MYiR MAC-B5760 — Fanless RK3576 Edge AI Industrial PC with Optional RK1828 LLM/VLM Module

Specifications

Overview

The MYiR MAC-B5760 is a fanless Edge AI industrial box PC built on the Rockchip RK3576 SoC, which integrates a 6 TOPS NPU. The system ships with 8GB of LPDDR5 and 64GB of onboard flash as standard, and is aimed at industrial vision, gateways and on-premise AI workloads that need a sealed, service-free enclosure.

Networking covers two Gigabit Ethernet ports, with four USB 3.0 ports plus a USB-C port for convenient serial console access. Video is driven over HDMI and mini DisplayPort, audio through a 3.5mm jack, and an optional Wi-Fi 6 / Bluetooth 5.4 module adds wireless connectivity. Expansion is split across an M.2 socket that accepts either an NVMe SSD or an AI accelerator, and a mini PCIe socket for 4G LTE or 5G cellular modems.

The most distinctive option is the RM1828MC0-F M.2 AI accelerator card built on the Rockchip RK1828 chipset, rated at 20 TOPS and specified for computer vision alongside large language and vision-language models supporting up to 7 billion parameters. That combination lets the same industrial chassis handle classic machine vision and local generative AI inference without changing platform.

Software is a Linux BSP based on kernel 6.1 from the Rockchip SDK with open-source drivers, and MYiR pre-builds three images: Debian, Yocto Linux and a Yocto Linux build with the Preempt RT kernel for deterministic control workloads. Two configurations are sold: the standard MAC-B5760-CC810000 at $299 with 8GB RAM and 64GB eMMC, and the MAC-B5760-CC81000C at $599 which adds the M.2 RK1828 AI module. Each ships with a quick start guide, a Wi-Fi/Bluetooth antenna and a 12V 5A power supply.

Key Benefits

Dual-path AI scaling — a built-in 6 TOPS NPU plus an optional 20 TOPS RK1828 M.2 card for 7B-parameter LLM and VLM inference. Fanless, sealed operation suited to dusty and vibration-prone industrial environments. Real-time option — a Preempt RT Yocto image is provided for deterministic control applications. Dual Gigabit Ethernet with an M.2 or mini PCIe path for NVMe, AI acceleration, or 4G/5G. Low entry cost at $299 with a clear upgrade to $599 for generative AI capability. Open-source BSP on kernel 6.1 for straightforward integration.

Applications

Industrial machine vision and inspection, on-premise LLM and VLM assistants, AI edge gateways and protocol conversion, autonomous equipment with RT control, retail and building analytics, and multi-camera video analytics where a fanless chassis is required.

Related Products