Nvidia has announced that the next major release of its CUDA Toolkit, likely version 13.x, will no longer support GPUs based on the Maxwell, Pascal, and Volta architectures. The deprecation applies specifically to the compute side, meaning these GPUs will continue to receive regular GeForce driver updates. However, developers will no longer be able to create new CUDA applications for these outdated architectures once the transition to the new toolkit version is complete. This decision is detailed in the official release notes for CUDA 12.9, which clarify that CUDA 12.x and earlier versions will still support developing applications for Maxwell, Pascal, and Volta GPUs. The deprecation primarily affects offline compilation and library support. Future CUDA compilers (nvcc) will not generate machine code compatible with these older GPUs, and CUDA-accelerated libraries such as cuBLAS and cuDNN will no longer offer support for them. While the exact date of the next major CUDA Toolkit release has not been announced, the interim 12.9.x releases are expected to continue supporting these architectures. This significant change signals Nvidia’s commitment to pushing the boundaries of AI and high-performance computing by focusing on more modern and capable hardware. Nvidia’s Maxwell architecture was introduced in early 2014 with GPUs like the GTX 745, GTX 750, and GTX 750 Ti, as well as the mobile GTX 800M series. It was also featured in the original Nintendo Switch’s Tegra SoC. Pascal followed in 2016, known for its iconic GTX 1080 Ti and its robust presence in both consumer and workstation markets, including the Quadro P-series and Tesla P4 accelerators. Volta, introduced in 2017, was pivotal in Nvidia’s AI acceleration efforts, marking the debut of Tensor Cores. The GV100, a Volta-based chip, is the second-largest in Nvidia’s lineup, surpassed only by the massive GA100 (Ampere) chip. Volta paved the way for subsequent architectures like Turing, Ampere, Hopper, and the latest Blackwell, contributing significantly to Nvidia’s growth and current valuation of nearly $2.8 trillion. Developers are advised to plan their migration to newer Nvidia architectures, as the upcoming CUDA Toolkit versions will not support Maxwell, Pascal, and Volta GPUs. This shift underscores the rapid pace of technological advancement in the GPU and AI industries, where continuous innovation is crucial for maintaining competitive edge.