NVIDIA H100 80GB SXM

The NVIDIA H100 80GB (SXM5) is a data center GPU built on the NVIDIA Hopper architecture, designed for large-scale AI training, inference, and high-performance computing (HPC). It pairs 80 GB of high-bandwidth HBM3 memory with very high memory bandwidth (approximately 3.35 TB/s on the SXM5 module).

• Hopper architecture with fourth-generation Tensor Cores and a dedicated Transformer Engine supporting FP8, accelerating transformer-based models such as large language models (LLMs).
• SXM5 form factor for NVIDIA HGX baseboards, with fourth-generation NVLink for high-speed GPU-to-GPU links.
• 80 GB HBM3 for large model and dataset residency, reducing the need to shard across devices.

Typical use cases: training and fine-tuning LLMs and other large neural networks, high-throughput inference, scientific computing, and multi-GPU cluster deployments.

This listing is for the SXM5 module, intended for HGX H100 systems rather than standard PCIe slots; appropriate server infrastructure, power, and cooling are required. Contact us for configuration and volume pricing.