Intelligent Computing Full-Stack Solution 2.0 Released; Four Liquid-Cooling Industry Standards Launched

9On the 29th, during the China Computing Power Conference, the China Academy of Information and Communications Technology (CAICT) and Chindata Group jointly hosted the 2024 forum on data-center technology leadership and transformation in the intelligent computing era. Under the theme of an integrated industry ecosystem and a new intelligent computing future, experts from CAICT, China Telecom, Intel China, Baidu, Alibaba Cloud, Inspur, Seagate, Chindata and other organizations discussed AI computing, IT equipment, and emerging data-center challenges and solutions.

Responding to higher density, greater scale and heterogeneous computing, Chindata introduced Intelligent Computing Data Center Full-Stack Solution 2.0. It also partnered with ecosystem organizations to launch a series of industry standards covering lifecycle quality control for cold-plate liquid-cooling systems and establish an expert group to accelerate standardization.

Full-Stack Solution 2.0 Upgraded for AI Applications

Increasing GPU and server power consumption and computing clusters exceeding 100 MW create new challenges for conventional data centers. Chindata has developed and continuously improved its Panshi modular framework to address high computing power, hyperscale deployment, flexibility, efficiency and intelligence.

Within the Panshi framework, Chindata and partners introduced Solution 2.0 around four areas: flexible building models, the Xuantie simplified power architecture, the Xuanbing cooling architecture and intelligent operations. Compared with its predecessor, the solution is more advanced, modular and integrated, supporting larger computing needs for large models and AGI and enabling faster delivery and flexible expansion.

The flexible building model allows configurable modules and phased prefabrication, supports flexible deployment, accommodates CPU and GPU workloads and improves IT output per unit area. It addresses rapidly growing high-density demand and heterogeneous computing and enables faster delivery.

The Xuantie power architecture combines more than ten conventional distribution stages into six electromechanical modules: substations, diesel generation, medium voltage, low voltage, batteries and IT. Modules support phased deployment and flexible expansion for fast, high-quality delivery.

The Xuanbing architecture includes air cooling, liquid cooling and hybrid modules to cool cabinets with power densities from 8 to 150 kW. Technologies can be combined flexibly to suit building, climate and regional conditions.

Intelligent operations use Chindata's integrated monitoring and IDC management platforms to standardize and improve data-center operation. Joint development with AI companies aims to optimize cooling-system control.

Chindata CTO Zhang Binghua said that AI is reshaping data-center technology and that the company pursues forward-looking, systematic innovation throughout the business lifecycle. The upgrade emphasizes modular, distributed, integrated and prefabricated designs to improve flexibility across workloads and efficiently convert electrical power into computing capacity.

Four Industry Standards to Accelerate Cold-Plate Liquid-Cooling Standardization

Led by CAICT and Chindata with dozens of industry leaders, the series covers design quality control, coolant management and testing methods, testing quality control, and operations quality control for cold-plate liquid cooling. It aims to provide practical lifecycle quality guidance, strengthen the ecosystem and support faster innovation and larger-scale commercial deployment.

Before launching the standards, CAICT, Chindata and partners produced a white paper on lifecycle quality control for data-center cold-plate liquid cooling. Drawing on Chindata's 100 MW-scale deployment, experience from more than 30 companies and international standards, it presents a comprehensive quality-management framework and will inform the new standards.

Zhang Binghua said the standards are essential to consistent technical requirements and a sound industry ecosystem. The goal is broad consensus and a balance between rigorous metrics and large-scale deployment, supporting cooling innovation and sustainable data-center development.