AI Accelerator IP

Our proprietary AI accelerator IP is optimized for hardware acceleration of mainstream neural networks including CNN and Transformer models. With a configurable compute array architecture, it delivers efficient AI inference within tight power and area budgets, suitable for embedded AI, smart cameras, and ADAS edge computing.

Features

• CNN/Transformer Acceleration
• INT8/INT16 Quantized Inference
• Configurable Compute Array
• Low-power Edge Deployment
• Mainstream AI Framework Support