Kneron KL520 AI SoC

Kneron KL520 AI (Artificial Intelligence) SoC (System on a Chip) is designed to provide smart-home and IoT applications with on-device Edge AI and Machine Learning. The KL520 combines proprietary software and hardware designs to create a highly efficient and ultra-low-power Neural Processing Unit (NPU). Running AI computations on the end device helps generate real-time insights without relying on the Cloud. The KL520's reconfigurable architecture allows it to run several different Convolutional Neural Networks (CNNs) based on different applications, regardless of its kernel size, architecture requirements, or input size. This feature enables the KL520 to achieve a very high MAC (Media Access Control) efficiency, a key metric for efficient computation.

The design of the Kneron KL520 allows the chip to dynamically adjust its function based on the application's needs to achieve maximum efficiency. The KL520 is also equipped with a proprietary compression technology that can compress a large AI neural network into a very small footprint to process on the Edge device, without sacrificing large amounts of power.

Features

NPU
- 300MHz maximum frequency
- 345 GOPS, 576MAC/cycle peak throughput of 8-bit mode:
CPU
- Arm® Cortex®-M4 @ 200MHz for system control
- Arm Cortex-M4 @ 250MHz as AI coprocessor
Memory
- 32MB or 64MB SIP SRAM, 16-bit LPDDR2-1066
- Up to 64MB SPI NOR Flash
OS Support
- CMSIS RTX
Power
- 380mW average power consumption
- 1.1V core voltage
- 3.3V I/O voltage
Video in interface
- 2x MIPI-CSI-2 RX ports with 2-lanes each
- DPI