News
IndustryTSMC’s grip on advanced chipmaking will face competitionIndustryCadence joins European Chip Design Platform to support semiconductor startups and SMEsMaking chipsMarc Hijink on ASML’s Culture, Supply Chain, and GeopoliticsProcessorsIntel's post-Nova Lake CPUs take shape as Razor Lake desktop and laptop variants surfaceAI chipsHow Google’s New AI Chip Software Helps It Battle NvidiaIndiaMore IC Packaging Plants Emerge In IndiaIndustryUS Tightens China AI Chip Access as Nvidia, TSMC and Semiconductor Stocks SlideDesignAxelera AI moves into HPC with TitaniaMaking chipsChinese envoy says chip equipment export curbs threaten EU-China trade thawIndiaSEMICON India 2026 Day 2 Sees 25 MoUs, Strategic Collaborations Across Semiconductor Value ChainIndustryTSMC’s grip on advanced chipmaking will face competitionIndustryCadence joins European Chip Design Platform to support semiconductor startups and SMEsMaking chipsMarc Hijink on ASML’s Culture, Supply Chain, and GeopoliticsProcessorsIntel's post-Nova Lake CPUs take shape as Razor Lake desktop and laptop variants surfaceAI chipsHow Google’s New AI Chip Software Helps It Battle NvidiaIndiaMore IC Packaging Plants Emerge In IndiaIndustryUS Tightens China AI Chip Access as Nvidia, TSMC and Semiconductor Stocks SlideDesignAxelera AI moves into HPC with TitaniaMaking chipsChinese envoy says chip equipment export curbs threaten EU-China trade thawIndiaSEMICON India 2026 Day 2 Sees 25 MoUs, Strategic Collaborations Across Semiconductor Value Chain
EveryChip
NVDLA

NVDLA

Open-source AI accelerator hardware by Nvidia

Nvidia created the NVIDIA Deep Learning Accelerator (NVDLA) in 2017 as an open-source neural network engine. Written in Verilog, it scales for various needs and powers edge devices like the Jetson Xavier NX for autonomous driving tasks.

Share on WhatsApp

  • Introduced by Nvidia in 2017 as open-source hardware
  • Written in Verilog for configurable scaling
  • Powers dual engines in Jetson Xavier NX
  • Delivers 14 TOPS under 10 W in Jetson NX
  • Used for object recognition in autonomous driving

What it is

The NVIDIA Deep Learning Accelerator, known as NVDLA, is a piece of open-source hardware designed to speed up artificial intelligence tasks. Nvidia introduced this neural network accelerator in 2017. It is built using Verilog, a language for describing electronic circuits. The design allows engineers to configure and scale the hardware to fit different architectural requirements. Unlike a full processor, NVDLA acts only as an accelerator. It requires an external entity, such as a central processing unit, to schedule tasks and manage access to resources. This setup lets it focus purely on heavy mathematical operations needed for deep learning.

How it came about

Nvidia launched NVDLA in 2017 to provide a standard, open hardware reference for neural network acceleration. By making the design available under an open-source model, the company aimed to help developers integrate efficient AI capabilities into their own products. This move aligned with broader industry trends towards open hardware, similar to Nvidia’s use of RISC-V processors in other GPU lines. The initiative allowed for wider adoption and customization without starting from scratch, fostering a community around shared accelerator designs.

Inside the chip

The core of NVDLA is written in Verilog, which makes it highly configurable. Engineers can adjust the design to meet specific performance or size constraints. It functions as a dedicated engine for neural network calculations, offloading work from the main CPU. Because it is just an accelerator, it does not handle task scheduling itself. A host processor must arbitrate processes and send instructions to the NVDLA engines. This separation ensures the accelerator remains efficient and focused on inference tasks, such as recognizing objects in video feeds.

How it was made

Specific manufacturing details for the standalone NVDLA IP are not provided, but its implementation appears in commercial products. For instance, the Jetson Xavier NX module integrates dual NVDLA engines. This small circuit board, roughly the size of a credit card, combines these accelerators with other components. The integration demonstrates how the open-source design translates into physical silicon within a larger system-on-chip environment, leveraging Nvidia’s existing fabrication partnerships for the final product.

Where it went

NVDLA targets edge computing applications where power efficiency and low latency are critical. It is prominently featured in the Nvidia Jetson Xavier NX platform. This device supports autonomous driving systems by handling object recognition tasks. The accelerator helps deliver high computational throughput for inference engines without consuming excessive energy. By enabling these capabilities on small form-factor boards, NVDLA supports devices that need to process data locally rather than sending it to a remote cloud server.

Versions and family

The primary variant discussed is the implementation within the Jetson Xavier NX. This module includes two NVDLA engines working alongside a six-core ARMv8.2 CPU and a Volta GPU with Tensor Cores. While the base NVDLA design is scalable, the Jetson version represents a specific configuration optimized for embedded use. Nvidia’s broader involvement in open hardware also includes RISC-V processors, showing a pattern of sharing architectural designs beyond just proprietary silicon.

Why it matters

NVDLA contributed to the growth of open-source hardware in the AI sector. By providing a verified, scalable accelerator design, Nvidia lowered the barrier for companies wanting to add neural network capabilities to their devices. Its inclusion in the Jetson series helped standardize high-performance edge AI development. The claim of delivering significant compute power under strict power limits demonstrated the viability of specialized accelerators for mobile and automotive markets.

Questions people ask about NVDLA

What is NVDLA?

NVDLA is an open-source neural network accelerator created by Nvidia. It is written in Verilog and designed to be configurable and scalable for various AI architecture needs.

Does NVDLA work alone?

No, NVDLA is only an accelerator. It requires an outside entity, such as a CPU, to schedule processes and arbitrate tasks before sending them to the accelerator.

Where is NVDLA used?

It is used in edge computing devices like the Nvidia Jetson Xavier NX. Applications include object recognition for autonomous driving and other inference engines requiring low power.

Good reading and sources

Where this page comes from

The facts on this page come from the Wikipedia article "NVDLA", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryChip rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.

It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

Photo credits on this page

The small pictures above come from Wikimedia Commons under free licences. Some are cropped to fit.

The weekly digest

Every Monday: the chip of the day, new pages on EveryChip, and the conferences and anniversaries of the week ahead.