NVIDIA Launches CUDA Toolkit 11.8, Promises Easier and Quicker Upgrades for Jetson Users
With a new compatibility package, Jetson users can upgrade to a new CUDA release without having to refresh the entire operating system.
NVIDIA has announced the latest version of its CUDA Toolkit general-purpose graphics processing unit (GPU) software — alongside improvements in how users of its Jetson embedded products can keep the software stack up-to-date.
"This release is focused on enhancing the programming model and CUDA application speed-up through new hardware capabilities," NVIDIA staffers write in their announcement of the latest CUDA Toolkit release, version 11.8. "New architecture-specific features in NVIDIA Hopper and Ada Lovelace are initially being exposed through libraries and framework enhancements."
Among the other new features in CUDA Toolkit 11.8, which is designed for those looking to offload highly-parallelizable workloads from the system's CPU to an NVIDIA GPU for a major gain in performance, are support for lazy loading of CPU modules to reduce latency, improved signal handling in Multi-Process Service (MPS) environments, and enhanced versions of the Nsight Compute and Nsight Systems profiling tools — plus the aforementioned support for next-generation Hopper and Ada Lovelace architecture processor, which will be finalized in CUDA 12.
For those working in the company's Jetson embedded hardware ecosystem, though, there's bigger news: CUDA Toolkit 11.8 brings NVIDIA's first support for the ability to upgrade CUDA versions without having to completely refresh the entire operating system.
"Starting from CUDA 11.8, [NVIDIA] has introduced an upgrade path that provides Jetson developers with an option to update the CUDA driver and the CUDA Toolkit to the latest versions," the company explains. "These upgrades are made possible by the introduction of the CUDA driver upgrade (also referred to as the CUDA compatibility package.) Without this package, you were previously limited to the functionality provided by the default CUDA driver that was packaged in the Jetson Linux BSP [Board Support Package]. You had no mechanism to upgrade to the latest CUDA driver and toolkit."
"With this package," NVIDIA's staffers continue, "Jetson users who have invested in long and thorough validation cycles for the existing Jetson Linux BSP can upgrade to the latest CUDA versions. This upgrade is done over the existing Jetson Linux BSP, keeping it unchanged. We are working towards a future where Jetson developers can migrate to newer versions of the compute libraries without upgrading Jetson Linux. This CUDA feature that enables upgrading CUDA is a step in that direction."
Instructions on taking advantage of the new feature to upgrade the CUDA release on any compatible Jetson board or module are available on the NVIDIA website.
