HPE NVIDIA GB300 NVL72 by Base Rack

Do you need a highly performant, integrated solution for training, tuning or inferencing of AI models over one trillion parameters? The NVIDIA GB300 NVL72 by HPE is a liquid-cooled, rack-scale solution featuring a 72-GPU NVLink™ domain that acts as a single massive GPU. It integrates 72 NVIDIA Blackwell Ultra GPUs, 36 NVIDIA Grace™ CPUs, NVIDIA networking and software and HPE Services to deliver extreme performance for AI models of over one trillion parameters. The offering enables AI service providers and ambitious AI model builders to accelerate deployment with proven services, backed by our trusted partnership with NVIDIA and a global presence. These organizations can operate confidently with worry-free direct liquid cooling from Hewlett Packard Enterprise, the market leader ^[1] in large DLC deployments and increase their investment with world-renowned performance tuning and optimization by expert HPE engineers.

What is new supporting point:
NVIDIA ConnectX-8 SuperNICs
72 NVIDIA Blackwell Ultra GPUs and 36 NVIDIA Grace™ CPUs
High-speed NVIDIA NVLink™ Switch for GPU-to-GPU connectivity
Up to 20 TB total HBM3E memory with 569 TB/s total memory bandwidth
HPE direct liquid cooling technology
HPE Services