Neural Magic LogoNeural Magic Logo
Products
DeepSparse Engine
Sparsity-aware neural network inference engine for GPU-class performance on CPUs.
SparseML
Open-source libraries for applying sparsification recipes to neural networks with a few lines of code.
SparseZoo
Open-source model repository for sparse and sparse-quantized models.
Use Cases
Natural Language Processing (NLP)
Get started with Sparse NLP use cases. Apply your data and redeploy on CPUs with ease.
Compute Vision
Get started with sparse computer vision use cases. Apply your data and deploy on CPUs with ease.
Community
Deep Sparse Slack Community
Engage, discuss, and get direct access to our engineering teams.
Code
Explore our code on GitHub.
Docs
Get support or get started.
Labs
Labs by Neural Magic
Partner with experts to ensure your ML training and deployment pipelines are optimized with SOTA research and industry best practices.
Resources
Blogs & News
Stay current & discover SOTA performance techniques.
Research Papers
Explore our published research papers that are driving ML simplicity & performance.
Upcoming Events
Explore upcoming events Neural Magic will be attending. Come say hello!
On-Demand Videos
Listen to on-demand ML performance content on your time.
Our Technology
Learn about the "magic" behind Neural Magic!
Subscribe
Sing up for the regular Neural Magic email updates.
Company
Team & Investors
Explore our team and our investors.
Careers
Discover open positions. Apply or refer a friend.
Contact
Reach out to us directly.
Legal
View Neural Magic's legal documentation.
menu-icon
Products
DeepSparse EngineSparseMLSparseZoo
Use Cases
Natural Language Processing (NLP)Compute Vision
Community
Deep Sparse Slack CommunityCodeDocs
Labs
Labs by Neural Magic
Resources
Blogs & NewsResearch PapersUpcoming EventsOn-Demand VideosOur TechnologySubscribe
Company
Team & InvestorsCareersContactLegal
Home
Get Started
Use Cases
User Guides
Products
Details
User Guides
DeepSparse
Supported Hardware
github
Edit on GitHub
slack
Ask on Slack

Supported Hardware for DeepSparse

With support for AVX2, AVX-512, and VNNI instruction sets, DeepSparse is validated to work on x86 Intel (Haswell generation and later) and AMD (Zen 2 and later) CPUs running Linux. Mac and Windows require running Linux in a Docker or virtual machine.

Here is a table detailing specific support for some algorithms over different microarchitectures:

x86 ExtensionMicroarchitecturesKernel SparsitySparse Quantization
AMD AVX2Zen 2, Zen 3optimizedemulated
AMD AVX-512 VNNIZen 4optimizedoptimized
Intel AVX2Haswell, Broadwell, and neweroptimizedemulated
Intel AVX-512Skylake, Cannon Lake, and neweroptimizedemulated
Intel AVX-512 VNNI (DL Boost)Cascade Lake, Ice Lake, Cooper Lake, Tiger Lakeoptimizedoptimized
User Guides for DeepSparse Engine
Inference Types with DeepSparse Scheduler