During the Huawei Full Connectivity Conference held in Shanghai, Xu Zhijun, the rotating chairman of Huawei, and Liao Heng, chief scientist of Hisilicon, had questions and answers with media reporters about Huawei's newly released Peerium computing architecture and Lingqu bus for the AI era. During this year's HC conference, Huawei exhibited the Ascend 950 and Atlas 950 supernodes. Xu Zhijun said that “supernodes” are now widely mentioned, but different manufacturers have very different implementations. The real challenge is to make thousands or tens of thousands of processors form a single computer. Huawei named this new computing architecture Peerium, which can turn one million processors into one computer. Currently, Huawei is deploying and testing an Atlas 950 SuperPod supernode with a size of 256,000 cards. Liao Heng believes that in the next two years, it is expected that about 6 to 7 cutting-edge AI laboratories will appear in China. The goal is to train basic language models with parameters between 10 trillion and 40 trillion yuan. These numbers roughly match the memory capacity and size of supernodes, so infrastructure of this scale is required for training.

Disclaimer:Webull uses external vendor Google Translation Service for news translations where we endeavour to ensure these are correct, however, we recommend that you please double-check this information accordingly. Webull is not responsible for translation errors or issues.