#France #AI Infrastructure #Rueil-Malmaison #NVIDIA Mission Control #GB300 NVL72

Schneider Electric's Latest Innovations: AI Infrastructure Reference Designs

On September 18, 2023, Schneider Electric announced two cutting-edge reference designs developed in collaboration with NVIDIA, aimed at enhancing the deployment of AI infrastructures. These designs integrate power management with liquid cooling control in a way that has never been done before, spotlighting the partnership between cutting-edge technology and AI applications.

Industry-First Integration of Power and Cooling Control

The newly launched reference designs are pivotal in achieving seamless connectivity between OT (Operational Technology) and IT (Information Technology) via the NVIDIA Mission Control. This connection is vital for the efficient operation of AI-centric data centers, creating a synergy that can significantly cut down the deployment time and simplify operator involvement with AI infrastructure.

The first reference design introduces a controlling framework that integrates both power management and liquid cooling mechanisms. It utilizes Motivair's liquid cooling technology, allowing for the complex handling of various AI infrastructure components. This design supports the NVIDIA Mission Control’s operational and orchestration software that manages clusters and workload effectively. Notably, these designs can also be utilized in conjunction with Schneider Electric's NVIDIA Grace Blackwell system, enabling comprehensive control over power and cooling systems while embracing advanced accelerated computing technologies.

Focused on High-Density AI Deployment

The second reference design targets the deployment of infrastructures for AI factories, specifically intended for the NVIDIA GB300 NVL72 racks. It facilitates the setup of these advanced systems within a single data hall, allowing for the integration of the next-generation NVIDIA Blackwell Ultra architecture. This framework encompasses facility power, facility cooling, IT space, and lifecycle software, adhering to both ANSI (American National Standards Institute) and IEC (International Electrotechnical Commission) standards.

As the demand for AI capabilities escalates, current data center operators are adopting these reference designs to address the challenge of deploying high-density GPU-accelerated AI clusters rapidly. Schneider Electric has long been a leader in energy management and digital transformation, committed to providing proven and tested designs that optimize efficiency, reliability, and cost-effectiveness in AI infrastructure.

Enhanced Control through End-to-End Systems

The newly revealed control reference designs operate as “plug-and-play” systems, enhancing the relationships between energy management and liquid cooling devices. Based on the MQTT protocol, these systems bridge OT infrastructure and IT systems, enabling operators to utilize data from every layer of the operation to optimize performance.

This design places significant focus on ensuring interoperability between the building management software and AI infrastructure management software. By establishing redundant systems for power and cooling, as well as introducing guidelines for measuring power profiles of AI racks, Schneider Electric aims to ensure precise, real-time management of critical power and cooling resources. Thus, guaranteeing optimal uptime, reliability, and peak performance in AI deployments.

Advancing the Future of Data Centers

Schneider Electric's reference design for the NVIDIA GB300 NVL72 clusters supports deployments with maximum density up to 142kW. The setup is optimized to accommodate high-performance cores such as NVIDIA DGX SuperPOD equipped with the DGX GB300 system, employing liquid cooling CDU and high-temperature chillers to operate efficiently.

The reference design also features Schneider Electric’s leading ETAP and EcoStruxure IT Design CFD models, allowing users to simulate specific power and cooling scenarios, ensuring tailored and efficient design for unique applications. By continually collaborating with NVIDIA, Schneider Electric aims to meet the most pressing needs within the data center domain.

Jim Simonelli, Senior Vice President and CTO of Schneider Electric, remarked, "With the new reference designs, we are streamlining the processes of designing, deploying, and operating advanced AI infrastructure. By leveraging integrated power management and liquid cooling controls, we offer a scalable solution that meets real-world application demands and supports data center operators amid soaring AI needs."

Scott Wallace, Director of Data Center Engineering at NVIDIA, added, "Schneider Electric's latest control reference design provides rigorously validated blueprints that connect essential infrastructure data with NVIDIA Mission Control, facilitating operation within today’s and tomorrow’s advanced computing frameworks."

With these innovations, Schneider Electric continues to define the forefront of AI infrastructure, setting a smart path for future developments and deployments.