Aixin Yuanzhi was selected as the preferred partner of Xuantie in 2024: developing AI computing and joining hands with RISC-V to reshape thousands of industries

China March 18, 2024 - Aixin Yuanzhi, an artificial intelligence chip R&D and basic computing platform company, announced that the second Xuantie RISC-V Ecological Conference with the theme of "Open·Connection" will be held on March 14 in Held in Shenzhen , Liu Jianwei, co-founder and vice president of Aixin Yuanzhi, was invited to deliver a keynote speech and shared with the guests the Aixin Tongyuan mixed-precision NPU, an AI processor oriented to edge-end computing power layout, and RISC-based The AI ​​computing power improvement and industry implementation exploration carried out by V Ecosystem.

At the same time, the conference also announced the "Xuantie Preferred Partners", and 8 companies including Aixin Yuanzhi were selected. It is reported that this honor is intended to promote the implementation of Xuantie’s ecological achievements, connect the RISC-V industrial chain ecological network, and promote the development of Xuantie’s business ecology. This selection represents the recognition of Aixin Yuanzhi’s contribution to the Xuantie RISC-V ecosystem.

AI is everywhere, RISC-V adds power to AI computing

From large cloud models represented by ChatGPT to computing boxes on the edge and cameras scattered on the end, AI has become ubiquitous with the rapid development of technology. AI has begun to play a role in multiple end-side business scenarios such as smart driving, smart robots, smart factories, smart cities, and smart retail, bringing productivity improvements to all walks of life.

As a platform chip company that supports edge-side and terminal-side intelligence, Aixin Yuanzhi takes "inclusive AI to create a better life" as its corporate mission and focuses on the research and development of high-performance, low-power artificial intelligence vision processor chips. Liu Jianwei introduced that Aixin Yuanzhi’s series of chip products have been widely used in the two major product lines of AIoT and ADAS and achieved steady development, promoting the deployment of AI on the edge and end sides.

However, the implementation of AI on the edge side also means higher computing power requirements. In this regard, the streamlined, flexible and efficient RISC-V instruction set architecture has shown its advantages and is expected to become one of the core processor architectures in the Internet of Everything era. As the preferred partner of Xuantie this time, Aixin Yuanzhi also shared the industry implementation practice based on Xuantie RISC-V. For example, in the field of Internet of Things, Aixin Yuanzhi applied Xuantie RISC-V processor and Combining the two self-developed core technologies of Aixin Zhimo AI-ISP and Aixin Tongyuan mixed precision NPU, it achieves fast and low-power AI detection wake-up, which can greatly reduce system power consumption and easily meet the needs of battery products.

At the same time, Aixin Yuanzhi has also created a high-energy-efficiency AI computing platform through the combination of Aixin Tongyuan mixed-precision NPU and Xuantie RISC-V IP, realizing large computing power applications such as graph segmentation and text-based image search to meet the needs of edge users. The demand for sensing, computing and data processing capabilities in different end-user scenarios.

Create a native AI processor and join hands with Xuantie RISC-V to reshape thousands of industries

Computing power is the foundation of AI, so in the AI ​​era, what kind of computing power infrastructure do we need?

Liu Jianwei pointed out that nowadays, the computing density of AI programs has become higher, calling for the emergence of a new processor called NPU. “As an AI-specific processor, NPU should have its own instruction set, rather than being simply understood as an accelerator or Expanded instruction set." It said that in AI computing, the object of instruction operation becomes a high-dimensional tensor, so the energy consumption of the entire processor needs to be focused on data calculation, and the energy consumption of data handling, sorting and combination is reduced. . From the perspective of reasoning, most AI algorithms/operators are already in a state of convergence. Therefore, NPU design using operators, the basic computing unit of the AI ​​model, as the instruction set can open up the hardware microarchitecture design space, allowing more design ideas to be used and improving energy efficiency.

As the core technology self-developed by Aixin Yuanzhi, Aixin Tongyuan mixed precision NPU is an AI computing processor with operators as the atomic instruction set. It adopts a multi-threaded heterogeneous multi-core design to realize operators, network microstructure, Data flow and memory access optimization, efficient support for mixed-precision algorithm design, and native support for Transformer network structure provide a good foundation for the application of large models on the edge and end sides.

At the same time, Aixintongyuan mixed-precision NPU also provides a mature and easy-to-use compiler, thereby greatly reducing the development and deployment costs of AI software, helping customers quickly get started within an hour, and promoting the vigorous development of AI computing.

Judging from the evaluation indicators of AI processors in the visual field, Aixin Tongyuan's mixed-precision NPU also performed well. When running SwinT, its quantification accuracy reached 80.45% higher than the market average , and it has a high performance of 416 FPS. and low power consumption of 199FPS/W.

In order to meet the diverse AI computing needs, the NPU as an AI processor also needs more flexible dynamic control. This is why Aixin Yuanzhi and Xuantie RISC-V have launched in-depth cooperation. According to reports, Aixin Yuanzhi is exploring the creation of generative AI solutions based on Xuantie RISC-V master control and Aixin Tongyuan mixed precision NPU . "We hope to join hands with Xuantie RISC-V to create the optimal platform choice for AI hardware, bring AI to thousands of industries, and help more customers improve efficiency and create value," said Liu Jianwei.