NVIDIA HGX H100 4x NVIDIA H100 SXM Baseboard 320 GB HBM3

The NVIDIA HGX H100 4-GPU Baseboard is a powerful computing platform designed to accelerate AI, high-performance computing (HPC), and enterprise data center workloads. Featuring 4 NVIDIA H100 SXM GPUs interconnected through third-generation NVLink, it provides seamless multi-GPU communication with up to 900 GB/s of bandwidth, ensuring efficient scaling across demanding applications. With a total of 320 GB of HBM3 high-bandwidth memory, the HGX H100 enables rapid access to massive datasets, making it ideal for training and deploying large language models (LLMs), powering generative AI workloads, and running complex scientific simulations. Built on the NVIDIA Hopper architecture, the system introduces groundbreaking features such as Transformer Engine acceleration, FP8 precision support, and advanced sparsity optimizations, delivering unprecedented performance for both training and inference. Optimized for integration into modern data centers, the HGX H100 4-GPU Baseboard provides the compute density, scalability, and energy efficiency required to drive the next era of AI factories, hyperscale cloud infrastructure, and exascale scientific research.