With CUDA 13.2 that is now shipping, NVIDIA has provided official support for Red Hat Enterprise Linux compatible distributions/downstreams like AlmaLinux to CUDA. With this official NVIDIA CUDA support for these RHEL-compatible distributions, NVIDIA is also allowing the NVIDIA packages to be distributed directly from the OS package repositories.
AlmaLinux relayed the word today of NVIDIA now officially supporting RHEL-derived but compatible distributions with CUDA:
“NVIDIA has added official support for enterprise Linux compatible distributions, including AlmaLinux, ensuring that our users can get support from NVIDIA when using NVIDIA infrastructure.
We also now have an agreement with NVIDIA that allows us to distribute NVIDIA’s packages directly from our repositories, rather than users getting the drivers separately from us.
…
Right now, when NVIDIA releases a new driver, there is a very slight delay in our release of the updates, which can result in brief version mismatches for users of NVIDIA hardware and AlmaLinux. Shipping the open source drivers along with the userspace and CUDA components ourselves means that all the packages are updated in tandem. There won’t be any delay between the release of the two package sets, ensuring the versions are always in sync.”
More details on this shift can be found via the AlmaLinux blog.
As for the newly-released CUDA 13.2, the release notes mention host tasks having a new spin-wait dispatch mode to reduce execution latency, new PTX features, support for new host compilers, improved C++20 standards conformance with the NVCC compiler, and a variety of other new features and improvements.
