BM1684X

Key features

SOPHGO high-performance AI inference processor for edge AI servers, video analytics, smart security and industrial vision
Provides higher TPU inference performance than BM1684 for larger models and higher-concurrency vision workloads
Supports multi-precision inference such as INT8, FP16 and FP32 for performance, accuracy and deployment flexibility
Supports multi-channel video codec, image preprocessing and AI inference pipelines
Works with SOPHON SDK and BModel deployment flows for migration from mainstream training frameworks to edge inference

Specification	Details
AI	• Integrated high-performance SOPHGO TPU• Supports INT8 / FP16 / FP32 inference workloads• Suitable for high-concurrency CV inference and edge AI services
Video	• Supports multi-channel video decode and encode applications• Channel count depends on card, module and software configuration
Image Processing	• Supports image preprocessing, scaling, color conversion and AI pipeline acceleration
Host Interface	• Supports PCIe accelerator card, module and embedded product designs
Memory	• Supports high-bandwidth memory configuration for AI inference• Capacity depends on product design
Software	• Supports SOPHON SDK• Supports model conversion, BModel compilation, runtime inference and multimedia pipeline development