[ { "50.00 percentile latency (ns)": 334196547327, "90.00 percentile latency (ns)": 584323952611, "95.00 percentile latency (ns)": 615605941541, "97.00 percentile latency (ns)": 628115142337, "99.00 percentile latency (ns)": 640625025605, "99.90 percentile latency (ns)": 646253278632, "Max latency (ns)": 646878936231, "Mean latency (ns)": 8182871281, "Min duration satisfied": "Yes", "Min latency (ns)": 22116905320, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 2448680.0, "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": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.samples_per_second": 2448680.0, "characteristics.samples_per_second.normalized_per_core": 306085.0, "characteristics.samples_per_second.normalized_per_processor": 306085.0, "ck_system": "GIGABYTE-G492-ID0_A100-SXM-80GBx8_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, CUDA 11.4", "host_memory_capacity": "1 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 32, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel Xeon Platinum 8362", "host_processors_per_node": 2, "host_storage_capacity": "3.84 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": 8, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/GIGABYTE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/GIGABYTE/results/GIGABYTE-G492-ID0_A100-SXM-80GBx8_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.2 LTS", "other_hardware": "", "other_software_stack": "TensorRT 8.0, CUDA 11.4, cuDNN 8.2.1, Driver 470.57.02, DALI 0.30.0", "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": "preview", "submitter": "GIGABYTE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/GIGABYTE", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/GIGABYTE-G492-ID0_A100-SXM-80GBx8_TRT", "system_name": "GIGABYTE-G492-ID0_A100-SXM-80GBx8", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2400000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 64, "uid": "9f6b8147dcbc81f4", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 334196547327, "90.00 percentile latency (ns)": 584323952611, "95.00 percentile latency (ns)": 615605941541, "97.00 percentile latency (ns)": 628115142337, "99.00 percentile latency (ns)": 640625025605, "99.90 percentile latency (ns)": 646253278632, "Max latency (ns)": 646878936231, "Mean latency (ns)": 8182871281, "Min duration satisfied": "Yes", "Min latency (ns)": 22116905320, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 2448680.0, "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": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "boot_firmware_version": "", "characteristics.samples_per_second": 2448680.0, "characteristics.samples_per_second.normalized_per_core": 306085.0, "characteristics.samples_per_second.normalized_per_processor": 306085.0, "ck_system": "GIGABYTE-G492-ID0_A100-SXM-80GBx8_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, CUDA 11.4", "host_memory_capacity": "1 TB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "", "host_processor_core_count": 32, "host_processor_frequency": "", "host_processor_interconnect": "", "host_processor_model_name": "Intel Xeon Platinum 8362", "host_processors_per_node": 2, "host_storage_capacity": "3.84 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": 8, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/GIGABYTE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/GIGABYTE/results/GIGABYTE-G492-ID0_A100-SXM-80GBx8_TRT", "number_of_nodes": 1, "number_of_type_nics_installed": "", "operating_system": "Ubuntu 20.04.2 LTS", "other_hardware": "", "other_software_stack": "TensorRT 8.0, CUDA 11.4, cuDNN 8.2.1, Driver 470.57.02, DALI 0.30.0", "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": "preview", "submitter": "GIGABYTE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/GIGABYTE", "sw_notes": "", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/GIGABYTE-G492-ID0_A100-SXM-80GBx8_TRT", "system_name": "GIGABYTE-G492-ID0_A100-SXM-80GBx8", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2400000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 64, "uid": "71c8f612996d2765", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 689724095655, "90.00 percentile latency (ns)": 1212395389056, "95.00 percentile latency (ns)": 1277980499711, "97.00 percentile latency (ns)": 1304113194534, "99.00 percentile latency (ns)": 1330356885114, "99.90 percentile latency (ns)": 1342182910050, "Max latency (ns)": 1343473505805, "Mean latency (ns)": 3018765898, "Min duration satisfied": "Yes", "Min latency (ns)": 37406192722, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 1179030.0, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-SXM-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 1179030.0, "characteristics.samples_per_second.normalized_per_core": 147378.75, "characteristics.samples_per_second.normalized_per_processor": 147378.75, "ck_system": "GIGABYTE_G492-ZD0_A100-SXM-80GBx8", "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, "division": "closed", "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": 2, "host_storage_capacity": "3.84 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99.9", "input_data_types": "int8", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 8, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/GIGABYTE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/GIGABYTE/results/GIGABYTE_G492-ZD0_A100-SXM-80GBx8", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, 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, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 1584000000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "GIGABYTE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/GIGABYTE", "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/GIGABYTE_G492-ZD0_A100-SXM-80GBx8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2400000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "d2449db6e358bb1f", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" }, { "50.00 percentile latency (ns)": 689724095655, "90.00 percentile latency (ns)": 1212395389056, "95.00 percentile latency (ns)": 1277980499711, "97.00 percentile latency (ns)": 1304113194534, "99.00 percentile latency (ns)": 1330356885114, "99.90 percentile latency (ns)": 1342182910050, "Max latency (ns)": 1343473505805, "Mean latency (ns)": 3018765898, "Min duration satisfied": "Yes", "Min latency (ns)": 37406192722, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "DLRM SERVER", "Samples per second": 1179030.0, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "80GB", "accelerator_memory_configuration": "HBM2", "accelerator_model_name": "NVIDIA A100-SXM-80GB", "accelerator_on-chip_memories": "", "accelerators_per_node": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.samples_per_second": 1179030.0, "characteristics.samples_per_second.normalized_per_core": 147378.75, "characteristics.samples_per_second.normalized_per_processor": 147378.75, "ck_system": "GIGABYTE_G492-ZD0_A100-SXM-80GBx8", "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, "division": "closed", "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": 2, "host_storage_capacity": "3.84 TB", "host_storage_type": "NVMe SSD", "hw_notes": "", "informal_model": "dlrm-99", "input_data_types": "int8", "max_async_queries": 1, "max_duration (ms)": 0, "max_query_count": 0, "min_duration (ms)": 600000, "min_query_count": 1, "mlperf_version": 1.0, "normalize_cores": 8, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/GIGABYTE/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/GIGABYTE/results/GIGABYTE_G492-ZD0_A100-SXM-80GBx8", "number_of_nodes": 1, "operating_system": "Ubuntu 18.04.5 LTS (Linux-5.4.0-1055-azure-x86_64-with-Ubuntu-18.04-bionic)", "other_software_stack": "TensorRT 7.2.3, CUDA 11.1, cuDNN 8.1.1, 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, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "N", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 1584000000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "tb00_40M.pt", "status": "available", "submitter": "GIGABYTE", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/GIGABYTE", "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/GIGABYTE_G492-ZD0_A100-SXM-80GBx8", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 2400000.0, "task": "recommendation", "task2": "recommendation", "total_cores": 240, "uid": "54281ba7bdf1f934", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "quantization, affine fusion" } ]