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
XDNA

XDNA

AMD microarchitecture for neural processing units in Ryzen AI chips

XDNA is a microarchitecture for neural processing units developed by AMD. It handles machine learning tasks on Ryzen AI processors, using spatial dataflow to improve power efficiency and performance for artificial intelligence workloads.

Share on WhatsApp

  • Uses spatial dataflow architecture for parallel processing
  • First generation launched in April 2023 with Ryzen 7040
  • Achieves up to 1.3 GHz operating frequency
  • Integrates VLIW vector and scalar RISC processors
  • Successor XDNA 2 reaches up to 55 TOPS

What it is

XDNA is a microarchitecture designed by AMD for neural processing units. It targets deep learning and artificial intelligence tasks within consumer and mobile processors. The design integrates into Ryzen AI branded chips, working alongside Zen central processing units and RDNA graphics cores. This architecture focuses on efficient handling of machine learning software through specialized hardware tiles.

How it came about

AMD developed this microarchitecture after acquiring Xilinx in 2022. The technology builds upon Xilinx foundations to create dedicated AI accelerators. The first generation launched in April 2023 with the Ryzen 7040 series. Subsequent updates appeared in the Ryzen 8040 Hawk Point line. A second generation, known as XDNA 2, debuted later with significantly higher throughput capabilities.

Inside the chip

The core uses a spatial dataflow layout with parallel AI Engine tiles. Each tile combines a VLIW and SIMD vector processor for heavy math with a scalar RISC controller for logic. Local memory stores weights and data to cut external access. Dedicated DMA engines move data between tiles efficiently. This structure reduces latency and power use by keeping operations on-chip rather than relying on main system memory.

How it was made

Specific manufacturing details such as the foundry or process node are not provided in the available records. The design relies on modular tile arrays that scale to fit different power and area constraints. Operating frequencies reach up to 1.3 GHz, adjusted for thermal limits. The silicon integrates directly into system-on-chip packages for laptops and mobile devices.

Where it went

These processors appear in Ryzen AI branded laptops and portable workstations. They support local AI inference for applications like Windows Copilot features. Developers use ROCm and Vitis AI stacks to accelerate workloads in frameworks such as TensorFlow and PyTorch. The hardware enables real-time processing for edge devices without needing cloud connectivity for every task.

Versions and family

The first generation delivered up to 10 TOPS in mobile forms. A refresh called Hawk Point raised this to roughly 16 TOPS via firmware and clock tweaks. XDNA 2 followed with up to 55 TOPS on flagship models based on Zen 5. Entry-level variants may disable some functions to save die area and power. The successor architecture continues the lineage with greater capacity.

Why it matters

This architecture brings dedicated AI acceleration to mainstream personal computers. It allows thin devices to run complex models locally with better battery life than GPU-only approaches. The modular design supports scaling from lightweight tablets to more powerful workstations. Integration with major software tools helps developers adopt the hardware for diverse machine learning applications.

Problems and reception

Performance figures represent theoretical maximums and vary with thermal conditions. Actual speed depends on driver optimization and specific workload types. Some lower-end chips limit functionality to conserve resources. The software ecosystem continues to evolve as tools mature to fully exploit the hardware capabilities for consistent real-time inference.

Questions people ask about XDNA

What is the XDNA microarchitecture?

XDNA is a design by AMD for neural processing units. It handles machine learning tasks in Ryzen AI processors. The architecture uses spatial dataflow to process data in parallel, improving efficiency for artificial intelligence workloads on personal computers.

How does XDNA improve power efficiency?

It uses on-chip local memory to store weights and data. This reduces the need to access external DRAM, which lowers latency and power consumption. Dedicated DMA engines also help move data efficiently between processing tiles without stalling.

Which processors use XDNA technology?

The first generation appeared in Ryzen 7040 Phoenix series chips. Later versions were included in Ryzen 8040 Hawk Point and Ryzen AI 300 series processors. These chips combine the NPU with Zen CPU and RDNA GPU cores on the same silicon.

Good reading and sources

Where this page comes from

The facts on this page come from the Wikipedia article "AMD XDNA", 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 2026-11-15.

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.