NVIDIA HGX

NVIDIA HGX is a modular data center platform designed for OEMs. It provides powerful AI and high-performance computing capabilities similar to NVIDIA DGX, but allows OEMs to integrate these technologies into their own solutions. Despite its different categorization and intended use for OEMs, HGX fully supports the NVIDIA SuperPOD standard similar to DGX, enabling scalable and high-performance GPU clusters to be built.

NVIDIA HGX features

• Modular Architecture: HGX is delivered in modules that include 8 GPUs and related components, allowing OEMs to build custom server solutions.
• SuperPOD support: The HGX platform is fully compliant with SuperPOD standards, providing interoperability and the ability to build highly scalable AI and data analytics solutions.
• NVIDIA Technologies: HGX incorporates the same advanced technologies as DGX, such as NVSwitch and NVLink, enabling high bandwidth data transfer between GPUs.

[text_with_btn btn=”Learn more” link=”https://itglobal.com/ru-ru/services/virtual-infrastructure/arenda-superkompyutera-na-baze-nvidia-superpod/”]Supercomputer Rental[/text_with_btn]

Applications of NVIDIA HGX

• Artificial Intelligence and Machine Learning: Training and inference of large models, including generative AI and LLMs.
• High Performance Computing (HPC): Processing large amounts of data and performing complex computational tasks.
• Graphics and Visualization: Creating high-quality visual content, rendering and 3D visualization.

.