The table below illustrates the speed-up factor for the performance gain by switching from an FP32 representation of an OpenVINO™ supported model to its INT8 representation.
Intel® Core™ i7-8700T | Intel® Xeon® Gold 5218T | Intel® Xeon® Platinum 8270 | Intel® Core™ i7-1185G7 | ||
---|---|---|---|---|---|
OpenVINO benchmark model name | Dataset | Throughput speed-up FP16-INT8 vs FP32 | |||
bert-large- uncased-whole-word- masking-squad-0001 | SQuAD | 1.6 | 2.7 | 2.0 | 2.6 |
brain-tumor- segmentation- 0001-MXNET | BraTS | 1.5 | 1.9 | 1.7 | 1.8 |
deeplabv3-TF | VOC 2012 Segmentation | 1.5 | 2.4 | 2.8 | 3.1 |
densenet-121-TF | ImageNet | 1.6 | 3.2 | 3.2 | 3.2 |
facenet- 20180408- 102900-TF | LFW | 2.0 | 3.6 | 3.5 | 3.4 |
faster_rcnn_ resnet50_coco-TF | MS COCO | 1.7 | 3.4 | 3.4 | 3.4 |
googlenet-v1-TF | ImageNet | 1.8 | 3.6 | 3.7 | 3.5 |
inception-v3-TF | ImageNet | 1.8 | 3.8 | 4.0 | 3.5 |
mobilenet- ssd-CF | VOC2012 | 1.5 | 3.1 | 3.6 | 3.1 |
mobilenet-v1-1.0- 224-TF | ImageNet | 1.5 | 3.2 | 4.1 | 3.1 |
mobilenet-v2-1.0- 224-TF | ImageNet | 1.3 | 2.7 | 4.3 | 2.5 |
mobilenet-v2- pytorch | ImageNet | 1.4 | 2.8 | 4.6 | 2.4 |
resnet-18- pytorch | ImageNet | 1.9 | 3.7 | 3.8 | 3.6 |
resnet-50- pytorch | ImageNet | 1.8 | 3.6 | 3.9 | 3.4 |
resnet-50- TF | ImageNet | 1.8 | 3.6 | 3.9 | 3.4 |
squeezenet1.1- CF | ImageNet | 1.6 | 2.9 | 3.4 | 3.2 |
ssd_mobilenet_ v1_coco-tf | VOC2012 | 1.6 | 3.1 | 3.7 | 3.0 |
ssd300-CF | MS COCO | 1.8 | 3.7 | 3.7 | 3.8 |
ssdlite_ mobilenet_ v2-TF | MS COCO | 1.4 | 2.3 | 3.9 | 2.5 |
yolo_v3-TF | MS COCO | 1.8 | 3.8 | 3.9 | 3.6 |
The following table shows the absolute accuracy drop that is calculated as the difference in accuracy between the FP32 representation of a model and its INT8 representation.
Intel® Core™ i9-10920X CPU @ 3.50GHZ (VNNI) | Intel® Core™ i9-9820X CPU @ 3.30GHz (AVX512) | Intel® Core™ i7-6700 CPU @ 4.0GHz (AVX2) | Intel® Core™ i7-1185G7 CPU @ 4.0GHz (TGL VNNI) | |||
---|---|---|---|---|---|---|
OpenVINO Benchmark Model Name | Dataset | Metric Name | Absolute Accuracy Drop, % | |||
brain-tumor- segmentation- 0001-MXNET | BraTS | Dice-index@ Mean@ Overall Tumor | 0.08 | 0.08 | 0.08 | 0.08 |
deeplabv3-TF | VOC 2012 Segmentation | mean_iou | 0.73 | 1.10 | 1.10 | 0.73 |
densenet-121-TF | ImageNet | acc@top-1 | 0.73 | 0.72 | 0.72 | 0.73 |
facenet- 20180408- 102900-TF | LFW | pairwise_ accuracy _subsets | 0.02 | 0.02 | 0.02 | 0.47 |
faster_rcnn_ resnet50_coco-TF | MS COCO | coco_ precision | 0.21 | 0.20 | 0.20 | 0.21 |
googlenet-v1-TF | ImageNet | acc@top-1 | 0.03 | 0.01 | 0.01 | 0.03 |
inception-v3-TF | ImageNet | acc@top-1 | 0.03 | 0.01 | 0.01 | 0.03 |
mobilenet- ssd-CF | VOC2012 | mAP | 0.35 | 0.34 | 0.34 | 0.35 |
mobilenet-v1-1.0- 224-TF | ImageNet | acc@top-1 | 0.27 | 0.20 | 0.20 | 0.27 |
mobilenet-v2-1.0- 224-TF | ImageNet | acc@top-1 | 0.44 | 0.92 | 0.92 | 0.44 |
mobilenet-v2- PYTORCH | ImageNet | acc@top-1 | 0.25 | 7.42 | 7.42 | 0.25 |
resnet-18- pytorch | ImageNet | acc@top-1 | 0.26 | 0.25 | 0.25 | 0.26 |
resnet-50- PYTORCH | ImageNet | acc@top-1 | 0.18 | 0.19 | 0.19 | 0.18 |
resnet-50- TF | ImageNet | acc@top-1 | 0.15 | 0.11 | 0.11 | 0.15 |
squeezenet1.1- CF | ImageNet | acc@top-1 | 0.66 | 0.64 | 0.64 | 0.66 |
ssd_mobilenet_ v1_coco-tf | VOC2012 | COCO mAp | 0.24 | 3.07 | 3.07 | 0.24 |
ssd300-CF | MS COCO | COCO mAp | 0.06 | 0.05 | 0.05 | 0.06 |
ssdlite_ mobilenet_ v2-TF | MS COCO | COCO mAp | 0.14 | 0.43 | 0.43 | 0.14 |
yolo_v3-TF | MS COCO | COCO mAp | 0.12 | 0.35 | 0.35 | 0.12 |
For more complete information about performance and benchmark results, visit: www.intel.com/benchmarks and Optimization Notice. Legal Information.