[ { "50.00 percentile latency (ns)": 28192123, "90.00 percentile latency (ns)": 30435498, "90th percentile latency (ns)": 30435498, "95.00 percentile latency (ns)": 31079646, "97.00 percentile latency (ns)": 31449376, "99.00 percentile latency (ns)": 32030505, "99.90 percentile latency (ns)": 34758502, "Max latency (ns)": 46670145, "Mean latency (ns)": 28254589, "Min duration satisfied": "Yes", "Min latency (ns)": 20417290, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 35.15, "QPS w/o loadgen overhead": 35.39, "Result is": "VALID", "SUT name": "QAIC_SUT", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "PCIe 4 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "16 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 DM.2", "accelerator_on-chip_memories": "144 MB", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 30.435498, "characteristics.90th_percentile_latency_ns": 30435498.0, "characteristics.90th_percentile_latency_s": 0.030435498, "characteristics.90th_percentile_latency_us": 30435.498, "characteristics.mAP": 19.87, "characteristics.power": 0.5953544536768118, "characteristics.power.normalized_per_core": 0.5953544536768118, "characteristics.power.normalized_per_processor": 0.5953544536768118, "ck_system": "aedk_20w-qaic-v1.3.14-aic100", "ck_used": true, "cooling": "passive", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "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/octoml/mlops/tree/main/package, https://github.com/ctuning/ai/tree/main/package", "framework": "Qualcomm Cloud AI SDK v1.3.14", "host_memory_capacity": "12 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "1x (L1d: 64K; L1i: 64K; L2: 512K); 3x (L1d: 64K; L1i: 64K; L2: 256K); 4x (L1d: 64K; L1i: 64K; L2: 128K); L3: 4M", "host_processor_core_count": 8, "host_processor_frequency": "1x (Cortex-A77 @ 2840 MHz); 3x (Cortex-A77 @ 2420 MHz); 4x (Cortex-A55 @ 1800 MHz)", "host_processor_interconnect": "", "host_processor_model_name": "Qualcomm Snapdragon 865", "host_processors_per_node": 1, "host_storage_capacity": "894 GB", "host_storage_type": "NVMe SSD", "hw_notes": "With 20W 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": 1024, "mlperf_version": "v1.0", "normalize_cores": 1, "normalize_processors": 1, "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/aedk_20w-qaic-v1.3.14-aic100", "number_of_nodes": 1, "operating_system": "CentOS 8.0 (Linux kernel: 4.19.81 #1 SMP PREEMPT Wed Dec 2 13:10:17 UTC 2020 aarch64 aarch64 aarch64 GNU/Linux)", "other_software_stack": "Collective Knowledge v1.55.2", "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": 1, "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 Collective Knowledge 1.55.2", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/aedk_20w-qaic-v1.3.14-aic100", "system_name": "AI Development Kit", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 33.3333, "task": "object detection", "task2": "object detection", "total_cores": 8, "uid": "2b48cf0acaa1429d", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration/" }, { "50.00 percentile latency (ns)": 29482357, "90.00 percentile latency (ns)": 31703549, "90th percentile latency (ns)": 31703549, "95.00 percentile latency (ns)": 32236973, "97.00 percentile latency (ns)": 32603109, "99.00 percentile latency (ns)": 33188875, "99.90 percentile latency (ns)": 34913621, "Max latency (ns)": 59885836, "Mean latency (ns)": 29503958, "Min duration satisfied": "Yes", "Min latency (ns)": 21273352, "Min queries satisfied": "Yes", "Mode": "PerformanceOnly", "QPS w/ loadgen overhead": 33.62, "QPS w/o loadgen overhead": 33.89, "Result is": "VALID", "SUT name": "QAIC_SUT", "Scenario": "singlestream", "accelerator_frequency": "", "accelerator_host_interconnect": "PCIe 4 lane Gen4", "accelerator_interconnect": "", "accelerator_interconnect_topology": "", "accelerator_memory_capacity": "8 GB", "accelerator_memory_configuration": "LPDDR4x", "accelerator_model_name": "QUALCOMM Cloud AI 100 DM.2e", "accelerator_on-chip_memories": "72 MB", "accelerators_per_node": 1, "accuracy_log_probability": 0, "accuracy_log_rng_seed": 0, "accuracy_log_sampling_target": 0, "characteristics.90th_percentile_latency_ms": 31.703549, "characteristics.90th_percentile_latency_ns": 31703549.0, "characteristics.90th_percentile_latency_s": 0.031703549, "characteristics.90th_percentile_latency_us": 31703.549, "characteristics.mAP": 19.87, "characteristics.power": 0.6272178085003224, "characteristics.power.normalized_per_core": 0.6272178085003224, "characteristics.power.normalized_per_processor": 0.6272178085003224, "ck_system": "aedk_15w-qaic-v1.3.14-aic100", "ck_used": true, "cooling": "passive", "dataset": "COCO 2017 (300x300)", "dataset_link": "https://github.com/ctuning/ck/blob/master/docs/mlperf-automation/datasets/coco2017.md", "dim_x_default": "characteristics.90th_percentile_latency_ms", "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/octoml/mlops/tree/main/package, https://github.com/ctuning/ai/tree/main/package", "framework": "Qualcomm Cloud AI SDK v1.3.14", "host_memory_capacity": "12 GB", "host_memory_configuration": "", "host_networking": "", "host_networking_topology": "", "host_processor_caches": "1x (L1d: 64K; L1i: 64K; L2: 512K); 3x (L1d: 64K; L1i: 64K; L2: 256K); 4x (L1d: 64K; L1i: 64K; L2: 128K); L3: 4M", "host_processor_core_count": 8, "host_processor_frequency": "1x (Cortex-A77 @ 2840 MHz); 3x (Cortex-A77 @ 2420 MHz); 4x (Cortex-A55 @ 1800 MHz)", "host_processor_interconnect": "", "host_processor_model_name": "Qualcomm Snapdragon 865", "host_processors_per_node": 1, "host_storage_capacity": "894 GB", "host_storage_type": "NVMe SSD", "hw_notes": "With 15W 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": 1024, "mlperf_version": "v1.0", "normalize_cores": 1, "normalize_processors": 1, "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/aedk_15w-qaic-v1.3.14-aic100", "number_of_nodes": 1, "operating_system": "CentOS 8.0 (Linux kernel: 4.19.81 #1 SMP PREEMPT Wed Dec 2 13:10:17 UTC 2020 aarch64 aarch64 aarch64 GNU/Linux)", "other_software_stack": "Collective Knowledge v1.55.2", "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": 1, "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 Collective Knowledge 1.55.2", "system_link": "https://github.com/ctuning/ck-mlperf-inference/tree/main/bench.mlperf.system/aedk_15w-qaic-v1.3.14-aic100", "system_name": "AI Development Kit", "system_type": "edge", "target_latency (ns)": 0, "target_qps": 33.3333, "task": "object detection", "task2": "object detection", "total_cores": 8, "uid": "38cb457a74936842", "use_accelerator": true, "weight_data_types": "int8", "weight_transformations": "see calibration/" } ]