Specifications
Product
Axelera Embedded 111c single board computer (Metis Compute Board)
Host Processor
Rockchip RK3588
Host CPU Cores
Dual quad-core Arm Cortex-A class clusters
AI Acceleration
Metis AIPU with four dedicated cores
AIPU Link
Four-lane PCIe connection from the RK3588 host to the AIPU
Host Memory
16 GB LPDDR4x dedicated to the CPU
AIPU Memory
4 GB or 16 GB LPDDR4x dedicated to the Metis AIPU
Storage
64 GB on-board eMMC
Storage Expansion
2 x SATA ports plus a microSD slot
Networking
2 x Gigabit LAN
USB
4 x USB 3.1 Gen1, 1 x USB Type-C, 3 x USB 2.0 headers
Video Output
HDMI 2.0 and DisplayPort up to 4096 x 2160 at 24 Hz
Expansion Slots
1 x M.2 E key and 1 x M.2 B key
Field I/O
GPIO, UART, ADC, I2C and SPI
Power Input
12 V DC input
Software Stack
Board Support Package plus the Axelera Voyager toolchain
Model Formats
TensorFlow, PyTorch and ONNX
Workload Fit
Multimodal edge applications with parallel or cascaded model pipelines
Architecture
RISC-V control plus Digital In-Memory Computing compute fabric, with an evolving model zoo
Role
Turnkey evaluation and deployment board — no separate accelerator integration required
Supply
Global B2B supply via QS Compute — quote on request
Overview
The Axelera Embedded 111c is the turnkey member of Axelera's Embedded line: instead of asking an integrator to fit an M.2 accelerator into their own carrier, the 111c puts the host and the accelerator on one board. A Rockchip RK3588 application processor supplies the general-purpose compute, and a quad-core Metis AIPU is attached over a four-lane PCIe connection dedicated to AI inference.
The memory split is the detail that matters for inference sizing. 16 GB of LPDDR4x is reserved for the CPU, while a separate 4 GB or 16 GB of LPDDR4x belongs to the Metis AIPU. Because the accelerator's working set lives in its own memory rather than competing with the host, the board can run multimodal pipelines — cascaded or parallel models — with predictable latency.
Integration surface is deliberately conventional: 64 GB of on-board eMMC plus two SATA ports and a microSD slot for storage, dual Gigabit LAN, four USB 3.1 Gen1 ports and a USB Type-C port, HDMI 2.0 and DisplayPort outputs up to 4K at 24 Hz, an M.2 E key and an M.2 B key for expansion, and GPIO, UART, ADC, I2C and SPI for field wiring. Power is a single 12 V DC input.
Software ships as a board support package plus Axelera's Voyager toolchain, so TensorFlow, PyTorch and ONNX models can be compiled and deployed without building a stack from scratch. QS Compute supplies the 111c for evaluation and quotes volume pricing for production edge AI builds.
Key Benefits
Host and accelerator on one board: removes carrier design work and integration risk. Separate memory domains: 16 GB host LPDDR4x plus 4 or 16 GB dedicated to the AIPU keeps inference latency predictable. Broad interface set: dual GbE, SATA, M.2, HDMI 2.0 and field I/O cover machine and vision integration. Voyager SDK included path: TensorFlow, PyTorch and ONNX models deploy without a hand-built toolchain.
Applications
Multimodal edge AI appliances, machine vision and inspection controllers, robotics and AMR perception nodes, smart retail and access-control systems, video analytics gateways, industrial inspection stations needing LLM or VLM inference locally, and rapid prototyping of Metis-based edge products before committing to a custom carrier.
Request a Quote — AXELERA EMBEDDED 111C — RK3588 SINGLE BOARD COMPUTER WITH INTEGRATED METIS AIPU
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →