YY-AX650N-M.2 AI Computing Card

YY-AX650N-M.2
The YY-AX650N-M.2 is a low-power, mixed-precision high-performance AI computing card developed by YIYUE Tech, powered by Axera’s third-generation high-efficiency AI vision chip AX650N. This chip delivers an AI computing performance of 72 TOPS @ INT4 or 18 TOPS @ INT8, capable of running popular large-scale models such as DeepSeek. In addition to efficiently supporting CNN networks, it natively accommodates various large-scale model architectures like Transformer and Conformer.
Quantity
Contact Now

The NPU demonstrates exceptional performance. Using only one-third of its computing capacity, the card achieves the following results with common open-source AI models:

  • YOLOv5s: 130 FPS (7.66 ms) with 640x640 resolution input.
  • MobileNetV2: Over 1798 FPS (0.556 ms) with 224x224 resolution input.

The card supports a wide range of deep learning algorithms, enabling diverse application scenarios. It efficiently handles Transformer models, large vision models, and DeepSeek models. An SDK is provided to facilitate secondary development, allowing users to rapidly implement productization.

System Resources

Parameter Description

CPU

Octa-core Cortex-A55 1.7GHz, supports NEON acceleration, integrated FPU processing unit

NPU

Supports up to 72 TOPS @ INT4 or 18 TOPS @ INT8;
Supports mixed precision and multiple mainstream deep learning frameworks;

CODEC

Supports H.264/H.265 codec;
Maximum single-channel 8K@60fps decoding and 8K@30fps encoding;

RAM

16GB LPDDR4 (Configurable based on user requirements)

ROM

256MB NAND Flash

Operating System

Linux (Supports Ubuntu/Debian desktop systems)

Interface

M.2 interface

Physical Dimension

Parameter Description

Size (L*W*H)

22mm*80mm*5mm

Weight

50g

Contact us