Ahead in the Cloud: Revolution at the RISC-V Summit China

By:

Age: 2 weeks

Poster of the 2024 RISC-V Summit China (Source: IT Home)

Executive Summary:

  • A state-private partnership involving China Telecom, the Chinese Academy of Sciences, and Alibaba has unveiled the industry’s first cloud computing platform dedicated to RISC-V architecture, to accelerate the adoption of RISC-V in cloud-native applications.
  • The RISC-V Summit China, held in Hangzhou, signaled RISC-V’s growing role in artificial intelligence (AI) and high-performance computing. Beijing-based ESWIN Computing introduced a chip designed for efficient AI processing and DAMO Academy unveiled a new R908 processor intended for industrial and automotive applications.
  • Beijing is committed to advancing RISC-V, which aligns with national strategies to foster innovation and reduce reliance on foreign technologies.

The 2024 RISC-V China Summit was held from August 21 to 23 in Hangzhou, Zhejiang Province. As one of the three largest RISC-V professional exhibitions globally, the summit drew over 3,000 participants from more than 100 companies, research institutions, and open-source communities. It also served as a platform for unveiling cutting-edge innovations and discussing the future of RISC-V, particularly in areas such as artificial intelligence (AI), cloud computing, and high-performance computing (Baijiahao, August 21).

The summit brought together key figures in the RISC-V community. These included Ni Guangnan (倪光南), an academician of the Chinese Academy of Engineering and Director of the Strategic Committee of the RISC-V Working Committee; Calista Redmond, CEO of the RISC-V International Foundation; Krste Asanovic, Co-founder and Chief Architect of SiFive, a leading American semiconductor company that uses RISC-V; and Xie Tao (谢涛), chief scientist of the Beijing Institute of Open Source Chip Research Institute (China Daily, August 21). Their participation underscored the importance of collaboration and strategic planning in advancing the RISC-V ecosystem, both within China and on a global scale.

PRC Entities Announce Key Developments

A key announcement at the summit was the launch of the “Beihai” RISC-V cloud computing experimental platform. This is a collaborative effort by the China Telecom Research Institute, the Institute of Software at the Chinese Academy of Sciences, Alibaba’s DAMO Academy, and others. It represents the industry’s first cloud computing environment dedicated to RISC-V architecture (Zaker, August 27). Cloud computing often relies on powerful servers that can handle vast amounts of data and complex computations. The “Beihai” platform integrates RISC-V CPUs—central processing units based on the RISC-V instruction set architecture—with various specialized processing units like TPUs (Tensor Processing Units, which are optimized for AI tasks), NPUs (Neural Processing Units, which accelerate machine learning), and VPUs (Video Processing Units, which handle video data). This setup not only allows for testing and verifying RISC-V chip designs but also makes these resources available to the broader public, potentially accelerating the development and adoption of RISC-V in cloud-native applications in the PRC.

Another highlight was the introduction of a new AI chip by ESWIN Computing. The EIC7702X AI PC chip features an 8-core, 64-bit out-of-order RISC-V processor, designed to handle tasks more efficiently by processing multiple instructions simultaneously (Baijiahao, August 21). It also includes an NPU that accelerates tasks associated with AI, such as recognizing patterns in data. The chip can perform up to 40 trillion operations per second (TOPS). This level of performance is crucial for applications like AI-driven image recognition and real-time language translation.

DAMO Academy unveiled an upgraded R908 processor as part of its Xuantie series of RISC-V chips. The processor is designed with flexible vector computing capabilities. This means it can handle a wide range of data types and operations simultaneously, something that is ideal for applications requiring high levels of performance and stability such as industrial control systems and automotive computing (Baijiahao, August 22). The R908 allegedly boasts a sevenfold increase in GEMM (General Matrix Multiply) computation speed, a key operation in many machine and deep learning algorithms. It also claims a seventeenfold improvement in transformer operator performance, which is critical for tasks like natural language processing (Nvidia, last accessed September 1; Vidha, June 7). These enhancements may demonstrate RISC-V’s expanding role in the PRC’s high-performance computing scenarios.

RISC-V to Compete with Incumbents in AI and Cloud Computing

The “Beihai” platform integrates RISC-V into cloud computing. The latest announcements by ESWIN Computing and Alibaba DAMO Academy also highlight RISC-V’s growing importance in AI and high-performance computing (HPC). This positions RISC-V as a strong competitor to established architectures like ARM and x86 in these growing domains where flexibility and customization are key advantages.

The strong participation from organizations and companies from the People’s Republic of China (PRC) in the summit reflects the country’s strategic commitment to RISC-V. By focusing on developing and promoting RISC-V, Beijing is fostering innovation within its technology sector while reducing reliance on foreign proprietary technologies (See China Brief, December 15, 2023). As these technologies evolve, RISC-V is set to become a cornerstone of digital infrastructure and, as this year’s RISC-V China Summit indicates, this will be the case far beyond the PRC itself. But the more attention and resources the PRC expends on such open-source technologies, the more influence it will end up having at the expense of other players.