[ { "50.00 percentile latency (ns)": 336908875892, "90.00 percentile latency (ns)": 606837242332, "95.00 percentile latency (ns)": 640570053285, "97.00 percentile latency (ns)": 654056992283, "99.00 percentile latency (ns)": 667563523649, "99.90 percentile latency (ns)": 673636415170, "Max latency (ns)": 674300493334, "Mean latency (ns)": 336944384577, "Min duration satisfied": "Yes", "Min latency (ns)": 62768309, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 155.628, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.samples_per_second": 155.628, "characteristics.samples_per_second.normalized_per_core": 51.876, "characteristics.samples_per_second.normalized_per_processor": 51.876, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "R7525_A100-PCIE-40GBx3_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 104940, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 159, "task": "image segmentation", "task2": "image segmentation", "total_cores": 64, "uid": "4a5fbe65297e4849", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 336908875892, "90.00 percentile latency (ns)": 606837242332, "95.00 percentile latency (ns)": 640570053285, "97.00 percentile latency (ns)": 654056992283, "99.00 percentile latency (ns)": 667563523649, "99.90 percentile latency (ns)": 673636415170, "Max latency (ns)": 674300493334, "Mean latency (ns)": 336944384577, "Min duration satisfied": "Yes", "Min latency (ns)": 62768309, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 155.628, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.samples_per_second": 155.628, "characteristics.samples_per_second.normalized_per_core": 51.876, "characteristics.samples_per_second.normalized_per_processor": 51.876, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "R7525_A100-PCIE-40GBx3_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 104940, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 159, "task": "image segmentation", "task2": "image segmentation", "total_cores": 64, "uid": "c69c148cd6e4311b", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 429237297478, "90.00 percentile latency (ns)": 773024674656, "95.00 percentile latency (ns)": 815975975501, "97.00 percentile latency (ns)": 833152486157, "99.00 percentile latency (ns)": 850336799248, "99.90 percentile latency (ns)": 858070036787, "Max latency (ns)": 858923537913, "Mean latency (ns)": 429259151632, "Min duration satisfied": "Yes", "Min latency (ns)": 55245652, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 230.521, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.samples_per_second": 230.521, "characteristics.samples_per_second.normalized_per_core": 57.63025, "characteristics.samples_per_second.normalized_per_processor": 57.63025, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "R750xa_A100-PCIE-80GBx4_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 198000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 300, "task": "image segmentation", "task2": "image segmentation", "total_cores": 128, "uid": "f19844c09a21aa01", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 429237297478, "90.00 percentile latency (ns)": 773024674656, "95.00 percentile latency (ns)": 815975975501, "97.00 percentile latency (ns)": 833152486157, "99.00 percentile latency (ns)": 850336799248, "99.90 percentile latency (ns)": 858070036787, "Max latency (ns)": 858923537913, "Mean latency (ns)": 429259151632, "Min duration satisfied": "Yes", "Min latency (ns)": 55245652, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 230.521, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.samples_per_second": 230.521, "characteristics.samples_per_second.normalized_per_core": 57.63025, "characteristics.samples_per_second.normalized_per_processor": 57.63025, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "R750xa_A100-PCIE-80GBx4_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 198000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 300, "task": "image segmentation", "task2": "image segmentation", "total_cores": 128, "uid": "9b210b45c7b2a644", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 372752957878, "90.00 percentile latency (ns)": 672027733699, "95.00 percentile latency (ns)": 709476263254, "97.00 percentile latency (ns)": 724400576824, "99.00 percentile latency (ns)": 739404944151, "99.90 percentile latency (ns)": 746109575226, "Max latency (ns)": 746833154630, "Mean latency (ns)": 372887414863, "Min duration satisfied": "Yes", "Min latency (ns)": 132991946, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 44.1866, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.854, "characteristics.power": 572.8684471218209, "characteristics.power.normalized_per_core": 286.4342235609104, "characteristics.power.normalized_per_processor": 286.4342235609104, "characteristics.samples_per_second": 44.1866, "characteristics.samples_per_second.normalized_per_core": 22.0933, "characteristics.samples_per_second.normalized_per_processor": 22.0933, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9141, "ck_system": "XE2420_A10x2_TRT_MaxQ", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "filesystem": "ext3/ext4", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "5.00.00.00", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 33000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 50, "task": "image segmentation", "task2": "image segmentation", "total_cores": 48, "uid": "54745f4944a309de", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 372752957878, "90.00 percentile latency (ns)": 672027733699, "95.00 percentile latency (ns)": 709476263254, "97.00 percentile latency (ns)": 724400576824, "99.00 percentile latency (ns)": 739404944151, "99.90 percentile latency (ns)": 746109575226, "Max latency (ns)": 746833154630, "Mean latency (ns)": 372887414863, "Min duration satisfied": "Yes", "Min latency (ns)": 132991946, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 44.1866, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.854, "characteristics.power": 572.8684471218209, "characteristics.power.normalized_per_core": 286.4342235609104, "characteristics.power.normalized_per_processor": 286.4342235609104, "characteristics.samples_per_second": 44.1866, "characteristics.samples_per_second.normalized_per_core": 22.0933, "characteristics.samples_per_second.normalized_per_processor": 22.0933, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9141, "ck_system": "XE2420_A10x2_TRT_MaxQ", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "filesystem": "ext3/ext4", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "5.00.00.00", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 33000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 50, "task": "image segmentation", "task2": "image segmentation", "total_cores": 48, "uid": "6f08ee1cccbcc22c", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 432051066897, "90.00 percentile latency (ns)": 778042426981, "95.00 percentile latency (ns)": 821304218283, "97.00 percentile latency (ns)": 838605494846, "99.00 percentile latency (ns)": 855905414429, "99.90 percentile latency (ns)": 863684896101, "Max latency (ns)": 864547616422, "Mean latency (ns)": 432066181760, "Min duration satisfied": "Yes", "Min latency (ns)": 58991325, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "Triton_Server", "Samples per second": 229.022, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.samples_per_second": 229.022, "characteristics.samples_per_second.normalized_per_core": 57.2555, "characteristics.samples_per_second.normalized_per_processor": 57.2555, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "R750xa_A100-PCIE-80GBx4_TRT_Triton", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 198000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 300, "task": "image segmentation", "task2": "image segmentation", "total_cores": 128, "uid": "1adc5184964a2d4c", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 432051066897, "90.00 percentile latency (ns)": 778042426981, "95.00 percentile latency (ns)": 821304218283, "97.00 percentile latency (ns)": 838605494846, "99.00 percentile latency (ns)": 855905414429, "99.90 percentile latency (ns)": 863684896101, "Max latency (ns)": 864547616422, "Mean latency (ns)": 432066181760, "Min duration satisfied": "Yes", "Min latency (ns)": 58991325, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "Triton_Server", "Samples per second": 229.022, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.samples_per_second": 229.022, "characteristics.samples_per_second.normalized_per_core": 57.2555, "characteristics.samples_per_second.normalized_per_processor": 57.2555, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "R750xa_A100-PCIE-80GBx4_TRT_Triton", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 198000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 300, "task": "image segmentation", "task2": "image segmentation", "total_cores": 128, "uid": "4c8e44ee9b99e596", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 587457246943, "90.00 percentile latency (ns)": 1057279036185, "95.00 percentile latency (ns)": 1116013463362, "97.00 percentile latency (ns)": 1139464372627, "99.00 percentile latency (ns)": 1162997064771, "99.90 percentile latency (ns)": 1173558517460, "Max latency (ns)": 1174682004434, "Mean latency (ns)": 587439993869, "Min duration satisfied": "Yes", "Min latency (ns)": 167184406, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 168.556, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85407, "characteristics.samples_per_second": 168.556, "characteristics.samples_per_second.normalized_per_core": 21.0695, "characteristics.samples_per_second.normalized_per_processor": 21.0695, "characteristics.tumor core": 0.8691, "characteristics.whole tumor": 0.9142, "ck_system": "DSS8440_A30x8_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 198000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 300, "task": "image segmentation", "task2": "image segmentation", "total_cores": 96, "uid": "4c80c20e8b23a899", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 587457246943, "90.00 percentile latency (ns)": 1057279036185, "95.00 percentile latency (ns)": 1116013463362, "97.00 percentile latency (ns)": 1139464372627, "99.00 percentile latency (ns)": 1162997064771, "99.90 percentile latency (ns)": 1173558517460, "Max latency (ns)": 1174682004434, "Mean latency (ns)": 587439993869, "Min duration satisfied": "Yes", "Min latency (ns)": 167184406, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 168.556, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85407, "characteristics.samples_per_second": 168.556, "characteristics.samples_per_second.normalized_per_core": 21.0695, "characteristics.samples_per_second.normalized_per_processor": 21.0695, "characteristics.tumor core": 0.8691, "characteristics.whole tumor": 0.9142, "ck_system": "DSS8440_A30x8_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 198000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 300, "task": "image segmentation", "task2": "image segmentation", "total_cores": 96, "uid": "41f3b604413ba6d3", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 340805264135, "90.00 percentile latency (ns)": 614470597014, "95.00 percentile latency (ns)": 648671848818, "97.00 percentile latency (ns)": 662356903676, "99.00 percentile latency (ns)": 676041338869, "99.90 percentile latency (ns)": 682207168090, "Max latency (ns)": 682849643260, "Mean latency (ns)": 340945038932, "Min duration satisfied": "Yes", "Min latency (ns)": 97505394, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 89.1338, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85407, "characteristics.samples_per_second": 89.1338, "characteristics.samples_per_second.normalized_per_core": 29.711266666666663, "characteristics.samples_per_second.normalized_per_processor": 29.711266666666663, "characteristics.tumor core": 0.8691, "characteristics.whole tumor": 0.9142, "ck_system": "R7525_A30x3_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 60865, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 92.22, "task": "image segmentation", "task2": "image segmentation", "total_cores": 128, "uid": "d3ab348e644ba958", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 340805264135, "90.00 percentile latency (ns)": 614470597014, "95.00 percentile latency (ns)": 648671848818, "97.00 percentile latency (ns)": 662356903676, "99.00 percentile latency (ns)": 676041338869, "99.90 percentile latency (ns)": 682207168090, "Max latency (ns)": 682849643260, "Mean latency (ns)": 340945038932, "Min duration satisfied": "Yes", "Min latency (ns)": 97505394, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 89.1338, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85407, "characteristics.samples_per_second": 89.1338, "characteristics.samples_per_second.normalized_per_core": 29.711266666666663, "characteristics.samples_per_second.normalized_per_processor": 29.711266666666663, "characteristics.tumor core": 0.8691, "characteristics.whole tumor": 0.9142, "ck_system": "R7525_A30x3_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 60865, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 92.22, "task": "image segmentation", "task2": "image segmentation", "total_cores": 128, "uid": "61682bb25f13fedc", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 2639427896544, "90.00 percentile latency (ns)": 4748368384588, "95.00 percentile latency (ns)": 5012133150766, "97.00 percentile latency (ns)": 5118004331446, "99.00 percentile latency (ns)": 5224181569511, "99.90 percentile latency (ns)": 5271469163076, "Max latency (ns)": 5276379792515, "Mean latency (ns)": 2639036975261, "Min duration satisfied": "Yes", "Min latency (ns)": 1922330229, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "OV_SUT", "Samples per second": 4.65774, "Scenario": "offline", "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.enhancing tumor": 0.7872, "characteristics.mean": 0.85603, "characteristics.samples_per_second": 4.65774, "characteristics.samples_per_second.normalized_per_core": 0.08317392857142858, "characteristics.samples_per_second.normalized_per_processor": 2.32887, "characteristics.tumor core": 0.8679, "characteristics.whole tumor": 0.913, "ck_system": "R750_Intel6330_openvino", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "No change.", "key.accuracy": "characteristics.mean", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "We do not retrain the model weight.", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 24576, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/3928973/files/224_224_160.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/R750_Intel6330_openvino", "system_name": "Dell EMC PowerEdge R750 ICX-6330(2S 28C)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 10, "task": "image segmentation", "task2": "image segmentation", "total_cores": 56, "uid": "4fafb92881ee3f28", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "We transfer the weight from int8 datatype in ONNX file to int8 datatype in OpenVino IR file." }, { "50.00 percentile latency (ns)": 2639427896544, "90.00 percentile latency (ns)": 4748368384588, "95.00 percentile latency (ns)": 5012133150766, "97.00 percentile latency (ns)": 5118004331446, "99.00 percentile latency (ns)": 5224181569511, "99.90 percentile latency (ns)": 5271469163076, "Max latency (ns)": 5276379792515, "Mean latency (ns)": 2639036975261, "Min duration satisfied": "Yes", "Min latency (ns)": 1922330229, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "OV_SUT", "Samples per second": 4.65774, "Scenario": "offline", "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.enhancing tumor": 0.7872, "characteristics.mean": 0.85603, "characteristics.samples_per_second": 4.65774, "characteristics.samples_per_second.normalized_per_core": 0.08317392857142858, "characteristics.samples_per_second.normalized_per_processor": 2.32887, "characteristics.tumor core": 0.8679, "characteristics.whole tumor": 0.913, "ck_system": "R750_Intel6330_openvino", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "No change.", "key.accuracy": "characteristics.mean", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "We do not retrain the model weight.", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 24576, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/3928973/files/224_224_160.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/R750_Intel6330_openvino", "system_name": "Dell EMC PowerEdge R750 ICX-6330(2S 28C)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 10, "task": "image segmentation", "task2": "image segmentation", "total_cores": 56, "uid": "ae41a94cf4904914", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "We transfer the weight from int8 datatype in ONNX file to int8 datatype in OpenVino IR file." }, { "50.00 percentile latency (ns)": 391271278188, "90.00 percentile latency (ns)": 704703491318, "95.00 percentile latency (ns)": 743968984785, "97.00 percentile latency (ns)": 759636957619, "99.00 percentile latency (ns)": 775328304885, "99.90 percentile latency (ns)": 782378170961, "Max latency (ns)": 783158685669, "Mean latency (ns)": 391300345821, "Min duration satisfied": "Yes", "Min latency (ns)": 63176694, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 178.661, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.power": 1478.1427841634738, "characteristics.power.normalized_per_core": 369.53569604086846, "characteristics.power.normalized_per_processor": 369.53569604086846, "characteristics.samples_per_second": 178.661, "characteristics.samples_per_second.normalized_per_core": 44.66525, "characteristics.samples_per_second.normalized_per_processor": 44.66525, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "R750xa_A100-PCIE-40GBx4_TRT_MaxQ", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 139920, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 212, "task": "image segmentation", "task2": "image segmentation", "total_cores": 64, "uid": "04224ac1ae565886", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 391271278188, "90.00 percentile latency (ns)": 704703491318, "95.00 percentile latency (ns)": 743968984785, "97.00 percentile latency (ns)": 759636957619, "99.00 percentile latency (ns)": 775328304885, "99.90 percentile latency (ns)": 782378170961, "Max latency (ns)": 783158685669, "Mean latency (ns)": 391300345821, "Min duration satisfied": "Yes", "Min latency (ns)": 63176694, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 178.661, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.power": 1478.1427841634738, "characteristics.power.normalized_per_core": 369.53569604086846, "characteristics.power.normalized_per_processor": 369.53569604086846, "characteristics.samples_per_second": 178.661, "characteristics.samples_per_second.normalized_per_core": 44.66525, "characteristics.samples_per_second.normalized_per_processor": 44.66525, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "R750xa_A100-PCIE-40GBx4_TRT_MaxQ", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 139920, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 212, "task": "image segmentation", "task2": "image segmentation", "total_cores": 64, "uid": "9c4b0782f2227758", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 366562168793, "90.00 percentile latency (ns)": 660447658636, "95.00 percentile latency (ns)": 697217097105, "97.00 percentile latency (ns)": 711884431771, "99.00 percentile latency (ns)": 726593131577, "99.90 percentile latency (ns)": 733202478727, "Max latency (ns)": 733916484764, "Mean latency (ns)": 366588764492, "Min duration satisfied": "Yes", "Min latency (ns)": 131710366, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 44.9642, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.854, "characteristics.samples_per_second": 44.9642, "characteristics.samples_per_second.normalized_per_core": 22.4821, "characteristics.samples_per_second.normalized_per_processor": 22.4821, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9141, "ck_system": "XE2420_A10x2_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "filesystem": "ext3/ext4", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "5.00.00.00", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 33000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 50, "task": "image segmentation", "task2": "image segmentation", "total_cores": 48, "uid": "3de934ec4ad44670", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 366562168793, "90.00 percentile latency (ns)": 660447658636, "95.00 percentile latency (ns)": 697217097105, "97.00 percentile latency (ns)": 711884431771, "99.00 percentile latency (ns)": 726593131577, "99.90 percentile latency (ns)": 733202478727, "Max latency (ns)": 733916484764, "Mean latency (ns)": 366588764492, "Min duration satisfied": "Yes", "Min latency (ns)": 131710366, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 44.9642, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.854, "characteristics.samples_per_second": 44.9642, "characteristics.samples_per_second.normalized_per_core": 22.4821, "characteristics.samples_per_second.normalized_per_processor": 22.4821, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9141, "ck_system": "XE2420_A10x2_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "filesystem": "ext3/ext4", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "5.00.00.00", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "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": 33000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 50, "task": "image segmentation", "task2": "image segmentation", "total_cores": 48, "uid": "811e5616396a4f9d", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 375524168356, "90.00 percentile latency (ns)": 676893296917, "95.00 percentile latency (ns)": 714532217065, "97.00 percentile latency (ns)": 729600847079, "99.00 percentile latency (ns)": 744673452051, "99.90 percentile latency (ns)": 751433281638, "Max latency (ns)": 752164220800, "Mean latency (ns)": 375590889645, "Min duration satisfied": "Yes", "Min latency (ns)": 107402652, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 43.8734, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.854, "characteristics.samples_per_second": 43.8734, "characteristics.samples_per_second.normalized_per_core": 21.9367, "characteristics.samples_per_second.normalized_per_processor": 21.9367, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9141, "ck_system": "XR12_datacenter_A10x2_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "power_settings": "", "print_timestamps": 0, "problem": false, "psu_details": "", "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 33000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 50, "task": "image segmentation", "task2": "image segmentation", "total_cores": 28, "uid": "65aae7a35e24fd65", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 375524168356, "90.00 percentile latency (ns)": 676893296917, "95.00 percentile latency (ns)": 714532217065, "97.00 percentile latency (ns)": 729600847079, "99.00 percentile latency (ns)": 744673452051, "99.90 percentile latency (ns)": 751433281638, "Max latency (ns)": 752164220800, "Mean latency (ns)": 375590889645, "Min duration satisfied": "Yes", "Min latency (ns)": 107402652, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "LWIS_Server", "Samples per second": 43.8734, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.854, "characteristics.samples_per_second": 43.8734, "characteristics.samples_per_second.normalized_per_core": 21.9367, "characteristics.samples_per_second.normalized_per_processor": 21.9367, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9141, "ck_system": "XR12_datacenter_A10x2_TRT", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "filesystem": "", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "management_firmware_version": "", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "power_settings": "", "print_timestamps": 0, "problem": false, "psu_details": "", "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 33000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 50, "task": "image segmentation", "task2": "image segmentation", "total_cores": 28, "uid": "3a6649f49994b348", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 352130695172, "90.00 percentile latency (ns)": 633767932603, "95.00 percentile latency (ns)": 668963504124, "97.00 percentile latency (ns)": 683037587350, "99.00 percentile latency (ns)": 697126279040, "99.90 percentile latency (ns)": 703466463781, "Max latency (ns)": 704141257564, "Mean latency (ns)": 352126515310, "Min duration satisfied": "Yes", "Min latency (ns)": 93478593, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "Triton_Server", "Samples per second": 171.528, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.samples_per_second": 171.528, "characteristics.samples_per_second.normalized_per_core": 42.882, "characteristics.samples_per_second.normalized_per_processor": 42.882, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "XE8545_A100-SXM-80GBx4_TRT_Triton", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "power_management": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 120780, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 183, "task": "image segmentation", "task2": "image segmentation", "total_cores": 128, "uid": "a5ca59e25b43ab84", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 352130695172, "90.00 percentile latency (ns)": 633767932603, "95.00 percentile latency (ns)": 668963504124, "97.00 percentile latency (ns)": 683037587350, "99.00 percentile latency (ns)": 697126279040, "99.90 percentile latency (ns)": 703466463781, "Max latency (ns)": 704141257564, "Mean latency (ns)": 352126515310, "Min duration satisfied": "Yes", "Min latency (ns)": 93478593, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "Triton_Server", "Samples per second": 171.528, "Scenario": "offline", "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.enhancing tumor": 0.7789, "characteristics.mean": 0.85403, "characteristics.samples_per_second": 171.528, "characteristics.samples_per_second.normalized_per_core": 42.882, "characteristics.samples_per_second.normalized_per_processor": 42.882, "characteristics.tumor core": 0.869, "characteristics.whole tumor": 0.9142, "ck_system": "XE8545_A100-SXM-80GBx4_TRT_Triton", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "int8", "key.accuracy": "characteristics.mean", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "power_management": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "N", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 120780, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "224_224_160_dyanmic_bs.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)": 0, "target_qps": 183, "task": "image segmentation", "task2": "image segmentation", "total_cores": 128, "uid": "4b4dae2021613f06", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 2588941243548, "90.00 percentile latency (ns)": 4658701553783, "95.00 percentile latency (ns)": 4918947771948, "97.00 percentile latency (ns)": 5022506037252, "99.00 percentile latency (ns)": 5125460611619, "99.90 percentile latency (ns)": 5172040909389, "Max latency (ns)": 5176827620227, "Mean latency (ns)": 2588585930402, "Min duration satisfied": "Yes", "Min latency (ns)": 1879989102, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "OV_SUT", "Samples per second": 4.74731, "Scenario": "offline", "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.enhancing tumor": 0.7872, "characteristics.mean": 0.85603, "characteristics.samples_per_second": 4.74731, "characteristics.samples_per_second.normalized_per_core": 0.07417671875, "characteristics.samples_per_second.normalized_per_processor": 2.373655, "characteristics.tumor core": 0.8679, "characteristics.whole tumor": 0.913, "ck_system": "R750_Intel8352M_2S_32C", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "formal_model": "3d-unet", "formal_model_accuracy": 99.0, "formal_model_link": "", "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": "3d-unet-99", "input_data_types": "No change.", "key.accuracy": "characteristics.mean", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "We do not retrain the model weight.", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 24576, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/3928973/files/224_224_160.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/R750_Intel8352M_2S_32C", "system_name": "Dell EMC PowerEdge R750 ICX-8352M(2S 32C)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 10, "task": "image segmentation", "task2": "image segmentation", "total_cores": 64, "uid": "e5caad371a40949d", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "We transfer the weight from int8 datatype in ONNX file to int8 datatype in OpenVino IR file." }, { "50.00 percentile latency (ns)": 2588941243548, "90.00 percentile latency (ns)": 4658701553783, "95.00 percentile latency (ns)": 4918947771948, "97.00 percentile latency (ns)": 5022506037252, "99.00 percentile latency (ns)": 5125460611619, "99.90 percentile latency (ns)": 5172040909389, "Max latency (ns)": 5176827620227, "Mean latency (ns)": 2588585930402, "Min duration satisfied": "Yes", "Min latency (ns)": 1879989102, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "OV_SUT", "Samples per second": 4.74731, "Scenario": "offline", "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.enhancing tumor": 0.7872, "characteristics.mean": 0.85603, "characteristics.samples_per_second": 4.74731, "characteristics.samples_per_second.normalized_per_core": 0.07417671875, "characteristics.samples_per_second.normalized_per_processor": 2.373655, "characteristics.tumor core": 0.8679, "characteristics.whole tumor": 0.913, "ck_system": "R750_Intel8352M_2S_32C", "ck_used": false, "cooling": "", "dataset": "BraTS 2019", "dataset_link": "https://www.med.upenn.edu/cbica/brats2019/data.html", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mean", "dim_y_maximize": true, "division": "closed", "formal_model": "3d-unet", "formal_model_accuracy": 99.9, "formal_model_link": "", "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": "3d-unet-99.9", "input_data_types": "No change.", "key.accuracy": "characteristics.mean", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "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": 16, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "We do not retrain the model weight.", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 24576, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "The original weight filename: https://zenodo.org/record/3928973/files/224_224_160.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/R750_Intel8352M_2S_32C", "system_name": "Dell EMC PowerEdge R750 ICX-8352M(2S 32C)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 10, "task": "image segmentation", "task2": "image segmentation", "total_cores": 64, "uid": "54823da2490e8ac9", "use_accelerator": false, "weight_data_types": "No change.", "weight_transformations": "We transfer the weight from int8 datatype in ONNX file to int8 datatype in OpenVino IR file." } ]