NVIDIA to Drop Support for Older GPU Architectures in CUDA 13.0
NVIDIA's upcoming CUDA 13.0 Toolkit will cease offline compilation support for GPU architectures older than compute capability 7.5, effectively phasing out Maxwell, Pascal, and Volta generations. The change reflects the company's focus on modernizing its software stack amid rapid advancements in parallel computing.
The decision impacts developers relying on legacy hardware for CUDA-based applications. NVIDIA's driver ecosystem—spanning New Feature, Production, and Long-Term Support branches—continues to prioritize cutting-edge capabilities over backward compatibility. Market observers note this could accelerate adoption of newer GPU models in both enterprise and cryptocurrency mining sectors.