This articlemay rely excessively on sourcestoo closely associated with the subject, potentially preventing the article from beingverifiable andneutral. Please helpimprove it by replacing them with more appropriatecitations toreliable, independent sources.(March 2021) (Learn how and when to remove this message)

Intel'sDeep Learning Boost (DL Boost) is a marketing name forinstruction set architecture (ISA) features on thex86-64 designed to improve performance ondeep learning tasks such as training and inference.^[1]

DL Boost consists of two sets of features:

• AVX-512 VNNI, 4VNNIW, orAVX-VNNI: fast multiply-accumulation mainly forconvolutional neural networks.
• AVX-512 BF16: lower-precisionbfloat16 floating-point numbers for generally faster computation. Operations provided include conversion to/fromfloat32 anddot product.

DL Boost features were introduced in theCascade Lake architecture.

ATensorFlow-based benchmark run on theGoogle Cloud PlatformCompute Engine shows improved performance and reduced cost compared to previous CPUs and to GPUs, especially for small batch sizes.^[2]

• Deep Learning Boost at Intel
• Andres Rodrigues et al., "Lower Numerical Precision Deep Learning Inference and Training", Intel White paper[3]
• Intel and ML (2017), from Intel's Developer Relations Division

• SIMD computing
• Deep learning
• X86 architecture

• Articles with short description
• Short description is different from Wikidata
• Articles lacking reliable references from March 2021
• All articles lacking reliable references