Getting Started with Multi-GPU Scaling: Distributed Libraries

Explore multi-GPU scaling strategies in this 41-minute technical session from NVIDIA GTC 2025. Learn how distributed libraries can simplify the challenging process of scaling beyond a single GPU. Senior Software Engineer Mads Kristensen guides you through various approaches taken by multi-GPU libraries and examines their impact on user experience. Designed for intermediate-level developers, this talk covers CUDA and CUDA-X technologies, providing practical insights into accelerated computing libraries. Discover how to leverage NVIDIA's tools for distributed computing and optimize your applications for multi-GPU environments.