Understand this story
On October 26, 2026, the China Computer Federation (CCF) announced the results of its 2026 Science and Technology Achievement Awards. The project "Key Technologies and Applications of High-Performance Networks for Large-Scale Intelligent Computing Centers," jointly completed by Alibaba Cloud and Tsinghua University, won the CCF Science and Technology Achievement Award First Prize for Scientific and Technological Progress. According to Jiwei.com, the HPN solution proposed by the project adopts a two-tier flat, high-density multi-rail architecture. Third-party testing shows that the effective bandwidth of a 10,000-GPU cluster is nearly 30% higher than that of the traditional three-tier Clos architecture, with bandwidth utilization reaching 97%, and the average GPU idle time caused by fault diagnosis reduced by 97%. The network has supported dozens of 10,000-GPU intelligent computing clusters, carrying training and inference workloads for Qwen large models, Xiaohongshu, XPeng Motors and others. A report by Alibaba Cloud Developer Community AI also mentioned the above award information and bandwidth improvement data, and stated that in testing, bandwidth utilization was about 37% higher than Microsoft's WABS technology. There are currently no further developments.
AI-generated from public reports
Developments
1 confirmed developments
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 27
Alibaba Cloud Developer Community AI
Alibaba Cloud HPN 10,000-GPU Intelligent Computing Network Wins CCF Science and Technology Progress First Prize: Effective Bandwidth Increased by Nearly 30%
The "Key Technologies and Applications of High-Performance Networks for Large-Scale Intelligent Computing Centers" project, jointly completed by Alibaba Cloud and Tsinghua University, won the First Prize for Science and Technology Progress from the China Computer Federation (CCF). Its HPN solution compresses the three-tier network into a two-tier flat, multi-rail parallel architecture, increasing the effective bandwidth of a 10,000-GPU cluster by nearly 30% compared with traditional architectures. In testing, bandwidth utilization reached 97%, about 37% higher than Microsoft's WABS technology, and GPU idle time caused by failures was reduced by an average of 97%. In the same period, the Alibaba Cloud Bailian platform added 65 new applications and 11 MCP services in one week, and launched the DeepSeek-V4.1-Flash model, which strengthens image-text understanding and Agent capabilities.
Oct 26
Jiwei.com
Alibaba Cloud and Tsinghua Intelligent Computing Center's High-Performance Network HPN Wins CCF Science and Technology Progress First Prize
The China Computer Federation announced the 2026 Science and Technology Achievement Awards, and the project "Key Technologies and Applications of High-Performance Networks for Large-Scale Intelligent Computing Centers," jointly completed by Alibaba Cloud and Tsinghua University, won the Science and Technology Progress First Prize. HPN adopts a two-layer flat, high-density multi-rail architecture. Third-party testing shows that the effective bandwidth of a ten-thousand-card cluster is nearly 30% higher than that of the traditional three-layer Clos architecture, with bandwidth utilization reaching 97%, and GPU idle time caused by fault diagnosis reduced by an average of 97%. The network has supported dozens of ten-thousand-card intelligent computing clusters, carrying training and inference workloads for Qwen large models, Xiaohongshu, XPeng Motors, and others.