Schneider Electrical, a contributor within the digital transformation of power administration and automation, has introduced new reference designs developed with NVIDIA that speed up time to deployment and help operators as they undertake AI-ready infrastructure options.
The first reference design delivers the business’s first and solely important framework for built-in energy administration and liquid cooling management techniques, together with Motivair by Schneider Electrical liquid cooling applied sciences and allows seamless administration of complicated AI infrastructure elements. It contains interoperability with NVIDIA Mission Management, NVIDIA’s AI manufacturing unit operations and orchestration software program, together with cluster and workload administration options. The management techniques reference design can be utilised with Schneider Electrical’s information centre reference designs for NVIDIA Grace Blackwell techniques, enabling operators to maintain tempo with the newest developments in accelerated computing, with seamless management of their energy and liquid cooling techniques.
The second reference design focuses on the deployment of AI infrastructure for AI factories of as much as 142 kW per rack, particularly NVIDIA GB300 NVL72 racks, in a single information corridor. Created to offer a framework for the next-generation NVIDIA Blackwell Extremely structure, the reference design contains data on 4 technical areas: facility energy, facility cooling, IT area and lifecycle software program. The design is obtainable underneath configurations for each the American Nationwide Requirements Institute (ANSI) and the Worldwide Electrotechnical Fee (IEC) requirements.
The primary controls reference design can be utilised with Schneider Electrical’s information centre reference designs for NVIDIA Grace Blackwell techniques, enabling operators to maintain tempo with the newest developments in accelerated computing, whereas having seamless management of their energy and liquid cooling techniques.
Engineered to maneuver at tempo
As AI advances, at present’s information centre operators depend on reference design frameworks to beat the pace and deployment challenges of high-density, GPU-accelerated AI clusters. By offering validated, confirmed, and documented information centre bodily infrastructure designs, Schneider Electrical allows operators globally to design and incorporate next-generation energy and liquid cooling controls infrastructure earlier than the latest AI infrastructure options even arrive, whereas optimising for value, effectivity and reliability. Schneider Electrical’s absolutely engineered reference designs are laying the muse for the newest AI factories, empowering information centre operators to not solely meet the second however to be prepared earlier than it even occurs.
“Schneider Electrical is streamlining the method of designing, deploying and working superior AI infrastructure with its new reference designs,” stated Jim Simonelli, the senior vice chairman and chief expertise officer at Schneider Electrical. “Our newest reference designs, that includes built-in energy administration and liquid cooling controls, are future-ready, scalable and co-engineered with NVIDIA for real-world purposes, enabling information centre operators to maintain tempo with surging demand for AI.”
“We’re coming into a brand new period of accelerated computing, the place built-in intelligence throughout energy, cooling and operations will redefine information centre architectures,” stated Scott Wallace, Director of Information Centre Engineering at NVIDIA. “With its newest controls reference design, Schneider Electrical connects important infrastructure information with NVIDIA Mission Management, delivering a rigorously validated blueprint that permits AI manufacturing unit digital twins and empowers operators to optimise superior accelerated computing infrastructure.”
‘Plug-and-play’ end-to-end controls system
The controls reference design connects edge gadgets and facility controls for power administration and liquid cooling throughout NVIDIA GB300 NVL72 and NVIDIA GB200 NVL72 deployments utilizing NVIDIA Mission Management. Utilizing a “plug-and-play” structure primarily based on the MQTT protocol, it bridges operational expertise (OT) infrastructure and data expertise (IT) techniques, permitting operators to, for the primary time, harness information from each layer to optimise efficiency.
With a concentrate on seamless interoperability between constructing and AI infrastructure administration software program, the controls reference design establishes redundant techniques for energy and cooling, and introduces new steerage for measuring AI rack energy profiles. In the end, it ensures the very best requirements of uptime, reliability and peak efficiency for AI deployments by enabling exact, real-time administration of important energy and cooling assets.
The result’s an end-to-end controls system that gives:
- A standardised format interface publishing energy administration and liquid cooling controls information for consumption by native purposes and upstream information customers, and instruments, together with AI infrastructure administration software program, digital twins, AI/ML and different enterprise techniques.
- The controls structure is designed to handle redundancy throughout cooling and energy distribution infrastructure, together with coolant distribution models (CDUs) and distant energy panels (RPPs), guaranteeing resiliency in each white and gray area environments.
- New steerage for measuring AI rack energy profiles, with a concentrate on rack peak energy and energy high quality monitoring.
NVIDIA GB300 NVL72 reference design for high-density AI clusters
Schneider Electrical’s reference design for NVIDIA GB300 NVL72 helps the deployment of NVIDIA GB300 NVL72-based clusters with a most rack density of 142kW, comparable to NVIDIA DGX SuperPOD with DGX GB300 techniques. The information corridor is purpose-built and optimised to host three NVIDIA GB300 NVL72- primarily based clusters powered by as much as 1,152 GPUs utilizing liquid-to-liquid CDUs and excessive temperature chillers.
The reference design additionally contains Schneider Electrical’s ETAP and EcoStruxure IT Design CFD fashions, permitting customers to make use of digital twins to simulate particular energy and cooling situations to optimise designs on distinctive purposes. It builds upon a earlier blueprint for the NVIDIA GB200 NVL72 as Schneider Electrical continues to collaborate with NVIDIA to offer absolutely engineered, examined fashions in anticipation of the brand new NVIDIA GB300 NVL72 platform.
The reference designs introduced at present proceed Schneider Electrical’s long-established collaboration with NVIDIA to satisfy the info centre business’s most urgent AI calls for. Along with these new reference designs, Schneider Electrical has developed 9 AI reference designs for varied situations together with prefabricated modules, retrofit information centres, and AI infrastructure purpose-built for NVIDIA GB200 NVL72 and NVIDIA GB300 NVL72-based clusters.
Schneider Electrical additionally applies complete, real-world engineering to every of its reference designs, demonstrating a steady dedication to creating the business’s most superior, energy-efficient, and resilient high-performance information centre architectures. For extra details about its reference designs co-engineered with NVIDIA, go to se.com/data-centres-and-networks/reference-designs.
Touch upon this text through X: @IoTNow_ and go to our homepage IoT Now