BM1688
Intelligent Edge 16T Compute Platform
Product Brief
The BM1688 is a high-integration edge TPU processor designed to serve as a powerful 16T to 8T intelligent compute platform. It features an octa-core ARM Cortex-A53 architecture running at 1.6GHz, providing server-class compute power for edge applications.
Equipped with a self-developed TPU delivering up to 16 TOPS @ INT8 (and 32 TOPS @ INT4), it supports a wide range of frameworks including PyTorch, TensorFlow, and ONNX. It is specifically optimized for Large Language Models (LLM) and Vision-Language Models (VLM) at the edge.
The imaging system is highly advanced, capable of dual 8MP@30 HDR inputs or even 8K@15 SDR. It features hardware-level 360-degree stitching and fisheye dewarping, making it the perfect core for panoramic cameras and high-end security systems.
Connectivity is a major strength, with support for PCIe 3.0, SATA Gen3, USB 3.0, and dual Gigabit Ethernet. This allows for massive data throughput and storage expansion, ideal for smart micro-servers, AI NAS, and intelligent NVRs.
Built for reliability, the BM1688 supports a wide operating temperature range (-20°C to 60°C), enabling its deployment in various industrial and vehicle-based platforms, from autonomous robots to smart shipping systems.
Server-Class Edge AI
Delivers up to 16 TOPS INT8 compute power, enabling complex AI inference like LLM and multi-object tracking locally.
Octa-Core Compute
Eight Cortex-A53 cores provide robust multitasking capabilities for edge servers and robotics control.
Hardwired Panoramic Engine
Dedicated hardware blocks for 360° stitching and fisheye dewarping (Ceiling/Wall/Ground) reduce CPU load.
High-Speed Data Fabric
Flexible high-speed interfaces including PCIe 3.0 and SATA Gen3 for high-performance storage and expansion.
Advanced ISP Pipeline
Supports multi-sensor inputs (up to 6 channels) with 2f-HDR, 3DNR, and LDC for superior imaging clarity.
Global Framework Support
Seamlessly compatible with Caffe, PyTorch, TensorFlow, ONNX, and MXNet for rapid deployment.
System Architecture
AI Performance
High-performance TPU architecture with mixed-precision support for high-throughput AI workloads.
Benchmarks & Metrics
Commercial Deployment Scenarios
Smart Micro-Server
Smart NVR
AI NAS
Smart Lawn Mower
Robotics Platform
Vehicle/Ship Platform
Datasheet
| Processor | |
| CPU | 8-Core ARM Cortex-A53 @ 1.6GHz |
| AI | |
| TPU | 16 TOPS INT8 / 32 TOPS INT4 |
| Memory | |
| LPDDR4/4X | 2x32-bit (Up to 16GB Support) |
| Video | |
| ISP | 8K SDR / Dual 8MP HDR / 360° Stitching |
| Codec | H.264/H.265 (Decode 16x2MP, Encode 10x2MP) |
| Interface | |
| High Speed | PCIe 3.0, SATA Gen3, USB 3.0 |
| Network | 2x Gigabit Ethernet |
| Video In | 6-Channel MIPI CSI / subLVDS / HiSPI |
| Video Out | HDMI 2.0, MIPI DSI, LVDS, BT.1120 |
| Audio | |
| Codec | Integrated 16-bit (ADC/DAC/I2S) |
| Peripherals | |
| Control | CAN 2.0, UART x8, I2C x10, SPI x4, PWM x19 |
| System | |
| Environment | -20°C to 60°C Operating Temp |