Single-chip programmable system system-on-chip RE3200

SoC

Product Overview
The RE3200 AI Vision Intelligent Computing Chip is a highly integrated, low‑power RISC‑V SoC designed for edge AI vision applications. It integrates CPU, dedicated AI acceleration, video and audio processing, security features and flexible I/O into a compact QFN40L package.

Key Advantages

• Highly integrated architecture combining CPU, AI accelerator, video, audio and security subsystems
• Heterogeneous multi‑core computing for parallel workloads and intelligent resource scheduling
• Configurable hardware acceleration modules to support customized development and model optimization

AI & Performance

• Integrated dedicated AI accelerator delivering up to 0.6 TOPS
• Supports mainstream deep‑learning models for real‑time vision inference
• Optimized power architecture delivering up to 300% improvement in compute efficiency
• Typical ultra‑low‑power operation for always‑on edge devices

Power Consumption

• Peak current: <45 mA @ 25°C
• Deep sleep mode (GPIO or timer wake‑up): <10 µA

Interfaces & Connectivity

• Rich peripheral support including MIPI‑CSI, HDMI, USB 3.0 and Ethernet
• Flexible I/O configuration to meet diverse system integration requirements

Reliability & Operating Conditions

• Ambient temperature: -40°C to 85°C
• Junction temperature: -40°C to 125°C
• ESD protection: 4 kV (HBM)
• Latch‑up immunity: 200 mA

Process & Packaging

• 55 nm low‑power process (1P7M)
• QFN40L package, 5 mm × 5 mm × 0.75 mm, 0.4 mm pitch

Power Supply

• Single external 3.3 V supply (range 3.0 V–3.6 V)
• Integrated LDOs: 3.3 V→1.2 V and 3.3 V→1.8 V

Target Applications

• Smart surveillance systems
• Intelligent transportation
• Industrial machine vision
• Smart home and IoT devices

Technical specifications

• Model: RE3200
• Architecture: RISC‑V based SoC with CPU, AI, video, audio and security subsystems
• AI accelerator: up to 0.6 TOPS
• Compute: heterogeneous multi‑core with configurable hardware accelerators
• Process: 55 nm low‑power (1P7M)
• Package: QFN40L, 5 mm × 5 mm × 0.75 mm, 0.4 mm pitch
• Power supply: external 3.3 V (3.0–3.6 V); integrated LDOs
• Peak current: <45 mA @ 25°C; Deep sleep <10 µA
• Peripherals: MIPI‑CSI, HDMI, USB 3.0, Ethernet; flexible I/O
• Operating temp: -40°C to 85°C; Junction temp: -40°C to 125°C
• ESD: 4 kV HBM; Latch‑up immunity: 200 mA
• Designed for always‑on low‑power edge vision applications and supports mainstream deep‑learning models