NVIDIA Multi-Node NVLink Systems
GB200 NVL72 connects 36 Grace CPUs and 72 Blackwell GPUs using a 72-GPU NVLink domain in a rack-scale, liquid-cooled design.
User Guides
The Multi-Node Tuning Guide is designed to provide detailed insights and practical tips for fine-tuning multi-node environments, ensuring you achieve maximum efficiency and scalability. Whether you're managing high-performance computing clusters, large-scale data processing, or complex simulations.
The IMEX service supports GPU memory export and import (NVLink P2P) and shared memory operations across OS domains in an NVLink multi-node deployment.
The NVIDIA Firmware Update Tool (NVFWUPD) user guide provides information about using the tool to version, update, and activate system component firmware for NVIDIA Grace Hopper and Grace Blackwell systems.
A generic, platform vendor-independent, guide for initial Multi-node NVLink systems.
NVDebug is a comprehensive log collection and aggregation tool suite designed for NVIDIA's Multi-Node NVLink Systems that streamlines the debugging process by gathering all targeted logs in a single operation.
Partition Guides
Latest version of the GB200 partitioning guide.
The GB200 partitioning guide provides an overview of partitions in NVIDIA GB200 systems.