End-to-end critical power and cooling reference designs for NVIDIA Blackwell architectures up to 7MW, with OCP infrastructure options. This is Part 2 of a two-part series.
End-to-end critical power and cooling reference designs. Supplied by Vertiv
The architecture’s complete critical power infrastructure is designed to significantly reduce stranded power by aligning AI clusters to data centre capacity blocks. The hybrid liquid- and air-cooling infrastructure leverages the interdependent impact of the two approaches to efficiently manage high-density heat removal. In addition, the design includes guidance for optional Open Compute Project-inspired systems, such as DC power shelves.
The reference architecture is part of the Vertiv 360AI portfolio of reference designs for retrofit and greenfield data centres, designed to help customers navigate integrated solutions for power and cooling for AI and other high-performance computing applications. Key benefits of the co-developed reference architecture for the NVIDIA GB200 NVL72 design include:
- Rapid Deployment and Retrofit: Enabling the use of preconfigured modules and factory integration, VertivTM MegaModTM CoolChip delivers turnkey AI critical infrastructure up to 50% faster than onsite builds.
- Space-Saving Power Management: Using Vertiv’s advanced power technologies, including Vertiv Trinergy uninterruptible power supply system (UPS) and Vertiv EnergyCore lithium battery cabinet, the design delivers industry-leading reliability and energy-efficient power management in ~40% less space compared to legacy offerings.
- Energy-Efficient Cooling: Integrating liquid and low-GWP (Global Warming Potential) air cooling technologies at scale – including Vertiv AFC chiller, Vertiv Liebert CW chilled water-based room cooling system and Vertiv XDU coolant distribution units – offers up to 20% lower annual cooling costs compared to fixed screw solutions.
- Dynamic Workload Management: Integrated load averaging via lithium-ion battery and next generation UPS provides support for dynamic GPU workloads.
- Installation and Operations Services: With an industry-leading scale, scope, and reach of ~4,000 field service engineers globally, Vertiv is the trusted lifecycle service partner and complex system-level expert for retrofit and newbuild.
As enterprises embrace AI at an unprecedented pace, Vertiv is reshaping the future of critical power and cooling to support accelerated computing, with the most complete portfolio of critical digital infrastructure that enables AI-ready infrastructure capable of managing the unique requirements of AI and other accelerated compute applications.
Vertiv’s collaboration with NVIDIA sets a future-forward roadmap for technical co-development and enables deployment of accelerated computing at scale.
