[ { "50.00 percentile latency (ns)": 673129179919, "90.00 percentile latency (ns)": 1181738079315, "95.00 percentile latency (ns)": 1245283343210, "97.00 percentile latency (ns)": 1270796194545, "99.00 percentile latency (ns)": 1296264144364, "99.90 percentile latency (ns)": 1307709825368, "Max latency (ns)": 1308978233066, "Mean latency (ns)": 9324384751, "Min duration satisfied": "Yes", "Min latency (ns)": 38123305684, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 1210100.0, "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-SXM-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.samples_per_second": 1210100.0, "characteristics.samples_per_second.normalized_per_core": 302525.0, "characteristics.samples_per_second.normalized_per_processor": 302525.0, "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT", "ck_used": false, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "formal_model_accuracy": 99.9, "formal_model_link": "", "framework": "TensorRT 8.0.2, CUDA 11.4", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 64, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7742", "host_processors_per_node": 1, "host_storage_capacity": "5 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99.9", "input_data_types": "int8", "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/HPE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.4, 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": 204800, "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": 1584000000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "HPE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT", "system_name": "HPE ProLiant XL645d Gen10 Plus (4x A100-SXM-80GB, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2400000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 64, "uid": "3a6b5fe107f04acc", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 673129179919, "90.00 percentile latency (ns)": 1181738079315, "95.00 percentile latency (ns)": 1245283343210, "97.00 percentile latency (ns)": 1270796194545, "99.00 percentile latency (ns)": 1296264144364, "99.90 percentile latency (ns)": 1307709825368, "Max latency (ns)": 1308978233066, "Mean latency (ns)": 9324384751, "Min duration satisfied": "Yes", "Min latency (ns)": 38123305684, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 1210100.0, "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-SXM-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.samples_per_second": 1210100.0, "characteristics.samples_per_second.normalized_per_core": 302525.0, "characteristics.samples_per_second.normalized_per_processor": 302525.0, "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT", "ck_used": false, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "TensorRT 8.0.2, CUDA 11.4", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 64, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7742", "host_processors_per_node": 1, "host_storage_capacity": "5 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99", "input_data_types": "int8", "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/HPE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.2", "other_hardware": "", "other_software_stack": "TensorRT 8.0.2, CUDA 11.4, 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": 204800, "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": 1584000000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "HPE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT", "system_name": "HPE ProLiant XL645d Gen10 Plus (4x A100-SXM-80GB, TensorRT)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2400000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 64, "uid": "d8b3c4e1c87bdc76", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 600476175278, "90.00 percentile latency (ns)": 1083054576217, "95.00 percentile latency (ns)": 1142812327456, "97.00 percentile latency (ns)": 1166712015914, "99.00 percentile latency (ns)": 1190601739005, "99.90 percentile latency (ns)": 1201351311135, "Max latency (ns)": 1202545691063, "Mean latency (ns)": 3634885481, "Min duration satisfied": "Yes", "Min latency (ns)": 10597235, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "Triton_Server", "Samples per second": 1207440.0, "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-SXM-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.samples_per_second": 1207440.0, "characteristics.samples_per_second.normalized_per_core": 301860.0, "characteristics.samples_per_second.normalized_per_processor": 301860.0, "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT_Triton", "ck_used": false, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "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": 64, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7742", "host_processors_per_node": 1, "host_storage_capacity": "15 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99.9", "input_data_types": "int8", "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/HPE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.4", "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": 204800, "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": 1452000000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "HPE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT_Triton", "system_name": "HPE ProLiant XL645d Gen10 Plus (4x A100-SXM-80GB, TensorRT, Triton)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2200000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 64, "uid": "5d43b3aa5b07c39b", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 600476175278, "90.00 percentile latency (ns)": 1083054576217, "95.00 percentile latency (ns)": 1142812327456, "97.00 percentile latency (ns)": 1166712015914, "99.00 percentile latency (ns)": 1190601739005, "99.90 percentile latency (ns)": 1201351311135, "Max latency (ns)": 1202545691063, "Mean latency (ns)": 3634885481, "Min duration satisfied": "Yes", "Min latency (ns)": 10597235, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "Triton_Server", "Samples per second": 1207440.0, "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-SXM-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 4, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.samples_per_second": 1207440.0, "characteristics.samples_per_second.normalized_per_core": 301860.0, "characteristics.samples_per_second.normalized_per_processor": 301860.0, "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT_Triton", "ck_used": false, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "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": 64, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7742", "host_processors_per_node": 1, "host_storage_capacity": "15 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99", "input_data_types": "int8", "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/HPE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.4", "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": 204800, "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": 1452000000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "HPE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM-80GBx4_TRT_Triton", "system_name": "HPE ProLiant XL645d Gen10 Plus (4x A100-SXM-80GB, TensorRT, Triton)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2200000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 64, "uid": "843f0f5442995046", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 354212648512, "90.00 percentile latency (ns)": 638671756128, "95.00 percentile latency (ns)": 674371668429, "97.00 percentile latency (ns)": 688652266038, "99.00 percentile latency (ns)": 702913340221, "99.90 percentile latency (ns)": 709359769869, "Max latency (ns)": 710092579082, "Mean latency (ns)": 16214850213, "Min duration satisfied": "Yes", "Min latency (ns)": 4111870, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "Triton_Server", "Samples per second": 1152530.0, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40 GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-SXM4-40GB", "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.samples_per_second": 1152530.0, "characteristics.samples_per_second.normalized_per_core": 288132.5, "characteristics.samples_per_second.normalized_per_processor": 288132.5, "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT_Triton", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "formal_model_accuracy": 99.9, "formal_model_link": "", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "3.2 TB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "dlrm-99.9", "input_data_types": "int8", "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": 1.0, "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_v1.0/tree/master/closed/HPE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)", "other_hardware": "", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.27.04, DALI 0.30.0, Triton 21.02; GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 204800, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 818400000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "HPE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE", "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT_Triton", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 1240000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 120, "uid": "6f96544b3b865784", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 354212648512, "90.00 percentile latency (ns)": 638671756128, "95.00 percentile latency (ns)": 674371668429, "97.00 percentile latency (ns)": 688652266038, "99.00 percentile latency (ns)": 702913340221, "99.90 percentile latency (ns)": 709359769869, "Max latency (ns)": 710092579082, "Mean latency (ns)": 16214850213, "Min duration satisfied": "Yes", "Min latency (ns)": 4111870, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "Triton_Server", "Samples per second": 1152530.0, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40 GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-SXM4-40GB", "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.samples_per_second": 1152530.0, "characteristics.samples_per_second.normalized_per_core": 288132.5, "characteristics.samples_per_second.normalized_per_processor": 288132.5, "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT_Triton", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "3.2 TB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "dlrm-99", "input_data_types": "int8", "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": 1.0, "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_v1.0/tree/master/closed/HPE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT_Triton", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)", "other_hardware": "", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.27.04, DALI 0.30.0, Triton 21.02; GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 204800, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 818400000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "HPE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE", "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT_Triton", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 1240000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 120, "uid": "8bda92a904d5b9ce", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 337162763099, "90.00 percentile latency (ns)": 598853768075, "95.00 percentile latency (ns)": 631584327230, "97.00 percentile latency (ns)": 644677772771, "99.00 percentile latency (ns)": 657780079539, "99.90 percentile latency (ns)": 663672214753, "Max latency (ns)": 664325766887, "Mean latency (ns)": 21790629622, "Min duration satisfied": "Yes", "Min latency (ns)": 11515394441, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 1231930.0, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40 GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-SXM4-40GB", "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.samples_per_second": 1231930.0, "characteristics.samples_per_second.normalized_per_core": 307982.5, "characteristics.samples_per_second.normalized_per_processor": 307982.5, "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "formal_model_accuracy": 99.9, "formal_model_link": "", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "3.2 TB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "dlrm-99.9", "input_data_types": "int8", "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": 1.0, "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_v1.0/tree/master/closed/HPE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)", "other_hardware": "", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.27.04, DALI 0.30.0; GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 204800, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 818400000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "HPE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE", "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 1240000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 120, "uid": "5e038994b7ca1935", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 337162763099, "90.00 percentile latency (ns)": 598853768075, "95.00 percentile latency (ns)": 631584327230, "97.00 percentile latency (ns)": 644677772771, "99.00 percentile latency (ns)": 657780079539, "99.90 percentile latency (ns)": 663672214753, "Max latency (ns)": 664325766887, "Mean latency (ns)": 21790629622, "Min duration satisfied": "Yes", "Min latency (ns)": 11515394441, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 1231930.0, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "40 GB", "accelerator_memory_configuration": "HBM2e", "accelerator_model_name": "NVIDIA A100-SXM4-40GB", "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.samples_per_second": 1231930.0, "characteristics.samples_per_second.normalized_per_core": 307982.5, "characteristics.samples_per_second.normalized_per_processor": 307982.5, "ck_system": "HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT", "ck_used": true, "cooling": "", "dataset": "1TB Click Logs", "dataset_link": "", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.AUC", "dim_y_maximize": true, "disk_controllers": "", "disk_drives": "", "division": "closed", "filesystem": "", "formal_model": "dlrm", "formal_model_accuracy": 99.0, "formal_model_link": "", "framework": "TensorRT 7.2.3, CUDA 11.1", "host_memory_capacity": "512 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 120, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 1, "host_storage_capacity": "3.2 TB", "host_storage_type": "SSD", "hw_notes": "", "informal_model": "dlrm-99", "input_data_types": "int8", "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": 1.0, "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_v1.0/tree/master/closed/HPE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/HPE/results/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)", "other_hardware": "", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, Driver 460.27.04, DALI 0.30.0; GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 204800, "power_management": "", "power_supply_details": "", "power_supply_quantity_and_rating_watts": "", "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 818400000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "HPE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/HPE", "sw_notes": "Powered by CK v2.5.8 (https://github.com/ctuning/ck)", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/HPE-ProLiant-XL645d-Gen10-Plus_A100-SXM4-40GBx4_TRT", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 1240000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 120, "uid": "e4697810e9f609f2", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" } ]