On April 1, 2025, Dalian University Holdings, a leading international semiconductor component distributor dedicated to the Asia Pacific market, announced that its subsidiary, Shiping, has launched an edge AI multi-channel physical inspection solution based on the MemryX MX3 AI inference acceleration card and Rockchip Orange Pi 5 Plus module (RK3588).

With the rapid development of technologies such as the Internet of Things, big data, and cloud computing, edge AI, as an important branch of artificial intelligence, is gradually becoming a key force in promoting the popularization and deepening of intelligent applications. Edge AI deploys AI algorithms directly to edge devices for processing and inference near the data generation source, greatly improving the real-time and efficiency of data processing while reducing data transmission latency and costs. Taking advantage of the advantages of edge AI, Da Da Da Da Da Da Shi Ping has launched an edge AI multi-channel object detection solution based on MemryX MX3 AI inference acceleration card and Ruixin Micro Orange Pi 5 Plus module (RK3588), aiming to achieve efficient, low latency, and cost-effective intelligent monitoring and object recognition applications, meeting the needs of diverse scenarios.

MemryX is a semiconductor company focused on edge AI processing solutions, dedicated to providing high-performance, low-power AI processing capabilities for various industrial and consumer application scenarios. The MX3 AI inference acceleration card equipped in this solution encapsulates four MX3 AI chips onto an M.2 2280 expansion card, which can be easily integrated into systems equipped with PCIe 3.0 M.2 slots. The MX3 AI chip can provide a computing power performance of 5 TOPS per watt and supports Brain Floating Point operations to ensure module accuracy for users. Each chip is equipped with 10.5MB of static random access memory for accessing modules, which does not occupy the resources of the main system. Up to 16 chips can be connected in series to expand performance. With the performance of the MX3 AI chip, the MX3 AI acceleration card has excellent computing performance of up to 20 TOPS, which can bring a plug and play convenient experience to various industrial computers.

With the powerful computing capability of MemryX, this solution can easily achieve multi-channel object detection applications. Developers can easily achieve various monitoring functions such as parking lot management system, intelligent parking pillars, intelligent traffic monitoring, mall pedestrian flow monitoring, and comprehensive home accident detection by simply using a regular USB camera or connecting through a network. In addition, to accelerate the R&D process, MemryX also provides a Developer Hub development environment, which enables developers to easily convert modules from mainstream deep learning frameworks such as TensorFlow Lite, ONNX, PyTorch, Keras, etc. into the DFP framework required for MX3 AI chips, thereby quickly completing AI application designs.

Core technological advantages:

Using floating-point BF16 for calculation to ensure module accuracy:

The module design is based on BF16 (Brain Floating Point 16) for computation. Compared to traditional floating-point formats, BF16 can significantly reduce memory usage while providing calculation accuracy close to FP32. This makes it particularly suitable for inference and training scenarios in artificial intelligence and deep learning models, ensuring the accuracy of the results.

Not occupying system memory:

The module adopts an independent memory architecture during runtime, which does not require the use of RAM resources in the main system, effectively reducing the impact on the overall performance of the system. This design feature ensures that the module can run efficiently while still reserving sufficient system resources for other applications.

Highly scalable:

Supports connecting up to 16 modules and achieves high scalability through modular design. This enables the system to flexibly expand its computing power according to requirements to meet the computing requirements of different scenarios, such as data centers or edge computing that need higher performance.

Best data flow optimization to minimize data movement:

The module has been highly optimized for data flow internally, and through intelligent routing and caching mechanisms, it can minimize the frequency of data movement during operation, thereby improving processing performance and reducing latency. In addition, such a design also helps to reduce energy consumption and further enhance the operational efficiency of the system.

High cost-effectiveness and low-power solution:

By pairing the main platform Orange Pi 5 Plus with the MemryX MX3 AI chip, it can be upgraded to a higher-level AI platform that can run object detection at approximately 480 frames per second (YOLOv8); And MX3 has a performance of 5 TOPS/W, and the entire multi-channel object detection solution only consumes about 14W of power.

New concept of multi-channel application:

Compared with edge computing, which has emerged in recent years, applying its concept to regional scenarios may be a novel solution that can significantly reduce costs. By using an easily accessible camera and pairing it with an intelligent industrial host, many applications can be implemented, and the front-end camera can be easily replaced and configured.

Scheme specifications:

The main platform development board is based on the Ruixin Micro RK3588 platform, equipped with four Cortex-A76 processors and four Cortex-A55 processors, and provides powerful core architectures such as high-performance image processor Arm Mali-G610 and neural computing processor NPU.

The I/O Board development board provides powerful peripheral configurations such as Gigabit Ethernet, USB Type A/C 3.0 Universal Serial Bus interface, HDMI high-definition multimedia interface, M.2 E-Key transmission interface, M.2 M-Key transmission interface, and can simulate commonly used UART, I2C, SPI, CAN and other signals through an extended 40 pin pin configuration.

The MemryX MX3 AI chip provides powerful AI computing capabilities (20 TOPS), mainly based on PCIe Gen3 M.2 2280 M-Key interface. Its M.2 acceleration card is equipped with four MX3 AI chips, each chip can provide 5 TOPS/W performance, and has a built-in 10.5MB static random access memory for accessing the module. Supports both Linux and Windows operating systems, and provides rich software resources for developers to use. It can directly port popular deep learning frameworks such as Tensorflow, ONNX, Pytorch, Keras, etc.

Regarding Dahua University Holdings:

Da Da Da Da Da Holdings is an internationally leading semiconductor component distributor dedicated to the Asia Pacific market. It was founded in 2005 and celebrates its 20th anniversary this year. Headquartered in Taipei, it has subsidiaries such as Shiping, Pinjia, Quanding, and Youshang, with approximately 5000 employees and over 250 authorized product suppliers. It has 69 service centers worldwide and achieved a revenue of NT $8805.5 billion in 2024. Dahua University has established an industrial holding platform, focusing on international operational scale and localized flexibility, and has long been deeply rooted in the Asia Pacific market. With the vision of "industry first choice - channel benchmark", it fully implements the core values of "team, integrity, professionalism, and efficiency", and has won the "Global Distributor Excellence Performance Award" for 24 consecutive years. At the same time, it continues to strengthen ESG sustainable development and has won international recognition for 3 consecutive years, with an MSCI ESG rating of A-level. Faced with the trend of new manufacturing, Dalian University is committed to transforming into a data-driven enterprise, establishing an online digital platform - "DaDaWang", and advocating the intelligent logistics service (LaaS, Logistics as a Service) model to assist customers in facing the challenges of intelligent manufacturing together. Starting from benevolence, building trust through technology, and adhering to the corporate purpose of "co creating partner value and achieving the future", Da Da Da Da Da hopes to build a highly competitive ecosystem with the industry "La Bang Jie Pai".

\This is reported by Top Components, a leading supplier of electronic components in the semiconductor industry

They are committed to providing customers around the world with the most necessary, outdated, licensed, and hard-to-find parts.

Media Relations

Name: John Chen

Email: salesdept@topcomponents.ru