PCIe AI accelerator card. Powered by quad-core Metis AIPU

Axelera® AI’s PCIe card, powered by a Quad-core Metis® AIPU, offers the highest performance inference acceleration on the market, combining ease of use, power efficiency, and scalability.

Now available in two configurations:

• 4 GB memory
• 16 GB memory

Key Benefits:

• The highest-performance AI-accelerator PCIe card in the market for edge AI applications. Powered by a quad-core Metis AIPU.
• A single board can run inference on dozens of cameras as well as support multiple parallel neural networks.
• A wide range of end-to-end AI pipelines and models are available out of the box.
• Hassle free evaluation and SW integration thanks to Voyager® SDK.
• Uncompromised prediction accuracy thanks to advanced quantization tools.

System Requirements

Metis PCIe Card has the following system integration requirements. Our team actively tests integration with different systems from vendors such as Dell, Lenovo, Advantech and Aetina.

• Host Interface: PCIe Gen. 3.0 x4
• Mechanical form factor: PCIe half-height, half-length, single slot (HH, HL, SS)
• Power Rating: PCI-SIG compliant power slot, 75W is recommended. Refer to Typical Power specs in the Metis PCIe 1 AIPU Datasheet for typical average power consumed when running inference.
• Host CPU: Intel® Core™ Processors and Xeon(R) Processors, Ryzen™ Processors, Arm64 (aarch64) based processors.
• Host Operating System:
  • Native Linux support is tested on Ubuntu 22.04 and Ubuntu 24.04
  • A Docker Guide is available to evaluate or develop with other Linux distributions.
  • Native Windows support for inference is available on Windows 10/11 and Windows Server 2025. SDK development is only supported on Linux.

Download the PCIe Product Brief here!

·