Huawei Ascend 950 AI Cluster Set for Global Launch on November 30

Huawei continues to expand its footprint in the artificial intelligence sector by announcing major hardware advancements. During its recent Connect event, the tech giant officially revealed that the highly anticipated Huawei Ascend 950 AI cluster will make its global debut later this year, bringing cutting-edge computing power to enterprises worldwide.

• ✨ The Huawei Ascend 950 AI cluster is officially scheduled for a global launch on November 30.
• ✨ Powered by the advanced UnifiedBus Interconnect technology protocol and a modernized supernode architecture.
• ✨ Delivers massive 1 EFLOPS FP8 and 2 EFLOPS FP4 performance metrics via a robust 1024-card computing power cluster.
• ✨ Features 256TB of globally unified memory addressing storage space to streamline ultra-large-scale model training.

Next-Generation Architecture and Commercial Availability

Zhou Yuefeng, Executive Director of Huawei and CEO of Huawei Cloud, shared key details regarding the rollout timeline. While the global release is set for late November, the intelligent computing cluster cloud service will see commercial availability in China by the end of September.

The core strength of the Ascend 950 supernode AI computing cluster lies in its proprietary UnifiedBus Interconnect technology protocol. This architecture enables the system to effortlessly manage the entire lifecycle of ultra-large-scale model training involving hundreds of billions of nodes. Crucially, it accomplishes this complex task without requiring additional cluster architecture splitting or adaptation, making it an efficient solution for enterprise environments.

Furthermore, Huawei Cloud will integrate the AI cluster as a ready-to-use cloud service for the commercial market. Organizations seeking to upgrade their infrastructure with Huawei's latest innovations can expect full deployment options following the forthcoming global rollout.

When will the Huawei Ascend 950 AI cluster launch globally?

The Huawei Ascend 950 AI cluster is scheduled to make its official global debut on November 30.

What technologies power the Ascend 950 cluster?

The cluster utilizes Huawei's self-developed UnifiedBus Interconnect technology protocol alongside a newly upgraded supernode architecture.

What level of computing power does the cluster deliver?

It features a 1024-card computing power cluster capable of delivering 1 EFLOPS FP8 and 2 EFLOPS FP4 computing power, supported by 256TB of globally unified memory addressing storage space.

How does the Ascend 950 handle large-scale models?

It is fully capable of managing the entire process of ultra-large-scale model training with hundreds of billions of nodes without needing extra cluster architecture splitting or adaptation.

🔎 As Huawei continues to push the boundaries of semiconductor innovation and cloud computing services, the upcoming launch of the Ascend 950 cluster marks a significant milestone in the global artificial intelligence landscape, offering organizations a powerful alternative for large-scale model training and deployment.