GB/T 46346-2025Artificial intelligence - Computing center - Computing capacity assessment (English PDF)

Foreword

This document conforms to GB/T 1.1-2020 "Standardization Work Guidelines Part

1.Structure and Drafting Rules of Standardization Documents". Drafting is scheduled. Please note that some content in this document may involve patents. The issuing organization of this document assumes no responsibility for identifying patents. This document was proposed and is under the jurisdiction of the National Information Technology Standardization Technical Committee (SAC/TC 28). This document was drafted by: China Electronics Technology Standardization Institute, Huawei Technologies Co., Ltd., and Inspur Electronic Information Industry Co., Ltd. China Telecom Group Corporation, China Mobile Research Institute, Alibaba Cloud Computing Co., Ltd., Beijing Academy of Artificial Intelligence Beijing University of Aeronautics and Astronautics, ZTE Corporation, Shanghai Suiyuan Technology Co., Ltd., Shanghai Biren Technology Co., Ltd. Shanghai Tianshu Intelligent Semiconductor Co., Ltd., Super Fusion Digital Technology Co., Ltd., Hygon Information Technology Co., Ltd., and Shanghai SenseTime Intelligent Technology Co., Ltd., Baidu Netcom Technology Co., Ltd., Pengcheng National Laboratory, North China Electric Power Research Institute Co., Ltd. China Mobile (Suzhou) Software Technology Co., Ltd., NARI Technology Co., Ltd., Guoneng Information Technology Co., Ltd., Peking University, China China Railway Construction Corporation Limited, China Railway Fifth Survey and Design Institute Group Co., Ltd., China Telecom Cloud Technology Co., Ltd., Shenzhen Kunyun Information Technology Co., Ltd. Pingtouge (Shanghai) Semiconductor Technology Co., Ltd., Shanghai Artificial Intelligence Innovation Center, H3C Information Technology Co., Ltd., Ant Group Joint-stock limited companies, China Southern Power Grid Science Research Institute Co., Ltd., iFlytek Co., Ltd., and Hangzhou Hikvision Digital Technology Co., Ltd. China Mobile Systems Integration Co., Ltd., Henan Kunlun Technology Co., Ltd., Moore Threads Intelligent Technology (Beijing) Co., Ltd. Kunlun Core (Beijing) Technology Co., Ltd., Zhejiang Dahua Technology Co., Ltd., Peking University Changsha Institute of Computing and Digital Economy, Qingdao Hong Kong International Co., Ltd., Guangzhou CESI Standard Testing & Research Institute Co., Ltd., Guilin University of Electronic Technology, Newland Digital Technology Co., Ltd. Limited Liability Company, Beyond Technology Co., Ltd., Inspur Communication Information Systems Co., Ltd., Mashang Consumer Finance Co., Ltd., Nanjing NARI Ruiteng Technology Co., Ltd., China Southern Power Grid Co., Ltd. Ultra-High Voltage Transmission Company, Shanghai Wenyao Information Technology Co., Ltd. Beijing Simou Intelligent Technology Co., Ltd., Tongfang Knowledge Network Digital Publishing Technology Co., Ltd., Guangdong Research Institute of China Telecom Corporation Limited Xinjiang Institute of Physics and Chemistry, Chinese Academy of Sciences; China Mobile Xiong'an Information and Communication Technology Co., Ltd.; Zhejiang University Institute of Computer Innovation Technology Shanghai Artificial Intelligence Research Institute Co., Ltd., Shanghai Artificial Intelligence Industry Association, Shenzhen UBTECH Robotics Corp., Ltd., Beijing Ansheng Technology company. The main drafters of this document are. Fan Kefeng, Xu Yang, Zhang Liang, Liu Wenfeng, Yang Yuze, Xi Rui, Liang Zhihong, Liu Weichen, Qin Rizhen, and Yu Xiaobo. Lu Shun, Wu Wenli, Liu Xianglong, Li Jianpeng, Yang Heng, Liu Yu, Liu Zhen, Shen Zhiyue, Chen Leran, Shi Honghao, Ma Shanshan, Huang Cheng, Zhu Jing, Mei Jingqing Wang Sishan, Ding Ruiquan, Xing Feng, Hu Mingshan, Yu Xuesong, Zhou Xinzhong, Zhang Peng, Duan Aimin, Wu Zongyou, Wu Geng, Xiang Tieyao, Tian Xiaoli, Xiao Song, Zhang Yehong Yu Yue, Li Min, Zhang Zhihong, Li Xudong, Zhang Wancai, Jing Dichun, Zhang Tianlin, Yang Chao, Li Dong, Zheng Zhong, Yang Ming, Miao Zicong, Luan Lihong, Gao Hui, Zhang Yibo, Zhang Chengxing, Li Xiaoru, Xu Shenchao, Liu Rubing, Wang Heng, Guo Zhihui, Qiao Yuping, Zhang Lei, Ye Tingqun, Zhou Zhiqiang, Zhang Tian, Yang Jiali Liu Jinnan, Liang Yonggui, Shen Qian, Gu Canghai, Deng Zhiji, Kong Weisheng, Gou Haipeng, Guo Yiyun, Wu Yuzhen, Huang Shenghua, He Qian, Kong Hao, Cai Chunshui Lin Jing, Wu Dengyong, Xiao Hongmei, Liang Binghao, Feng Yue, Mu Jun, Shi Chao, Shi Yu, Wang Ning, Li Xuewu, Zhong Kaitao, Rui Ziwen, Su Chi, Yu Kai, Li Da Ma Xing, Shi Chunyu, Fu Deji, Yang Yating, Zhou Xi, Zheng Qingguo, Wang Jing, Xia Ruichen, Chen Xi, Xu Chunshan, Zhao Chunhao, Rao Xue, Liang Qiaoling, Pang Jianxin Liu Yifan. Computing capacity assessment of artificial intelligence computing centers

1 Scope

GB/T 46346-2025 is the Chinese national standard covering measuring what an AI computing centre can actually do - the throughput on representative training and inference workloads rather than the peak FLOPS on the datasheet, the interconnect and storage that usually bind, the utilisation achieved, and the reporting. First edition, and it belongs with the intelligent computing terminology GB/T 46572-2025. Already in force. Issued on 5 October 2025, it has been in force since 5 October 2025.

This document specifies the evaluation indicators for the computing capabilities of artificial intelligence computing centers and describes the corresponding evaluation methods. This document applies to the assessment of the computing capabilities of artificial intelligence computing centers, and also serves as a guide for the planning, design, construction, and service of artificial intelligence computing centers. It provides a reference for operation and maintenance.

2 Normative references

The contents of the following documents, through normative references within the text, constitute essential provisions of this document. Dated citations are listed below. For references to documents, only the version corresponding to that date applies to this document; for undated references, the latest version (including all amendments) applies. This document.

GB/T 41867-2022 Terminology for Information Technology and Artificial Intelligence

GB/T 42018-2022 Information Technology Artificial Intelligence Platform Computing Resource Specification

GB/T 45087-2024 Performance Testing Methods for Artificial Intelligence Server Systems

3 Terms and Definitions

The terms and definitions defined in GB/T 41867-2022 and GB/T 42018-2022, as well as the following terms and definitions, apply to this document.

3.1 Artificial intelligence computing center A structure or group of structures that can provide artificial intelligence computing services to multiple users. Note

1.Abbreviated as "Intelligent Computing Center," an intelligent computing center includes hardware devices with artificial intelligence computing characteristics, such as computing, storage, and networking equipment, as well as device drivers and computing... Accelerator libraries, management components, and other necessary software components that support artificial intelligence computing capabilities. Note

2.Intelligent computing centers generally provide services such as training calculations, inference calculations, model fine-tuning, and related data storage, processing, and transmission for artificial intelligence models.

3.2 Artificial intelligence computing capability The ability to perform artificial intelligence computing tasks or support intelligent applications.

3.3 performance Measurable characteristics when running computational tasks. Note

1.Performance includes qualitative and quantitative characteristics. Note

2.Performance is measured based on one or more parameters (such as runtime, energy consumption, throughput, actual throughput, floating-point operations per second, fault recovery time, etc.). Quantified or calculated to characterize the capability and efficiency of a technical process operating in a certain device (group).

3.4 computing node Devices or groups of devices that provide computing power.

Note. In intelligent computing centers, computing nodes typically include computing devices such as AI accelerator processors, AI accelerator cards, or AI servers, as well as... Any other storage and networking devices in the intelligent computing center used for processing computing tasks.