NVIDIA Corporation (NVDA.US) launches CUDA 13.1 and CUDA Tile, Huang Renxun claims it is the biggest upgrade in twenty years

NVIDIA Corporation (NVDA.US) has launched CUDA 13.1 and CUDA Tile, with the company's CEO Jensen Huang stating that this is the biggest advancement the platform has made since its launch about 20 years ago. NVIDIA Corporation engineers Jonathan Bentz and Tony Scudiero wrote in a blog post: "This exciting innovation introduces a virtual instruction set for module-based parallel programming, focusing on the ability to write algorithms at a higher level and abstracting details such as tensor cores and other specialized hardware." CUDA is a parallel computing platform and programming model developed by NVIDIA Corporation. It helps developers utilize the powerful computing capabilities of graphics processing units to enhance the speed of applications. The new primitive-based programming options allow developers to "fine-tune" the execution of their code, especially when dealing with multiple GPU architectures. NVIDIA Corporation engineers explained: "Tile-based programming allows you to write algorithms by specifying data blocks (or tiles), and then defining the computational operations performed on these tiles. You don't need to set the execution of the algorithm at the element level: the compiler and runtime will take care of all of that for you." CUDA Tile also has a version available in Python. NVIDIA Corporation plans to release a version compatible with C++ in the future. NVIDIA Corporation CUDA architect Stephen Jones said: "It has just been implemented in Python, and I think it integrates perfectly with artificial intelligence development frameworks. We intend to release CUDA Tile as an open-source project."