24 October 2024, Thu |
8:23 AM
Vertiv (NYSE: VRT), a provider of critical digital infrastructure, has announced the release of a complete 7MW reference architecture for the NVIDIA GB200 NVL72 platform. This collaborative effort with NVIDIA aims to help customers transform traditional data centers into AI-driven facilities capable of supporting enterprise-level AI applications.
The new reference architecture facilitates the rapid deployment of the NVIDIA GB200 NVL72 liquid-cooled rack-scale platform, which can support up to 132kW per rack. This end-to-end infrastructure design is tailored to optimize deployment speed, performance, resilience, cost, energy efficiency, and scalability for both current and future-generation data centers.
Giordano (Gio) Albertazzi, CEO of Vertiv, expressed pride in deepening collaboration with NVIDIA, stating, “As a leader in critical power and cooling infrastructure, Vertiv is uniquely positioned to support the NVIDIA GB200 NVL72 platform. Our portfolio of high-performance solutions, combined with our global scale, enables customers to deploy AI data centers faster and more efficiently while addressing the demands of densification, dynamic workloads, and retrofits.”
NVIDIA founder and CEO Jensen Huang highlighted the complexity of new data centers designed for accelerated computing and generative AI. He emphasized that Vertiv’s world-class power and cooling technologies are essential for realizing NVIDIA’s vision of reinventing computing and building AI factories that produce digital intelligence across various industries.
The reference architecture simplifies the deployment of AI workloads in both new and existing data centers, reducing risks and enabling standardization across sites. Its critical power infrastructure aims to minimize stranded power by aligning AI clusters with data center capacity blocks. The hybrid liquid and air-cooling infrastructure effectively manages high-density heat removal, and the design includes optional Open Compute Project-inspired systems such as DC power shelves.
Key benefits of the Vertiv and NVIDIA GB200 NVL72 reference architecture include:
– Rapid Deployment and Retrofit: Vertiv™ MegaMod™ CoolChip allows for the delivery of AI-critical infrastructure up to 50% faster than traditional onsite builds through preconfigured modules and factory integration.
– Space-Saving Power Management: Featuring advanced power technologies like the Vertiv™ Trinergy™ uninterruptible power supply system (UPS) and Vertiv™ EnergyCore lithium battery cabinet, this design provides industry-leading reliability and energy efficiency in approximately 40% less space than legacy systems.
– Energy-Efficient Cooling: The integration of liquid and low-GWP (Global Warming Potential) air cooling technologies, including Vertiv™ AFC chiller and Vertiv™ Liebert® CW chilled water-based systems, can reduce annual cooling costs by up to 20% compared to fixed screw solutions.
– Dynamic Workload Management: Integrated load averaging via lithium-ion battery and next-generation UPS systems supports dynamic GPU workloads.
– Comprehensive Installation and Operations Services: With around 4,000 field service engineers worldwide, Vertiv offers trusted lifecycle services and expertise for both retrofit and new build projects.
As enterprises rapidly adopt AI technologies, Vertiv is committed to reshaping critical power and cooling solutions to support accelerated computing. This collaboration with NVIDIA establishes a roadmap for future technical co-development, enabling large-scale deployments of accelerated computing infrastructure.
