Linux NPU Support

This article will teach you how to run LLMs on your AMD XDNA 2 NPU on Linux using FastFlowLM.
Get set up and then show us what you build!

Date: March 5, 2026
Authors: Lemonade-server🍋 and FastFlowLM contributors

📢 FastFlowLM Linux Support

FastFlowLM is a lightweight LLM runtime optimized for AMD NPUs.
Today, FastFlowLM is adding support for Ubuntu, Arch, and other distros to enable fast, low-power LLMs on Ryzen™ AI PCs that run Linux.

This article will help you:

• Understand Linux NPU support status and required platform versions
• Install the FLM + driver stack for your distribution
• Validate your setup with flm validate
• Fix common firmware, driver, and memlock issues

⚙️ Hardware Requirements

Supported processors

FastFlowLM on Linux requires an AMD XDNA 2 NPU.

Note: Ryzen AI 7000 / 8000 / 200-series chips have XDNA 1, which is not supported.

🧰 Software Requirements

Runtime stack

The NPU requires specific firmware, kernel version, driver, and runtime software to function.
The quickstart guide below will help you install these requirements.

🚀 Quickstart

Supported Distributions

• Ubuntu 24.04 LTS
• Ubuntu 25.10
• Ubuntu 26.04
• Arch Linux
• Other (Generic Linux)

1. Prerequisites

• amdxdna driver (included in kernel 7.0+, or via amdxdna-dkms)
• NPU firmware version 1.1.0.0 or later
• Python 3.8+
• XRT stack from AMD

2. System Preparation

Ubuntu (24.04, 25.10)

1. Add the AMD XRT PPA (Required for NPU/XDNA)

The AMD XRT stack is a prerequisite for NPU support. Add AMD’s PPA:

sudo add-apt-repository ppa:lemonade-team/stable
sudo apt update

See lemonade-team/stable PPA for details.

2. Install XRT and NPU Drivers

sudo apt install libxrt-npu2 amdxdna-dkms

3. Reboot

sudo reboot

4. Install FastFlowLM

• Download the latest .deb package from the Releases page:

sudo apt install ./fastflowlm*.deb

5. (NPU) Check memlock limit

• Run:

   ulimit -l
  

• If not unlimited, add to /etc/security/limits.conf:

  * soft memlock unlimited
* hard memlock unlimited

• Reboot system

Ubuntu 26.04, Arch, and Others

For Ubuntu 26.04 and other distributions, check this Linux NPU setup guide.

Arch Linux

Arch users need the kernel driver, matching kernel headers, XRT, and the AMD XDNA XRT plugin:

sudo pacman -Syu
sudo pacman -S linux-headers linux-firmware-other xrt xrt-plugin-amdxdna

Install amdxdna-dkms from the AUR using your preferred AUR workflow, then reboot. If needed, rebuild the DKMS module for the running kernel:

sudo dkms autoinstall -k "$(uname -r)"
sudo depmod -a
sudo reboot

If you need to rebuild a specific DKMS version, check dkms status and use that version explicitly.

After rebooting, confirm the DKMS module is selected:

modinfo -F filename amdxdna

The path should contain updates/dkms. If it points under kernel/drivers/accel/amdxdna/, the stock in-tree driver is still being used.

Then confirm XRT can see the NPU:

xrt-smi examine

If flm validate passes but flm run fails with No such device with index '0', XRT does not see a device. Make sure xrt-plugin-amdxdna is installed and xrt-smi examine lists the NPU.

Arch firmware note: Some linux-firmware-other versions ship both 1.0 and 1.1 NPU firmware for 17f0_10. On stock Linux 6.19, forcing npu.sbin.zst to 1.1 firmware can make the NPU disappear because the in-tree driver expects the older firmware protocol. Use amdxdna-dkms or a newer kernel that supports the protocol-7 firmware path, then verify flm validate reports firmware 1.1.x.

Building from Source

1. Clone the repository:

   git clone https://github.com/FastFlowLM/FastFlowLM.git
   cd FastFlowLM
   

2. Build:

   cd src
   cmake --preset linux-default
   cmake --build --preset linux-default -j$(nproc)
   cmake --install --preset linux-default
   

4. Validating NPU Setup

To validate your NPU setup, run:

flm validate

You should see output similar to:

[Linux]  Kernel: 7.0.0-rc1-00052-g27936bfca73d
[Linux]  NPU: /dev/accel/accel0
[Linux]  NPU FW Version: 1.1.2.64
[Linux]  Memlock Limit: infinity

On Linux, flm validate checks the kernel DRM path. flm run uses XRT. If validation succeeds but running a model fails with No such device with index '0', run xrt-smi examine and install the XRT AMD XDNA plugin for your distribution.

📚 Additional resources

• Lemonade-server🍋
• Lemonade GitHub issues
• Lemonade Discord