The industry is shifting from connected devices to intelligent terminals. Instead of relying on the cloud, BK7259 enables real-time autonomy through a unified silicon architecture that integrates compute, connectivity, AI acceleration, and multimedia processing on a single die.

 

By consolidating multiple subsystems into one platform, developers can reduce system complexity, save PCB space, and accelerate time-to-market — replacing multi-chip designs with a cohesive Edge AI solution.

​

Heterogeneous Edge AI Compute

​

BK7259 integrates a powerful heterogeneous compute subsystem designed for embodied AI applications:

​

• Arm Cortex-M52 and Cortex-M55 application processors running up to 480 MHz

• Integrated Arm Ethos-U65 NPU delivering up to ~0.25 TOPS AI acceleration

• Dual RISC wireless cores enabling concurrent connectivity and application workloads

• Helium vector extensions for DSP-level performance

 

This architecture allows devices to process vision, voice, and control tasks simultaneously — enabling cognitive reflex behavior without sacrificing efficiency.

​

Unified Connectivity and Wireless Performance

​

BK7259 combines advanced wireless technologies into a single integrated solution:

​

• Wi-Fi 6 (802.11 b/g/n/ax) 1x1 2.4 GHz

• Dual-mode Bluetooth 6.0

• IEEE 802.15.4 / Thread support

• Proprietary 2.4 GHz multi-protocol capability

 

With strong sensitivity (-98 dBm) and optimized low-power operation (80 μA DTIM10), BK7259 delivers reliable connectivity for always-on intelligent devices.

​

Graphics, Vision, and Multimedia Acceleration

​

BK7259 goes beyond traditional MCU architectures by integrating a full multimedia pipeline:

​

• 2.5D GPU for graphics acceleration

• Video processing unit for advanced codecs

• Display processor for rich UI rendering

• MIPI CSI-2 camera interface and RGB display support

 

This allows developers to build devices with high-quality visual interfaces and local AI perception without external processors.

​

Local-First Security and Privacy

​

Edge AI devices must be intelligent — but also secure. BK7259 integrates hardware-level protection:

​

• Arm TrustZone secure architecture

• Dedicated secure element for cryptography

• Local processing for sensitive vision and voice data

 

By keeping inference and analytics on-device, BK7259 enables privacy-first experiences where data never needs to leave the terminal.

​

Performance Designed for Edge AI

​

Typical inference performance:

​

• MobileNet V1 — ~79 inferences/sec

• MobileNet V2 — ~78 inferences/sec

• Tiny YOLO V2 — ~42 inferences/sec

​​

This level of performance enables responsive perception for robotics, smart cameras, AI toys, and HMI systems.

 

Example Applications

​

BK7259 is optimized for intelligent terminals across multiple industries:

• AI toys and companion devices

• Smart locks and visual doorbells

• Touchless HMI and control panels

• Always-on video IPC

• AI glasses and wearables

• Industrial predictive maintenance systems

​​

The unified architecture allows developers to build products that operate locally, respond instantly, and remain reliable even without cloud connectivity.

​

Why BK7259

​

Traditional architectures required separate MCUs, DSPs, AI accelerators, and wireless chips. BK7259 consolidates these components into a single cognitive hub — enabling developers to stop fighting hardware complexity and focus on building intelligent experiences.