DL Boost

E653484

Intel technology deep learning acceleration technology

DL Boost is Intel’s deep learning acceleration technology that enhances AI inference performance on its processors through specialized instruction set extensions.

Jump to: Statements Referenced by

Statements (40)

Predicate	Object
instanceOf	Intel technology ⓘ deep learning acceleration technology ⓘ
appliesTo	AI inference ⓘ deep neural networks ⓘ
architectureLevel	x86 instruction set architecture extension ⓘ
basedOn	specialized instruction set extensions ⓘ
benefit	higher performance per watt for AI inference ⓘ reduced need for discrete accelerators for some AI workloads ⓘ
category	AI hardware acceleration feature ⓘ CPU instruction set extension ⓘ
compatibleWith	Intel OpenVINO toolkit NERFINISHED ⓘ Intel oneDNN NERFINISHED ⓘ
designedFor	data center workloads ⓘ edge AI workloads ⓘ
developer	Intel NERFINISHED ⓘ
enables	faster matrix multiply-accumulate operations ⓘ higher operations per cycle for AI kernels ⓘ
focus	inference rather than training ⓘ
fullName	Intel Deep Learning Boost NERFINISHED ⓘ
improves	latency of AI inference ⓘ performance of image recognition models ⓘ performance of natural language processing inference ⓘ throughput of AI workloads ⓘ
introducedBy	Intel Xeon Scalable processors NERFINISHED ⓘ
introducedIn	Intel Xeon Scalable 2nd Generation NERFINISHED ⓘ
marketingNameFor	Intel VNNI features NERFINISHED ⓘ
optimizationFor	INT8 inference ⓘ low-precision arithmetic ⓘ
purpose	accelerate deep learning workloads ⓘ enhance AI inference performance ⓘ
supports	convolutional neural networks ⓘ fully connected layers ⓘ quantized neural network operations ⓘ recurrent neural networks ⓘ
targetHardware	Intel Core processors NERFINISHED ⓘ Intel Xeon processors NERFINISHED ⓘ
uses	AVX-512 extensions ⓘ VNNI NERFINISHED ⓘ vector neural network instructions ⓘ
vendor	Intel NERFINISHED ⓘ

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

Tiger Lake microarchitecture → supports → DL Boost ⓘ

subject surface form: Tiger Lake