The annual Global AI Chip Summit (GACS 2024) officially commenced today in Beijing, attracting significant attention with a packed audience and over 1.2 million viewers tuning in through live streams across various platforms.
The summit, organized by ZhiYi Technology and its subsidiary, Core Things, in collaboration with ZhiXingXing, is themed “Smart Computing Era: Building the Future of Chip Technology.” The event features over 50 esteemed speakers from diverse fields such as AI chips, Chiplet technology, RISC-V architecture, AI infrastructures, and more, gathered to share their insights and expertise.
This year marks the fifth anniversary of WallFace Technology, a domestic GPGPU unicorn, which unveiled a significant breakthrough in multi-core mixed training technology at the conference. The company introduced its HGCT (Heterogeneous GPU Cooperative Training) solution, marking an industry first that allows for the training of large models using three or more heterogeneous GPUs simultaneously.
ZhiYi Technology’s co-founder and CEO, Gong Lun Chang, delivered an opening address highlighting that this year marks the seventh installment of the Global AI Chip Summit, establishing it as the most influential conference in the country within this domain and a vital window into the development dynamics of AI chips both domestically and internationally.
The summit spans two days, featuring a main venue for the opening ceremony and three specialized sessions focusing on AI chip architecture, Datacenter AI chips, and Edge AI chips. Additional forums include discussions on Chiplet technology and innovations in RISC-V.
In a keynote presentation, Professor Yin Shou Yi, Vice Dean of the School of Integrated Circuits at Tsinghua University, tackled the developmental pathways for high-performance chips, covering the technological challenges present and examining five innovative technical routes including data flow chips, reconfigurable chips, and three-dimensional integrated chips.
A panel discussion highlighted contributions from industry experts and executives from leading AI chip companies focused on current developments, practices, and future directions. Significant commentary emerged regarding how to overcome the challenges surrounding computational power demands for large models, emphasizing architectural innovations that fully leverage available transistors for enhanced performance while addressing system limitations and bandwidth issues.
Leading firms like AMD and Qualcomm showcased their latest advancements at the summit. AMD’s Senior Director Wang Hongqiang shared insights on their end-to-end AI infrastructure that supports a multitude of applications while delivering competitive products, including a next-generation AI PC equipped with groundbreaking NPU technology.
Meanwhile, discussions delved into industry trends, such as the emergence of domestic AI startups focusing on specific applications in intelligent video processing and edge computing. These new enterprises emphasize the importance of cost-effectiveness and innovation in order to compete against major players like NVIDIA.
As the conference participants engage in conversations regarding scaling AI technology and infrastructure demands, there is a clear sense of urgency to address the stifling bandwidth bottlenecks and ensure the continual development of high-performance computing ecosystems.
The momentum gained through the summit is poised to trigger further advancements in the AI chip market, which continually evolves to meet the demands of cutting-edge technologies and applications emerging in sectors from smart cities to autonomous vehicles and beyond.
The Global AI Chip Summit will continue tomorrow, featuring dedicated sessions on AI chip architecture innovations, edge AI chips, and the announcement of top Chinese AI chip companies, ensuring that the conference remains a focal point for industry insights and progress in AI chip technology.