CUDA: New Features and Beyond | NVIDIA GTC 2025
NVIDIA CUDA Team
NVIDIA GTC 2025 · Session
In this comprehensive talk from NVIDIA GTC 2025, Stephen Jones, a CUDA Architect at NVIDIA, delves into the past, present, and future of the **CUDA** platform. The presentation, aptly titled "CUDA: New Features and Beyond," highlights the platform's evolution from a nascent C-based GPU programming model to a sprawling, multi-layered ecosystem essential for accelerated computing across diverse domains. Jones emphasizes that CUDA is far more than just CUDA C++, encompassing a vast array of SDKs, libraries, runtimes, and developer tools that collectively enable high-performance computing on NVIDIA GPUs.
AI review
A competent overview of where CUDA is heading — Pythonic runtime wrappers, the Coupile tile programming model, JIT compilation tooling, and early sketches of datacenter-scale CUDA. Stephen Jones clearly knows this stack from the ground up and has 17 years of receipts to show for it. But this article summary, and apparently the talk itself, stays mostly at the level of feature announcements and strategic framing rather than implementation depth. The Coupile result — Llama 3.1 within 10% of cuDNN in a few weeks — is the most interesting claim in the whole piece, and it gets about two…