Hewlett Packard Enterprise (HPE) has announced the shipment of its first system based on the NVIDIA Blackwell family, the NVIDIA GB200 NVL72. This high-performance system is designed to facilitate the deployment of large-scale artificial intelligence (AI) clusters, optimizing efficiency and performance with direct liquid cooling technology.
Key Features of the NVIDIA GB200 NVL72
The GB200 NVL72 stands out with its shared memory architecture and low latency, making it ideal for training AI models with over one trillion parameters in a single memory space. Its main features include:
- 72 NVIDIA Blackwell GPUs and 36 NVIDIA Grace CPUs, connected via high-speed NVIDIA NVLink.
- Up to 13.5 TB of HBM3e memory with a bandwidthBandwidth is the maximum transfer capacity of 576 TB/s.
- HPE’s direct liquid cooling technology, optimizing thermal and energy performance.
Innovation in Cooling and Energy Efficiency
With the growing energy demand and density of data centers, HPE brings five decades of experience in liquid cooling, enabling it to develop some of the most efficient supercomputers in the world. Currently, HPE has built seven of the ten fastest supercomputers and eight of the fifteen most efficient according to the Green500 list.
HPE’s liquid cooling approach offers benefits such as:
- Reduced energy consumption and optimized thermal performance for sustainable operations.
- Increased reliability and scalability in AI and high-performance computing (HPC) environments.
High-Level Services and Support
HPE provides a comprehensive suite of specialized services to facilitate the deployment and maintenance of these large-scale AI solutions:
- On-site support: Highly trained engineers work alongside IT teams to ensure maximum performance and availability.
- Optimization and benchmarking: HPC and AI experts fine-tune the system throughout its lifespan to ensure optimal performance.
- Sustainability services: Energy consumption reports, emission reduction workshops, and resource monitoring to minimize environmental impact.
Impact on AI and Supercomputing
The NVIDIA GB200 NVL72 is part of a range of high-performance solutions from HPE, designed to address use cases such as generative AI, scientific discoveries, and compute-intensive workloads. The combination of NVIDIA’s power with HPE’s infrastructure expertise allows for efficient deployment of customized AI clusters, offering rapid installation, greater scalability, and comprehensive support.
Conclusion
The shipment of the NVIDIA GB200 NVL72 marks a milestone in high-performance computing and artificial intelligence, providing businesses and service providers with a cutting-edge platform for developing massive AI models. With its advanced liquid cooling and optimized architecture, this system represents the future of efficient and scalable computing for AI and HPC applications.
Source: HPE