[ { "50.00 percentile latency (ns)": 322539776977, "90.00 percentile latency (ns)": 580474766013, "95.00 percentile latency (ns)": 612721207254, "97.00 percentile latency (ns)": 625618502239, "99.00 percentile latency (ns)": 638518171985, "99.90 percentile latency (ns)": 644325118110, "Max latency (ns)": 644974840487, "Mean latency (ns)": 322532731747, "Min duration satisfied": "Yes", "Min latency (ns)": 66074761, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "QAIC_SUT", "Samples per second": 6958.41, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "16x PCIe 8 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 PCIe HHHL", "accelerator_on-chip_memories": "144 MB", "accelerators_per_node": 16, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.mAP": 19.827, "characteristics.power": 1927.840310077522, "characteristics.power.normalized_per_core": 120.49001937984512, "characteristics.power.normalized_per_processor": 120.49001937984512, "characteristics.samples_per_second": 6958.41, "characteristics.samples_per_second.normalized_per_core": 434.900625, "characteristics.samples_per_second.normalized_per_processor": 434.900625, "ck_system": "g292_z43_q16-qaic-v1.5.6-aic100", "ck_used": true, "cooling": "air", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "Qualcomm Cloud AI SDK v1.5.6", "host_memory_capacity": "16x 64 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "L1d: 32K; L1i: 32K; L2: 512K; L3: 256M", "host_processor_core_count": 64, "host_processor_frequency": "2000 MHz", "host_processor_interconnect": "1x 16-20GT/s xGMI2 (?)", "host_processor_model_name": "AMD EPYC 7713 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1 TB", "host_storage_type": "1x SSD", "hw_notes": "With 75W Accelerator TDP constraints", "informal_model": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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": 16, "normalize_processors": 16, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Qualcomm/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Qualcomm/results/g292_z43_q16-qaic-v1.5.6-aic100", "number_of_nodes": 1, "operating_system": "CentOS 7 (Linux kernel: 5.4.1-1.el7.elrepo.x86_64 #1 SMP Fri Nov 29 10:21:13 EST 2019 x86_64 x86_64 x86_64 GNU/Linux)", "other_software_stack": "Docker 20.10.7; Collective Knowledge v1.55.5", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "No", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 4488000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "https://zenodo.org/record/3228411/files/resnet34-ssd1200.onnx", "status": "available", "submitter": "Qualcomm", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Qualcomm", "sw_notes": "Powered by Collective Knowledge v1.55.5", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/g292_z43_q16-qaic-v1.5.6-aic100", "system_name": "Gigabyte G292-Z43 16x QAIC100", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 6800, "task": "object detection", "task2": "object detection", "total_cores": 128, "uid": "6b70791a07b4f50a", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration.md" }, { "50.00 percentile latency (ns)": 322747617096, "90.00 percentile latency (ns)": 581885805272, "95.00 percentile latency (ns)": 614290268969, "97.00 percentile latency (ns)": 627254666002, "99.00 percentile latency (ns)": 640220338843, "99.90 percentile latency (ns)": 646052610663, "Max latency (ns)": 646699680965, "Mean latency (ns)": 323005529732, "Min duration satisfied": "Yes", "Min latency (ns)": 43056926, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "QAIC_SUT", "Samples per second": 3469.93, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "8x PCIe 8 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 PCIe HHHL", "accelerator_on-chip_memories": "144 MB", "accelerators_per_node": 8, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.mAP": 19.827, "characteristics.power": 862.791190108192, "characteristics.power.normalized_per_core": 107.848898763524, "characteristics.power.normalized_per_processor": 107.848898763524, "characteristics.samples_per_second": 3469.93, "characteristics.samples_per_second.normalized_per_core": 433.74125, "characteristics.samples_per_second.normalized_per_processor": 433.74125, "ck_system": "r282_z93_q8-qaic-v1.5.6-aic100", "ck_used": true, "cooling": "air", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "Qualcomm Cloud AI SDK v1.5.6", "host_memory_capacity": "8x 64 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "L1d: 32K; L1i: 32K; L2: 512K; L3: 16384K", "host_processor_core_count": 16, "host_processor_frequency": "2800 MHz", "host_processor_interconnect": "1x 16-20GT/s xGMI2", "host_processor_model_name": "AMD EPYC 7282 16-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1 TB", "host_storage_type": "1x SSD", "hw_notes": "With 75W Accelerator TDP constraints. 3x QAIC100 on riser CRS2033; 3x QAIC100 on riser CRS2033; 2x QAIC100 on riser CRS2026", "informal_model": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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": 8, "normalize_processors": 8, "note_code": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Qualcomm/code", "note_details": "https://github.com/mlcommons/inference_results_v1.1/tree/master/closed/Qualcomm/results/r282_z93_q8-qaic-v1.5.6-aic100", "number_of_nodes": 1, "operating_system": "CentOS 7 (Linux kernel: 5.4.1-1.el7.elrepo.x86_64 #1 SMP Fri Nov 29 10:21:13 EST 2019 x86_64 x86_64 x86_64 GNU/Linux)", "other_software_stack": "Docker 20.10.7; Collective Knowledge v1.55.5", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 1024, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 1624344308455410291, "retraining": "No", "sample_index_rng_seed": 517984244576520566, "samples_per_query": 2244000, "schedule_rng_seed": 10051496985653635065, "starting_weights_filename": "https://zenodo.org/record/3228411/files/resnet34-ssd1200.onnx", "status": "available", "submitter": "Qualcomm", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Qualcomm", "sw_notes": "Powered by Collective Knowledge v1.55.5", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/r282_z93_q8-qaic-v1.5.6-aic100", "system_name": "Gigabyte R282-Z93 8x QAIC100", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 3400, "task": "object detection", "task2": "object detection", "total_cores": 32, "uid": "69646418d92208f2", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration.md" }, { "50.00 percentile latency (ns)": 334410171022, "90.00 percentile latency (ns)": 601443569237, "95.00 percentile latency (ns)": 634836231128, "97.00 percentile latency (ns)": 648194854308, "99.00 percentile latency (ns)": 661563509704, "99.90 percentile latency (ns)": 667539524447, "Max latency (ns)": 668201234029, "Mean latency (ns)": 334386483871, "Min duration satisfied": "Yes", "Min latency (ns)": 77215105, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "Result is": "VALID", "SUT name": "QAIC_SUT", "Samples per second": 1777.91, "Scenario": "offline", "accelerator_frequency": "", "accelerator_host_interconnect": "5x PCIe 8 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 PCIe HHHL", "accelerator_on-chip_memories": "144 MB", "accelerators_per_node": 5, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.mAP": 19.81, "characteristics.power": 598.0932634730543, "characteristics.power.normalized_per_core": 119.61865269461086, "characteristics.power.normalized_per_processor": 119.61865269461086, "characteristics.samples_per_second": 1777.91, "characteristics.samples_per_second.normalized_per_core": 355.582, "characteristics.samples_per_second.normalized_per_processor": 355.582, "ck_system": "r282_z93_q5-qaic-v1.3.14-aic100", "ck_used": true, "cooling": "air", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.samples_per_second", "dim_x_maximize": true, "dim_y_default": "characteristics.mAP", "dim_y_maximize": true, "division": "closed", "formal_model": "ssd-mobilenet", "formal_model_accuracy": 99.0, "formal_model_link": "https://github.com/mlcommons/ck-mlops/tree/main/package", "framework": "Qualcomm Cloud AI SDK v1.3.14", "host_memory_capacity": "4x 32 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "L1d: 32K; L1i: 32K; L2: 512K; L3: 16384K", "host_processor_core_count": 120, "host_processor_frequency": "2800 MHz", "host_processor_interconnect": "1x 16-20GT/s xGMI2", "host_processor_model_name": "AMD EPYC 7V13 64-Core Processor", "host_processors_per_node": 2, "host_storage_capacity": "1 TB", "host_storage_type": "1x Seagate ST1000NM0045 HDD", "hw_notes": "With 75W Accelerator TDP constraints", "informal_model": "ssd-resnet34", "input_data_types": "int8", "key.accuracy": "characteristics.mAP", "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": 5, "normalize_processors": 5, "note_code": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Qualcomm/code", "note_details": "https://github.com/mlcommons/inference_results_v1.0/tree/master/closed/Qualcomm/results/r282_z93_q5-qaic-v1.3.14-aic100", "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": "GCC 7.5.0; Python 3.7.10", "performance_issue_same": 0, "performance_issue_same_index": 0, "performance_issue_unique": 0, "performance_sample_count": 64, "print_timestamps": 0, "problem": false, "qsl_rng_seed": 7322528924094909334, "retraining": "No", "sample_index_rng_seed": 1570999273408051088, "samples_per_query": 1188000, "schedule_rng_seed": 3507442325620259414, "starting_weights_filename": "https://zenodo.org/record/3228411/files/resnet34-ssd1200.onnx", "status": "available", "submitter": "Qualcomm", "submitter_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.submitter/Qualcomm", "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/r282_z93_q5-qaic-v1.3.14-aic100", "system_name": "Microsoft Corporation 7.0 (Virtual Machine)", "system_type": "datacenter", "target_latency (ns)": 0, "target_qps": 1800, "task": "object detection", "task2": "object detection", "total_cores": 240, "uid": "eaad457dddd09bda", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration/" } ]