Artificial Intelligence with 2nd Gen Intel® Xeon® Scalable Processor

The 2nd Gen Intel® Xeon® Scalable processor provides scalable performance for the widest variety of datacenter workloads – including deep learning. The new 2nd Gen Intel® Xeon® Scalable processor platform offers built-in Return on Investment (ROI), potent performance and production-ready support for AI deployments.

In our smart and connected world, machines are increasingly learning to sense, reason, act, and adapt in the real world. Artificial Intelligence (AI) is the next big wave of computing, and Intel uniquely has the experience to fuel the AI computing era. AI will let us accelerate solutions to large-scale problems that would otherwise take months, years, or decades to resolve.

AI is expected to unleash new scientific discoveries, automate monotonous tasks and extend our human senses and capabilities. Today, machine learning (ML) and deep learning (DL) are two underlying approaches to AI, as are reasoning-based systems.

Deep learning is the most rapidly emerging branch of machine learning, in many cases supplanting classic ML, relying on massive labeled data sets to iteratively “train” many-layered neural networks inspired by the human brain. Trained neural networks are used to “infer” the meaning of new data, with increased speed and accuracy for processes like image search, speech recognition, natural language processing, and other complex tasks.

The 2nd Generation Intel® Xeon® Scalable processors take AI performance to the next level with Intel® Deep Learning Boost (Intel DL Boost), a new set of embedded processor technologies designed to accelerate AI deep learning use cases such as image recognition, object detection, speech recognition, language translation and others. It extends Intel® Advanced Vector Extensions 512 (Intel® AVX-512) with a new Vector Neural Network Instruction (VNNI) that significantly increases deep learning inference performance over previous generations. With 2nd Gen Intel® Xeon® Platinum 8280 processors and Intel® Deep Learning Boost (Intel DL Boost), we project that image recognition with Intel optimized Caffe ResNet-50 can perform up to 14x1 faster than on prior generation Intel® Xeon® Scalable processors (at launch, July 2017).

Performance data as of April 2, 2019.1 2 3 4

ResNet-50 performance with Intel® Optimization for Caffe*

Designed for high performance computing, advanced artificial intelligence and analytics, and high density infrastructures Intel® Xeon® Platinum 9200 processors deliver breakthrough levels of performance. Using Intel® Deep Learning Boost (Intel DL Boost) combined with Intel optimized Caffe, new breakthrough levels of performance can be achieved. Here we show the throughput on an image classification topology – ResNet-50 on the 2nd Generation Intel® Xeon Scalable processor.

Learn more ›

Configuration Details

Max Inference throughput at <7ms

Intel® Xeon® Platinum 8280 processor: Tested by Intel as of 3/04/2019. 2S Intel® Xeon® Platinum 8280(28 cores per socket) processor, HT ON, turbo ON, Total Memory 384 GB (12 slots/ 32 GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0348.011820191451, Centos 7 Kernel 3.10.0-957.5.1.el7.x86_64, Intel® Deep Learning Framework: Intel® Optimization for Caffe* version: https://github.com/intel/caffe Commit id: 362a3b3, ICC 2019.2.187 for build, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a), model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS=10, synthetic Data:3x224x224, 2 instance/2 socket, Datatype: INT8; latency: 6.16 ms.

Intel® Xeon® Platinum 9242 processor: Tested by Intel as of 3/04/2019 2S Intel® Xeon® Platinum 9242(48 cores per socket) processor, HT ON, turbo ON, Total Memory 768 GB (24 slots/ 32 GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0403.022020190327, Centos 7 Kernel 3.10.0-957.5.1.el7.x86_64, Intel® Deep Learning Framework: Intel® Optimization for Caffe* version: https://github.com/intel/caffe Commit id: 362a3b3, ICC 2019.2.187 for build, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a), model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS= 8, synthetic Data:3x224x224, 4 instance/2 socket, Datatype: INT8; latency: 6.90 ms.

Intel® Xeon® Platinum 9282 processor: Tested by Intel as of 3/04/2019. DL Inference: Platform: Dragon rock 2S Intel® Xeon® Platinum 9282(56 cores per socket) processor, HT ON, turbo ON, Total Memory 768 GB (24 slots/ 32 GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0241.112020180249, Centos 7 Kernel 3.10.0-957.5.1.el7.x86_64, Intel® Deep Learning Framework: Intel® Optimization for Caffe* version: https://github.com/intel/caffe Commit id: 362a3b3, ICC 2019.2.187 for build, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a), model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS=12, synthetic Data:3x224x224, 4 instance/2 socket, Datatype: INT8; latency: 6.91 ms.

Max Inference throughput

Intel® Xeon® Platinum 8280 processor: Tested by Intel as of 3/04/2019. 2S Intel® Xeon® Platinum 8280(28 cores per socket) processor, HT ON, turbo ON, Total Memory 384 GB (12 slots/ 32 GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0348.011820191451, Centos 7 Kernel 3.10.0-957.5.1.el7.x86_64, Intel® Deep Learning Framework: Intel® Optimization for Caffe* version: https://github.com/intel/caffe Commit id: 362a3b3, ICC 2019.2.187 for build, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a), model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS=10, syntheticData:3x224x224, 14 instance/2 socket, Datatype: INT8.

Intel® Xeon® Platinum 9242 processor: Tested by Intel as of 3/04/2019 2S Intel® Xeon® Platinum 9242(48 cores per socket) processor, HT ON, turbo ON, Total Memory 768 GB (24 slots/ 32 GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0403.022020190327, Centos 7 Kernel 3.10.0-957.5.1.el7.x86_64, Intel® Deep Learning Framework: Intel® Optimization for Caffe* version: https://github.com/intel/caffe Commit id: 362a3b3, ICC 2019.2.187 for build, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a), model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS=8, synthetic Data:3x224x224, 24 instance/2 socket, Datatype: INT8.

Intel® Xeon® Platinum 9282 processor: Tested by Intel as of 3/04/2019. DL Inference: Platform: Dragon rock 2S Intel® Xeon® Platinum 9282(56 cores per socket) processor, HT ON, turbo ON, Total Memory 768 GB (24 slots/ 32 GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0241.112020180249, Centos 7 Kernel 3.10.0-957.5.1.el7.x86_64, Intel® Deep Learning Framework: Intel® Optimization for Caffe*
version: https://github.com/intel/caffe Commit id: 362a3b3, ICC 2019.2.187 for build, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a), model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS=8, synthetic Data:3x224x224, 28 instance/2 socket, Datatype: INT8.

BKMs for running multi-stream configurations on Xeon: https://www.intel.ai/wp-content/uploads/sites/69/TensorFlow_Best_Practices_Intel_Xeon_AI-HPC_v1.1_Q119.pdf

ResNet-50 Inference Throughput Performance

Inference generally happens instantaneously at the edge or in the data center, such as when a new photo is uploaded for inspection. Inference output can be fed into a number of different usages including – a dashboard for visualization or a decision tree for automatic decision making. Here we show inference throughput on an image database using multiple popular deep learning frameworks such as Caffe, TensorFlow, Pytorch and MxNet with the ResNet-50 topology.

View larger image

Learn more

Configuration Details

3.0x performance boost with MxNet on ResNet-50: Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013),CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: MxNet https://github.com/apache/incubator-mxnet/ -b master da5242b732de39ad47d8ecee582f261ba5935fa9, Compiler: gcc 4.8.5,MKL DNN version: v0.17, ResNet50, BS=64, synthetic data, 2 instance/2 socket, 0.12% accuracy loss ,Datatype: INT8 vs Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0286.121520181757, CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: MxNet https://github.com/apache/incubator-mxnet/ -b master da5242b732de39ad47d8ecee582f261ba5935fa9, Compiler: gcc 4.8.5,MKL DNN version: v0.17, ResNet50, BS=64, synthetic data, 2 instance/2 socket, Datatype: FP32

3.7x performance boost with Pytorch ResNet-50: Tested by Intel as of 2/25/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode: 0x4000013), Ubuntu 18.04.1 LTS, kernel 4.15.0-45-generic, SSD 1x sda INTEL SSDSC2BA80 SSD 745.2GB, 3X INTEL SSDPE2KX040T7 SSD 3.7TB , Intel® Deep Learning Framework: Pytorch with ONNX/Caffe2 backend: https://github.com/pytorch/pytorch.git (commit: 4ac91b2d64eeea5ca21083831db5950dc08441d6)and Pull Request link: https://github.com/pytorch/pytorch/pull/17464 (submitted for upstreaming), gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, MKL DNN version: v0.17.3 (commit hash: 0c3cb94999919d33e4875177fdef662bd9413dd4), ResNet-50: https://github.com/intel/optimized-models/tree/master/pytorch, BS=512, synthetic data, 2 instance/2 socket, 0.6% accuracy loss; Datatype: INT8 vs Tested by Intel as of 2/25/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 192 GB (12 slots/ 16GB/ 2666 MHz), BIOS: SE5C620.86B.00.01.0015.110720180833 (ucode: 0x200004d), CentOS 7.5, 3.10.0-693.el7.x86_64, Intel® SSD DC S4500 SERIES SSDSC2KB480G7 2.5’’ 6Gb/s SATA SSD 480G, Intel® Deep Learning Framework: https://github.com/pytorch/pytorch.git (commit:4ac91b2d64eeea5ca21083831db5950dc08441d6)and Pull Request link: https://github.com/pytorch/pytorch/pull/17464 (submitted for upstreaming), gcc (Red Hat 5.3.1-6) 5.3.1 20160406, MKL DNN version: v0.17.3 (commit hash: 0c3cb94999919d33e4875177fdef662bd9413dd4), ResNet-50: https://github.com/intel/optimized-models/tree/master/pytorch, BS=512, synthetic data, 2 instance/2 socket, Datatype: FP32

3.9x performance boost with TensorFlow ResNet-50:Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013),CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: https://hub.docker.com/r/intelaipg/intel-optimized-tensorflow:PR25765-devel-mkl (https://github.com/tensorflow/tensorflow.git commit: 6f2eaa3b99c241a9c09c345e1029513bc4cd470a + Pull Request PR 25765, PR submitted for upstreaming) Compiler: gcc 6.3.0,MKL DNN version: v0.17, ResNet50: https://github.com/IntelAI/models/tree/master/models/image_recognition/tensorflow/resnet50, (commit: 87261e70a902513f934413f009364c4f2eed6642) BS=128, synthetic data, 2 instance/2 socket, 0.45% accuracy loss Datatype: INT8 vs Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0286.121520181757, CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: https://hub.docker.com/r/intelaipg/intel-optimized-tensorflow:PR25765-devel-mkl 6f2eaa3b99c241a9c09c345e1029513bc4cd470a + PR25765, PR submitted for upstreaming) Compiler: gcc 6.3.0,MKL DNN version: v0.17, ResNet50: https://github.com/IntelAI/models/tree/master/models/image_recognition/tensorflow/resnet50 , (commit: 87261e70a902513f934413f009364c4f2eed6642) BS=128, synthetic data, 2 instance/2 socket, Datatype: FP32

4.0x performance boost with Intel® Optimization for Caffe* ResNet-50: Tested by Intel as of 2/20/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode: 0x4000013), Ubuntu 18.04.1 LTS, kernel 4.15.0-45-generic, SSD 1x sda INTEL SSDSC2BA80 SSD 745.2GB, 3X INTEL SSDPE2KX040T7 SSD 3.7TB , Intel® Deep Learning Framework: Intel® Optimization for Caffe* version: 1.1.3 (commit hash: 7010334f159da247db3fe3a9d96a3116ca06b09a) , ICC version 18.0.1, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a, model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS=64, syntheticData, 2 instance/2 socket, Datatype: INT8 vs Tested by Intel as of 2/21/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 192 GB (12 slots/ 16GB/ 2666 MHz), BIOS: SE5C620.86B.00.01.0015.110720180833 (ucode: 0x200004d), CentOS 7.5, 3.10.0-693.el7.x86_64, Intel® SSD DC S4500 SERIES SSDSC2KB480G7 2.5’’ 6Gb/s SATA SSD 480G, , Intel® Deep Learning Framework: Intel® Optimization for Caffe* version: 1.1.3 (commit hash: 7010334f159da247db3fe3a9d96a3116ca06b09a) , ICC version 18.0.1, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a, model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/benchmark/resnet_50/deploy.prototxt, BS=64, synthetic Data, 2 instance/2 socket, Datatype: FP32

Inference Throughput Performance

The 2nd Gen Intel® Xeon® Scalable processors are built specifically to run high-performance AI and IoT workloads on the same hardware as other existing workloads. Intel® Deep Learning Boost (Intel DL Boost) can benefit many inference applications ranging from recommendation systems, Object detection and image recognition and classification. Here we show inference throughput for image classification, object detection and a recommendation system. Multiple frameworks are used including TensorFlow*, Caffe2 and MxNet* and multiple topologies such as ResNet-101, Inception v3, RETINANET*, SSD-VGG16 and Wide and Deep.

View larger image

Learn more

Configuration Details

4.0x performance boost with TensorFlow ResNet-101: Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013),CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: https://hub.docker.com/r/intelaipg/intel-optimized-tensorflow:PR25765-devel-mkl (https://github.com/tensorflow/tensorflow.git commit: 6f2eaa3b99c241a9c09c345e1029513bc4cd470a + Pull Request PR 25765, PR submitted for upstreaming), Compiler: gcc 6.3.0,MKL DNN version: v0.17, ResNet 101 : https://github.com/IntelAI/models/tree/master/models/image_recognition/tensorflow/resnet101 commit: 87261e70a902513f934413f009364c4f2eed6642 , BS=128, synthetic data, 2 instance/2 socket, 0.58% accuracy loss Datatype: INT8 vs Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0286.121520181757, CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: https://hub.docker.com/r/intelaipg/intel-optimized-tensorflow:PR25765-devel-mkl (https://github.com/tensorflow/tensorflow.git commit: 6f2eaa3b99c241a9c09c345e1029513bc4cd470a + Pull Request PR 25765, PR submitted for upstreaming) Compiler: gcc 6.3.0,MKL DNN version: v0.17, ResNet 101 : https://github.com/IntelAI/models/tree/master/models/image_recognition/tensorflow/resnet101 commit: 87261e70a902513f934413f009364c4f2eed6642 , BS=128, synthetic data, 2 instance/2 socket, Datatype: FP32

3.8x performance boost with MxNet ResNet 101: Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013),CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: MxNet https://github.com/apache/incubator-mxnet.git commit: da5242b732de39ad47d8ecee582f261ba5935fa9 , Compiler: gcc 4.8.5,MKL DNN version: v0.17, ResNet 101, BS= 64, Synthetic data, 2 instance/2 socket, 0.56% accuracy loss, Datatype: INT8 vs Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0286.121520181757, CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: MxNet: https://github.com/apache/incubator-mxnet.git commit: da5242b732de39ad47d8ecee582f261ba5935fa9, Compiler: gcc 4.8.5,MKL DNN version: v0.17, ResNet 101, BS= 64, synthetic Data, 2 instance/2 socket, Datatype:FP32

3.1x performance boost with TensorFlow Inception v3: Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013),CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: https://hub.docker.com/r/intelaipg/intel-optimized-tensorflow:PR25765-devel-mkl (https://github.com/tensorflow/tensorflow.git commit: 6f2eaa3b99c241a9c09c345e1029513bc4cd470a + Pull Request PR 25765, PR submitted for upstreaming), Compiler: gcc 6.3.0,MKL DNN version: v0.17, Inception v3 : https://github.com/IntelAI/models/tree/master/models/image_recognition/tensorflow/inceptionv3 commit: 87261e70a902513f934413f009364c4f2eed6642 , BS=128, synthetic data, 2 instance/2 socket, Datatype: INT8 vs Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0286.121520181757, CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: https://hub.docker.com/r/intelaipg/intel-optimized-tensorflow:PR25765-devel-mkl (https://github.com/tensorflow/tensorflow.git commit: 6f2eaa3b99c241a9c09c345e1029513bc4cd470a + Pull Request PR 25765, PR submitted for upstreaming) Compiler: gcc 6.3.0,MKL DNN version: v0.17, Inception v3 : https://github.com/IntelAI/models/tree/master/models/image_recognition/tensorflow/inceptionv3 commit: 87261e70a902513f934413f009364c4f2eed6642 , BS=128, synthetic data, 2 instance/2 socket, Datatype: FP32

2.6x performance boost with PyTorch RetinaNet: Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode: 0x4000013), Ubuntu 18.04.1 LTS, kernel 4.15.0-45-generic, SSD 1x sda INTEL SSDSC2BA80 SSD 745.2GB, 3X INTEL SSDPE2KX040T7 SSD 3.7TB , Intel® Deep Learning Framework: Pytorch with ONNX/Caffe2 backend: https://github.com/pytorch/pytorch.git (commit: 4ac91b2d64eeea5ca21083831db5950dc08441d6)and Pull Request link: https://github.com/pytorch/pytorch/pull/17464 (submitted for upstreaming), gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, MKL DNN version: v0.17.3 (commit hash: 0c3cb94999919d33e4875177fdef662bd9413dd4), RetinaNet: https://github.com/intel/Detectron/blob/master/configs/12_2017_baselines/retinanet_R-101-FPN_1x.yaml BS=1, synthetic data, 2 instance/2 socket, 0.003mAP accuracy loss, Datatype: INT8 vs Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 192 GB (12 slots/ 16GB/ 2666 MHz), BIOS: SE5C620.86B.00.01.0015.110720180833 (ucode: 0x200004d), CentOS 7.5, 3.10.0-693.el7.x86_64, Intel® SSD DC S4500 SERIES SSDSC2KB480G7 2.5’’ 6Gb/s SATA SSD 480G, Intel® Deep Learning Framework: Pytorch with ONNX/Caffe2 backend: https://github.com/pytorch/pytorch.git (commit: 4ac91b2d64eeea5ca21083831db5950dc08441d6)and Pull Request link: https://github.com/pytorch/pytorch/pull/17464 (submitted for upstreaming), gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, MKL DNN version: v0.17.3 (commit hash: 0c3cb94999919d33e4875177fdef662bd9413dd4), RetinaNet: https://github.com/intel/Detectron/blob/master/configs/12_2017_baselines/retinanet_R-101-FPN_1x.yaml, BS=1, synthetic data, 2 instance/2 socket, Datatype: FP32

2.5x performance boost with MxNet SSD-VGG16: Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013),CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: MxNet https://github.com/apache/incubator-mxnet/ -b master da5242b732de39ad47d8ecee582f261ba5935fa9, Compiler: gcc 4.8.5,MKL DNN version: v0.17, SSD-VGG16: https://github.com/apache/incubator-MXNet/blob/master/example/ssd/symbol/vgg16_reduced.py ,BS= 224, Synthetic data, 2 instance/2 socket, 0.0001 mAP accuracy loss , Datatype: INT8 vs Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0286.121520181757, CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: MxNet https://github.com/apache/incubator-mxnet/ -b master da5242b732de39ad47d8ecee582f261ba5935fa9, Compiler: gcc 4.8.5,MKL DNN version: v0.17, SSD-VGG16: https://github.com/apache/incubator-MXNet/blob/master/example/ssd/symbol/vgg16_reduced.py ,BS= 224, synthetic Data, 2 instance/2 socket, Datatype:FP32

2.2x performance boost with Intel® Optimized Caffe on SSD-Mobilenet v1: Tested by Intel as of 2/20/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode: 0x4000013), Ubuntu 18.04.1 LTS, kernel 4.15.0-45-generic, SSD 1x sda Intel® SSDSC2BA80 SSD 745.2GB, 3X INTEL SSDPE2KX040T7 SSD 3.7TB , Intel® Deep Learning Framework: Intel® Optimization for Caffe* version: 1.1.3 (commit hash: 7010334f159da247db3fe3a9d96a3116ca06b09a) , ICC version 18.0.1, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a, model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/ssd_mobilenet_int8.prototxt, BS=64, synthetic Data, 2 instance/2 socket, 0.0096 mAP accuracy loss, Datatype: INT8 vs Tested by Intel as of 2/21/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 192 GB (12 slots/ 16GB/ 2666 MHz), BIOS: SE5C620.86B.00.01.0015.110720180833 (ucode: 0x200004d), CentOS 7.5, 3.10.0-693.el7.x86_64, Intel® SSD DC S4500 SERIES SSDSC2KB480G7 2.5’’ 6Gb/s SATA SSD 480G, , Intel® Deep Learning Framework: Intel® Optimization for Caffe* version: 1.1.3 (commit hash: 7010334f159da247db3fe3a9d96a3116ca06b09a) , ICC version 18.0.1, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a, model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/ssd_mobilenet_int8.prototxt, BS=64, synthetic Data, 2 instance/2 socket, Datatype: FP32

2.1x performance boost with TensorFlow on Wide & Deep: Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013),CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: https://github.com/tensorflow/tensorflow.git 3262818d9d8f9f630f04df23033032d39a7a413 + Pull Request PR26169 + Pull Request PR26261 + Pull Request PR26271 , PR submitted for upstreaming, Compiler:gcc 6.3.1,MKL DNN version: v0.17, Wide & Deep: https://github.com/IntelAI/models/tree/master/benchmarks/recommendation/tensorflow/wide_deep_large_ds commit: a044cb3e7d2b082aebae2edbe6435e57a2cc1f8f ,BS=512, Criteo Display Advertisement Challenge, 2 instance/2 socket, 0.007% accuracy loss, Datatype: INT8 vs Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0286.121520181757, CentOS 7.6, 4.19.5-1.el7.elrepo.x86_64, Intel® Deep Learning Framework: https://github.com/tensorflow/tensorflow.git 3262818d9d8f9f630f04df23033032d39a7a413 + Pull Request PR26169 + Pull Request PR26261 + Pull Request PR26271 , PR submitted for upstreaming, Compiler:gcc 6.3.1,MKL DNN version: v0.17, Wide & Deep:https://github.com/IntelAI/models/tree/master/benchmarks/recommendation/tensorflow/wide_deep_large_ds a044cb3e7d2b082aebae2edbe6435e57a2cc1f8f, BS= 512, Criteo Display Advertisement Challenge, 2 instance/2 socket,Datatype:FP32

2.1x performance boost with MXNet Wide & Deep: Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8280L processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0348.011820191451 (ucode:0x5000017), CentOS 7.6, Kernel 4.19.5-1.el7.elrepo.x86_64, SSD 1x INTEL SSDSC2KG96 960GB, Intel® Deep Learning Framework: MXNet https://github.com/apache/incubator-mxnet.git commit f1de8e51999ce3acaa95538d21a91fe43a0286ec applying, Compiler: gcc 6.3.1, MKL DNN version: commit: 08bd90cca77683dd5d1c98068cea8b92ed05784, Wide & Deep: https://github.com/intel/optimized-models/tree/v1.0.2/mxnet/wide_deep_criteo commit: c3e7cbde4209c3657ecb6c9a142f71c3672654a5, Dataset: Criteo Display Advertisement Challenge, Batch Size=1024, 2 instance/2 socket, 0.23% accuracy loss, Datatype: INT8 vs Tested by Intel as of 3/1/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2666 MHz), BIOS: SE5C620.86B.0D.01.0286.121520181757 (ucode:0x2000057), CentOS 7.6, Kernel 4.19.5-1.el7.elrepo.x86_64, SSD 1x INTEL SSDSC2KG96 960GB, Intel® Deep Learning Framework: MXNet https://github.com/apache/incubator-mxnet.git commit f1de8e51999ce3acaa95538d21a91fe43a0286ec applying, Compiler: gcc 6.3.1, MKL DNN version: commit: 08bd90cca77683dd5d1c98068cea8b92ed05784, Wide & Deep: https://github.com/intel/optimized-models/tree/v1.0.2/mxnet/wide_deep_criteo commit: c3e7cbde4209c3657ecb6c9a142f71c3672654a5, Dataset: Criteo Display Advertisement Challenge, Batch Size=1024, 2 instance/2 socket, Datatype:FP32

OpenVINO™ Toolkit5 Inference Throughput Performance

AI at the edge is opening up new possibilities in every industry, from predicting machine failures to personalizing retail. With the OpenVINO™ toolkit, businesses can take advantage of near real-time insights to help make better decisions, faster. The OpenVINO™ toolkit allows your business to implement computer vision and deep learning solutions quickly and effectively across multiple applications.

View larger image

Learn more

Configuration Details

2.4x performance boost with OpenVINO™ toolkit on SqueezeNet v1.1: Tested by Intel as of 1/30/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013), Linux-4.15.0-43-generic-x86_64-with-debian-buster-sid, Compiler: gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, Deep Learning Deployment Toolkit (DLDT): OpenVINO™ toolkit R5.01, SqueezeNet v1.1, BS=64, Imagenet, 1 instance/2 socket, Datatype: INT8 vs Tested by Intel as of 1/30/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 192 GB (12 slots/ 16GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605, Linux-4.15.0-29-generic-x86_64-with-Ubuntu-18.04-bionic, Compiler: gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, Deep Learning Deployment Toolkit (DLDT): OpenVINO™ toolkit R5 (DLDTK Version:1.0.19154), SqueezeNet v1.1, Imagenet images , 1 instance/2 socket, Datatype: FP32 (BS=16)

3.1x performance boost with OpenVino™ toolkit on MobileNet v1: Tested by Intel as of 1/30/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013), Linux-4.15.0-43-generic-x86_64-with-debian-buster-sid, Compiler: gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, Deep Learning Deployment Toolkit (DLDT): OpenVINO™ toolkit R5.01, MobileNet v1, BS=64, Imagenet, 1 instance/2 socket, Datatype: INT8 vs Tested by Intel as of 1/30/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 192 GB (12 slots/ 16GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605, Linux-4.15.0-29-generic-x86_64-with-Ubuntu-18.04-bionic, Compiler: gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, Deep Learning Deployment Toolkit (DLDT): OpenVINO™ toolkit R5 (DLDTK Version:1.0.19154), MobileNet v1, Imagenet, 1 instance/2 socket, Datatype: FP32 (BS=16)

3.2x performance boost with OpenVINO™ toolkit on Inception v4 :Tested by Intel as of 1/30/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013), Linux-4.15.0-43-generic-x86_64-with-debian-buster-sid, Compiler: gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, Deep Learning Deployment Toolkit (DLDT): OpenVINO™ toolkit R5.01, Inception v4, BS=128, Imagenet, 1 instance/2 socket, Datatype: INT8 vs Tested by Intel as of 1/30/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 192 GB (12 slots/ 16GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605, Linux-4.15.0-29-generic-x86_64-with-Ubuntu-18.04-bionic, Compiler: gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, Deep Learning Deployment Toolkit (DLDT): OpenVINO™ toolkit R5 (DLDTK Version:1.0.19154), Inception v4, Imagenet, 1 instance/2 socket, Datatype: FP32 (BS=16)

3.9x performance boost with OpenVINO™ toolkit ResNet-50: Tested by Intel as of 1/30/2019. 2 socket Intel® Xeon® Platinum 8280 processor, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode:0x4000013), Linux-4.15.0-43-generic-x86_64-with-debian-buster-sid, Compiler: gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, Deep Learning Deployment Toolkit (DLDT): OpenVINO™ toolkit R5 (DLDTK Version:1.0.19154 , AIXPRT CP (Community Preview) benchmark (https://www.principledtechnologies.com/benchmarkxprt/aixprt/) BS=64, Imagenet images, 1 instance/2 socket, Datatype: INT8 vs Tested by Intel as of 1/30/2019. 2 socket Intel® Xeon® Platinum 8180 processor, 28 cores HT On Turbo ON Total Memory 192 GB (12 slots/ 16GB/ 2633 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605, Linux-4.15.0-29-generic-x86_64-with-Ubuntu-18.04-bionic, Compiler: gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0, Deep Learning Deployment Toolkit (DLDT): OpenVINO™ toolkit R5 (DLDTK Version:1.0.19154), AIXPRT CP (Community Preview) benchmark (https://www.principledtechnologies.com/benchmarkxprt/aixprt/) BS=64, Imagenet images, 1 instance/2 socket, Datatype: FP32

产品和性能信息

1

在英特尔® 至强® 铂金 9282 处理器和英特尔® Deep Learning Boost(英特尔® DL Boost)上推断吞吐量高达 30 倍:由英特尔测试,截至 2019 年 2 月 26 日。平台:Dragon rock 2 插槽英特尔® 至强® 铂金 9282(每个插槽 56 个内核),HT 开启,睿频开启,总内存 768 GB(24 个插槽/32GB/2933 MHz),BIOS:SE5C620.86B.0D.01.0241.112020180249,CentOS 7 内核 3.10.0-957.5.1.el7.x86_64,深度学习框架:英特尔® Optimization for Caffe* 版本:https://github.com/intel/caffe d554cbf1,ICC 2019.2.187,MKL DNN 版本:v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a),模型,BS=64,无数据层综合数据:3x224x224,56 实例/2 插槽,数据类型:INT8 对比截至 2017 年 7 月 11 日英特尔进行的测试:2 个英特尔® 至强® 铂金 8180 处理器 CPU @ 2.50GHz(28 个内核),HT 已禁用,睿频已禁用,通过 intel_pstate 驱动程序将扩展调节器设置为“性能”,384GB DDR4-2666 ECC RAM。CentOS Linux* 版本 7.3.1611 (Core),Linux 内核 3.10.0-514.10.2.el7.x86_64。固态盘:英特尔® 数据中心级固态盘 S3700 系列(800GB,2.5 英寸 SATA 6Gb/秒,25 纳米,MLC)。性能评测标准基于:环境变量:KMP_AFFINITY='granularity=fine, compact‘,OMP_NUM_THREADS=56,CPU 频率设置为 CPU Power frequency-set -d 2.5G -u 3.8G -g performance Caffe:(http://github.com/intel/caffe/),修订版 f96b759f71b2281835f690af267158b82b150b5c。推断评测标准基于“caffe time --forward_only”命令,训练评测标准基于“caffe time”指令。对于“ConvNet”拓扑,使用合成数据集。对于其他拓扑,数据在本地存储,并且在训练之前在内存中缓存。拓扑规格来源于 https://github.com/intel/caffe/tree/master/models/intel_optimized_models (ResNet-50)。英特尔® C++ 编译器版本 17.0.2 20170213,英特尔® 数学核心函数库(英特尔® MKL)小型库版本 2018.0.20170425。使用“numactl -l”运行 Caffe。

2

性能结果基于截至配置中所示日期的测试,并且可能无法反映所有公开的安全更新。有关详细信息,请参见配置信息披露。没有任何产品或组件能够做到绝对安全。

3

性能测试中使用的软件和工作负载可能仅可基于英特尔® 微处理器进行性能优化。性能测试(如 SYSmark* 和 MobileMark*)使用特定计算机系统、组件、软件、操作和功能进行测量。对这些因素的任何更改可能导致不同的结果。您应该查询其他信息和性能测试以帮助您对正在考虑的采购作出全面的评估,包括该产品在与其他产品结合使用时的性能。如欲了解更多信息,请访问 www.intel.cn/benchmarks

4

优化通知:英特尔的编译器针对非英特尔微处理器的优化程度可能与英特尔® 微处理器相同(或不同)。这些优化包括英特尔® SIMD 流指令扩展 2(英特尔® SSE2),英特尔® SSE3,和追加 SIMD 流指令扩展 3 (SSSE3) 指令集和其它优化。对于在非英特尔制造的微处理器上进行的优化,英特尔不对相应的可用性、功能或有效性提供担保。此产品中依赖于处理器的优化仅适用于英特尔微处理器。某些不是专门面向英特尔® 微体系结构的优化保留专供英特尔微处理器使用。请参阅相应的产品用户和参考指南,以了解关于本通知涉及的特定指令集的更多信息。通知版本 #20110804。

5

OpenVINO 和 OpenVINO 标志是英特尔公司或其子公司在美国和/或其他国家(地区)的商标。