80GBx10_TRT", "system_name": "Dell EMC DSS 8440 (10x NVIDIA A100-PCIE-80GB, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 272250, "task": "image classification", "task2": "image classification", "total_cores": 48, "uid": "468f95c14a70b674", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 8242702, "90.00 percentile latency (ns)": 10703906, "95.00 percentile latency (ns)": 11826680, "97.00 percentile latency (ns)": 12495938, "99.00 percentile latency (ns)": 13740896, "99.90 percentile latency (ns)": 16130776, "Completed samples per second": 126016.28, "Max latency (ns)": 27858179, "Mean latency (ns)": 8479062, "Min duration satisfied": "Yes", "Min latency (ns)": 2154989, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "Triton_Server", "Scenario": "server", "Scheduled samples per second": 126017.55, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80 GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-PCIE-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.04, "characteristics.good": 38020, "characteristics.scheduled_queries_per_second": 126017.55, "characteristics.scheduled_queries_per_second.normalized_per_core": 31504.3875, "characteristics.scheduled_queries_per_second.normalized_per_processor": 31504.3875, "characteristics.total": 50000, "ck_system": "R750xa_A100-PCIE-80GBx4_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_memory_capacity": "1 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": "Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz", "host_processors_per_node": 2, "host_storage_capacity": "3.5 TB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 4, "normalize_processors": 4, "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/R750xa_A100-PCIE-80GBx4_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "CentOS 8.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, 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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R750xa_A100-PCIE-80GBx4_TRT_Triton", "system_name": "Dell EMC PowerEdge R750xa (4x A100-PCIE-80GB, TensorRT, Triton)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 126000, "task": "image classification", "task2": "image classification", "total_cores": 128, "uid": "326a4369b5bd684e", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 12185313, "90.00 percentile latency (ns)": 13276899, "95.00 percentile latency (ns)": 13615306, "97.00 percentile latency (ns)": 13851916, "99.00 percentile latency (ns)": 14339490, "99.90 percentile latency (ns)": 15464759, "Completed samples per second": 144018.32, "Max latency (ns)": 23799617, "Mean latency (ns)": 12228490, "Min duration satisfied": "Yes", "Min latency (ns)": 8646057, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 144020.75, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "24 GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A30", "accelerator_on-chip_memories": "", "accelerators_per_node": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.034, "characteristics.good": 38017, "characteristics.scheduled_queries_per_second": 144020.75, "characteristics.scheduled_queries_per_second.normalized_per_core": 18002.59375, "characteristics.scheduled_queries_per_second.normalized_per_processor": 18002.59375, "characteristics.total": 50000, "ck_system": "DSS8440_A30x8_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_memory_capacity": "1 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 48, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Gold 6248R CPU @ 3.00GHz", "host_processors_per_node": 2, "host_storage_capacity": "3 TB", "host_storage_type": "NVMe", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 8, "normalize_processors": 8, "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/DSS8440_A30x8_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "CentOS 8.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, 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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/DSS8440_A30x8_TRT", "system_name": "Dell EMC DSS 8440 (8x A30, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 144000, "task": "image classification", "task2": "image classification", "total_cores": 96, "uid": "45948c59d552b236", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 6600799, "90.00 percentile latency (ns)": 7629310, "95.00 percentile latency (ns)": 7886669, "97.00 percentile latency (ns)": 8053925, "99.00 percentile latency (ns)": 8419231, "99.90 percentile latency (ns)": 8983372, "Completed samples per second": 45005.9, "Max latency (ns)": 15073065, "Mean latency (ns)": 6604434, "Min duration satisfied": "Yes", "Min latency (ns)": 2969878, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 45006.13, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "24 GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A30", "accelerator_on-chip_memories": "", "accelerators_per_node": 3, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.042, "characteristics.good": 38021, "characteristics.scheduled_queries_per_second": 45006.13, "characteristics.scheduled_queries_per_second.normalized_per_core": 15002.043333333333, "characteristics.scheduled_queries_per_second.normalized_per_processor": 15002.043333333333, "characteristics.total": 50000, "ck_system": "R7525_A30x3_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_memory_capacity": "1 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 7763", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 3, "normalize_processors": 3, "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/R7525_A30x3_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "CentOS 8.4.2105", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, 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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R7525_A30x3_TRT", "system_name": "Dell EMC PowerEdge R7525 (3x A30, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 45000, "task": "image classification", "task2": "image classification", "total_cores": 128, "uid": "ebcfd02ea54097a1", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5768459, "90.00 percentile latency (ns)": 8194729, "95.00 percentile latency (ns)": 9318432, "97.00 percentile latency (ns)": 10179662, "99.00 percentile latency (ns)": 12128347, "99.90 percentile latency (ns)": 40801048, "Completed samples per second": 2060.45, "Max latency (ns)": 57590125, "Mean latency (ns)": 6353504, "Min duration satisfied": "Yes", "Min latency (ns)": 4836151, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "SUT\u0000", "Scenario": "server", "Scheduled samples per second": 2060.48, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.292, "characteristics.good": 38146, "characteristics.scheduled_queries_per_second": 2060.48, "characteristics.scheduled_queries_per_second.normalized_per_core": 36.794285714285714, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1030.24, "characteristics.total": 50000, "ck_system": "R750_Intel6330_openvino", "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.scheduled_queries_per_second", "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": "OpenVINO", "host_memory_capacity": "2.16TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 28, "host_processor_frequency": "2.0GHz", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Gold 6330 CPU @ 2.0GHz", "host_processors_per_node": 2, "host_storage_capacity": "1.5TB", "host_storage_type": "NVMe", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "normalize_cores": 56, "normalize_processors": 2, "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/R750_Intel6330_openvino", "number_of_nodes": 1, "operating_system": "Ubuntu 20.10 (GNU/Linux 5.8.0-45-generic x86_64)", "other_software_stack": "MKL-DNN; URL: github.com/intel/mkl-dnn", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "no", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "resnet50_int8.xml", "status": "available", "submitter": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R750_Intel6330_openvino", "system_name": "Dell EMC PowerEdge R750 ICX-6330(2S 28C)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 2060, "task": "image classification", "task2": "image classification", "total_cores": 56, "uid": "5bc2d555c7b2daa8", "use_accelerator": false, "weight_data_types": "int8", "weight_transformations": "TF -> OpenVINO (please refer to closed/Intel/calibration/OpenVINO)" }, { "50.00 percentile latency (ns)": 3707742, "90.00 percentile latency (ns)": 4188070, "95.00 percentile latency (ns)": 4439431, "97.00 percentile latency (ns)": 4653539, "99.00 percentile latency (ns)": 4966591, "99.90 percentile latency (ns)": 8808295, "Completed samples per second": 96014.27, "Max latency (ns)": 45455502, "Mean latency (ns)": 3773699, "Min duration satisfied": "Yes", "Min latency (ns)": 2802849, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 96015.03, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40 GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-PCIE-40GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.042, "characteristics.good": 38021, "characteristics.power": 1446.1594999999995, "characteristics.power.normalized_per_core": 361.5398749999999, "characteristics.power.normalized_per_processor": 361.5398749999999, "characteristics.scheduled_queries_per_second": 96015.03, "characteristics.scheduled_queries_per_second.normalized_per_core": 24003.7575, "characteristics.scheduled_queries_per_second.normalized_per_processor": 24003.7575, "characteristics.total": 50000, "ck_system": "R750xa_A100-PCIE-40GBx4_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_memory_capacity": "256 GB", "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 8368 CPU @ 2.40GHz", "host_processors_per_node": 2, "host_storage_capacity": "543 GB", "host_storage_type": "SSD", "hw_notes": "Result Measured for Power; GPU Power Limited to 175W", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 4, "normalize_processors": 4, "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/R750xa_A100-PCIE-40GBx4_TRT_MaxQ", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "CentOS 8.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, DALI 0.31.0", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 2048, "power_management": "", "power_settings": "closed/Dell/power/R750xa_A100-PCIE-40GBx4_TRT_MaxQ.md", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "1x2400W", "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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R750xa_A100-PCIE-40GBx4_TRT_MaxQ", "system_name": "Dell EMC PowerEdge R750xa (4x A100-PCIE-40GB, MaxQ, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 96000, "task": "image classification", "task2": "image classification", "total_cores": 64, "uid": "18e0c2173143e987", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 8716639, "90.00 percentile latency (ns)": 11240756, "95.00 percentile latency (ns)": 12610036, "97.00 percentile latency (ns)": 13287987, "99.00 percentile latency (ns)": 14492870, "99.90 percentile latency (ns)": 17016237, "Completed samples per second": 25096.14, "Max latency (ns)": 33023507, "Mean latency (ns)": 8722144, "Min duration satisfied": "Yes", "Min latency (ns)": 3210142, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 25096.36, "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": 2, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "2.12.3", "characteristics.accuracy": 76.03, "characteristics.good": 38015, "characteristics.scheduled_queries_per_second": 25096.36, "characteristics.scheduled_queries_per_second.normalized_per_core": 12548.18, "characteristics.scheduled_queries_per_second.normalized_per_processor": 12548.18, "characteristics.total": 50000, "ck_system": "XE2420_A10x2_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.scheduled_queries_per_second", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "filesystem": "ext3/ext4", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.2, CUDA 11.3", "host_cooling_type": "air", "host_memory_capacity": "384 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 6252 CPU @ 2.10GHz", "host_processors_per_node": 2, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "5.00.00.00", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "normalize_cores": 2, "normalize_processors": 2, "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_A10x2_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": 2048, "power_settings": "closed/Dell/power/XE2420_A10x2_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": "resnet50_v1.onnx", "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_A10x2_TRT", "system_name": "Dell EMC PowerEdge XE2420 (2x A10, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 25100, "task": "image classification", "task2": "image classification", "total_cores": 48, "uid": "32ba726ee4bc60b4", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 9355010, "90.00 percentile latency (ns)": 12501545, "95.00 percentile latency (ns)": 13014414, "97.00 percentile latency (ns)": 13330413, "99.00 percentile latency (ns)": 13922616, "99.90 percentile latency (ns)": 16022148, "Completed samples per second": 99015.25, "Max latency (ns)": 31704151, "Mean latency (ns)": 9780329, "Min duration satisfied": "Yes", "Min latency (ns)": 2565092, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "Triton_Server", "Scenario": "server", "Scheduled samples per second": 99015.75, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "24 GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A30", "accelerator_on-chip_memories": "", "accelerators_per_node": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.04, "characteristics.good": 38020, "characteristics.scheduled_queries_per_second": 99015.75, "characteristics.scheduled_queries_per_second.normalized_per_core": 12376.96875, "characteristics.scheduled_queries_per_second.normalized_per_processor": 12376.96875, "characteristics.total": 50000, "ck_system": "DSS8440_A30x8_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_memory_capacity": "1 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 48, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Gold 6248R CPU @ 3.00GHz", "host_processors_per_node": 2, "host_storage_capacity": "3 TB", "host_storage_type": "NVMe", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "network_speed_mbit": "", "nics_enabled_connected": "", "nics_enabled_firmware": "", "nics_enabled_os": "", "normalize_cores": 8, "normalize_processors": 8, "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/DSS8440_A30x8_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "CentOS 8.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.3, cuDNN 8.2.1, Driver 470.42.01, 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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/DSS8440_A30x8_TRT_Triton", "system_name": "Dell EMC DSS 8440 (8x A30, TensorRT, Triton)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 99000, "task": "image classification", "task2": "image classification", "total_cores": 96, "uid": "b47d75bad743a1b2", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 8574554, "90.00 percentile latency (ns)": 11418550, "95.00 percentile latency (ns)": 12664844, "97.00 percentile latency (ns)": 13328393, "99.00 percentile latency (ns)": 14587726, "99.90 percentile latency (ns)": 17150377, "Completed samples per second": 24495.0, "Max latency (ns)": 35553069, "Mean latency (ns)": 8627663, "Min duration satisfied": "Yes", "Min latency (ns)": 2899185, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 24495.2, "accelerator_cooling_type": "", "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": 2, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.accuracy": 76.03, "characteristics.good": 38015, "characteristics.scheduled_queries_per_second": 24495.2, "characteristics.scheduled_queries_per_second.normalized_per_core": 12247.6, "characteristics.scheduled_queries_per_second.normalized_per_processor": 12247.6, "characteristics.total": 50000, "ck_system": "XR12_datacenter_A10x2_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.scheduled_queries_per_second", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "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 8.0.2, CUDA 11.3", "host_cooling_type": "", "host_memory_capacity": "512 GB", "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 6330 CPU @ 2.00GHz", "host_processors_per_node": 1, "host_storage_capacity": "4 TB", "host_storage_type": "NVMe SSD", "hw_notes": "ECC on", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "normalize_cores": 2, "normalize_processors": 2, "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/XR12_datacenter_A10x2_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": 2048, "power_settings": "", "print_timestamps": 0, "problem": false, "psu_details": "", "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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_cooling_type": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/XR12_datacenter_A10x2_TRT", "system_name": "Dell EMC PowerEdge XR12 (2x A10, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 24500, "task": "image classification", "task2": "image classification", "total_cores": 28, "uid": "f94617d0293177d6", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 7803336, "90.00 percentile latency (ns)": 10396271, "95.00 percentile latency (ns)": 10989735, "97.00 percentile latency (ns)": 11449801, "99.00 percentile latency (ns)": 12493557, "99.90 percentile latency (ns)": 14820160, "Completed samples per second": 69942.93, "Max latency (ns)": 69768814, "Mean latency (ns)": 7706863, "Min duration satisfied": "Yes", "Min latency (ns)": 1118716, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "Triton_MultiMigServer", "Scenario": "server", "Scheduled samples per second": 69943.42, "accelerator_cooling_type": "", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-SXM-80GB (7x1g.10gb MIG)", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.04, "characteristics.good": 38020, "characteristics.scheduled_queries_per_second": 69943.42, "characteristics.scheduled_queries_per_second.normalized_per_core": 17485.855, "characteristics.scheduled_queries_per_second.normalized_per_processor": 17485.855, "characteristics.total": 50000, "ck_system": "XE8545_A100-SXM-80GB-MIG_28x1g.10gb_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_cooling_type": "", "host_memory_capacity": "1 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 7763", "host_processors_per_node": 2, "host_storage_capacity": "3 TB", "host_storage_type": "NVMe SSD", "hw_notes": "500W A100-SXM-80GB", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "normalize_cores": 4, "normalize_processors": 4, "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/XE8545_A100-SXM-80GB-MIG_28x1g.10gb_TRT_Triton", "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": 2048, "power_management": "", "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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/XE8545_A100-SXM-80GB-MIG_28x1g.10gb_TRT_Triton", "system_name": "Dell EMC PowerEdge XE8545 (4x A100-SXM-80GB-7x1g.10gb, TensorRT, Triton)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 69937, "task": "image classification", "task2": "image classification", "total_cores": 128, "uid": "48d1841794af47c8", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 5161783, "90.00 percentile latency (ns)": 6136846, "95.00 percentile latency (ns)": 6350393, "97.00 percentile latency (ns)": 6544868, "99.00 percentile latency (ns)": 7078720, "99.90 percentile latency (ns)": 13443947, "Completed samples per second": 109829.23, "Max latency (ns)": 1005396779, "Mean latency (ns)": 5238965, "Min duration satisfied": "Yes", "Min latency (ns)": 3789866, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "Triton_Server", "Scenario": "server", "Scheduled samples per second": 110013.13, "accelerator_cooling_type": "", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-SXM-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.044, "characteristics.good": 38022, "characteristics.scheduled_queries_per_second": 110013.13, "characteristics.scheduled_queries_per_second.normalized_per_core": 27503.2825, "characteristics.scheduled_queries_per_second.normalized_per_processor": 27503.2825, "characteristics.total": 50000, "ck_system": "XE8545_A100-SXM-80GBx4_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.scheduled_queries_per_second", "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 8.0.2, CUDA 11.3", "host_cooling_type": "", "host_memory_capacity": "1 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 7763", "host_processors_per_node": 2, "host_storage_capacity": "3 TB", "host_storage_type": "NVMe SSD", "hw_notes": "500W A100-SXM-80GB", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "normalize_cores": 4, "normalize_processors": 4, "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/XE8545_A100-SXM-80GBx4_TRT_Triton", "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": 2048, "power_management": "", "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": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/XE8545_A100-SXM-80GBx4_TRT_Triton", "system_name": "Dell EMC PowerEdge XE8545 (4x A100-SXM-80GB, TensorRT, Triton)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 110000, "task": "image classification", "task2": "image classification", "total_cores": 128, "uid": "58e6fff034270a08", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 6005031, "90.00 percentile latency (ns)": 7590924, "95.00 percentile latency (ns)": 8062288, "97.00 percentile latency (ns)": 8313542, "99.00 percentile latency (ns)": 8709306, "99.90 percentile latency (ns)": 9323400, "Completed samples per second": 140019.54, "Max latency (ns)": 13028737, "Mean latency (ns)": 6182108, "Min duration satisfied": "Yes", "Min latency (ns)": 4188953, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "LWIS_Server", "Scenario": "server", "Scheduled samples per second": 140021.21, "accelerator_cooling_type": "", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-SXM-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "2.2.5", "characteristics.accuracy": 76.046, "characteristics.good": 38023, "characteristics.power": 2092.9846666666654, "characteristics.power.normalized_per_core": 523.2461666666663, "characteristics.power.normalized_per_processor": 523.2461666666663, "characteristics.scheduled_queries_per_second": 140021.21, "characteristics.scheduled_queries_per_second.normalized_per_core": 35005.3025, "characteristics.scheduled_queries_per_second.normalized_per_processor": 35005.3025, "characteristics.total": 50000, "ck_system": "XE8545_A100-SXM-80GBx4_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.scheduled_queries_per_second", "dim_y_default": "characteristics.accuracy", "dim_y_maximize": true, "division": "closed", "filesystem": "ext3/ext4", "formal_model": "resnet50-v1.5", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "TensorRT 8.0.2, CUDA 11.3", "host_cooling_type": "", "host_memory_capacity": "1 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 7763", "host_processors_per_node": 2, "host_storage_capacity": "3 TB", "host_storage_type": "NVMe SSD", "hw_notes": "500W A100-SXM-80GB", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "management_firmware_version": "4.40.40.151", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "normalize_cores": 4, "normalize_processors": 4, "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/XE8545_A100-SXM-80GBx4_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": 2048, "power_management": "", "power_settings": "closed/Dell/power/XE8545_A100-SXM-80GBx4_power_settings.md", "print_timestamps": 0, "problem": false, "psu_details": "4x2400W", "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": "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/XE8545_A100-SXM-80GBx4_TRT", "system_name": "Dell EMC PowerEdge XE8545 (4x A100-SXM-80GB, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 140000, "task": "image classification", "task2": "image classification", "total_cores": 128, "uid": "fc640b3a8cdddf0e", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 6725465, "90.00 percentile latency (ns)": 9631184, "95.00 percentile latency (ns)": 10975082, "97.00 percentile latency (ns)": 12055005, "99.00 percentile latency (ns)": 14673301, "99.90 percentile latency (ns)": 37784836, "Completed samples per second": 2030.18, "Max latency (ns)": 52019977, "Mean latency (ns)": 7406542, "Min duration satisfied": "Yes", "Min latency (ns)": 5118531, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Performance constraints satisfied": "Yes", "Result is": "VALID", "SUT name": "SUT\u0000", "Scenario": "server", "Scheduled samples per second": 2030.19, "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "N/A", "accelerator_memory_configuration": "", "accelerator_model_name": "N/A", "accelerator_on-chip_memories": "", "accelerators_per_node": 0, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.accuracy": 76.758, "characteristics.good": 786, "characteristics.scheduled_queries_per_second": 2030.19, "characteristics.scheduled_queries_per_second.normalized_per_core": 31.72171875, "characteristics.scheduled_queries_per_second.normalized_per_processor": 1015.095, "characteristics.total": 1024, "ck_system": "R750_Intel8352M_2S_32C", "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.scheduled_queries_per_second", "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": "OpenVINO", "host_memory_capacity": "512GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 32, "host_processor_frequency": "2.30GHz", "host_processor_interconnect": "", "host_processor_model_name": "Intel(R) Xeon(R) Platinum 8352M CPU @ 2.30GHz", "host_processors_per_node": 2, "host_storage_capacity": "2.9TB", "host_storage_type": "NVMe", "hw_notes": "", "informal_model": "resnet50", "input_data_types": "int8", "key.accuracy": "characteristics.accuracy", "max_async_queries": 0, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 270336, "mlperf_version": 1.1, "normalize_cores": 64, "normalize_processors": 2, "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/R750_Intel8352M_2S_32C", "number_of_nodes": 1, "operating_system": " Ubuntu 20.04.2 LTS (GNU/Linux 5.4.0-80-generic x86_64)", "other_software_stack": "MKL-DNN; URL: github.com/intel/mkl-dnn", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "no", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 1, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "resnet50_int8.xml", "status": "available", "submitter": "Dell", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Dell", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/R750_Intel8352M_2S_32C", "system_name": "Dell EMC PowerEdge R750 ICX-8352M(2S 32C)", "system_type": "datacenter", "target_latency (ns)": 15000000, "target_qps": 2030, "task": "image classification", "task2": "image classification", "total_cores": 64, "uid": "543f168c442a8939", "use_accelerator": false, "weight_data_types": "int8", "weight_transformations": "TF -> OpenVINO (please refer to closed/Intel/calibration/OpenVINO)" } ]