Nvidia fully transitioned to open-source GPU kernel modules in its driver stack.

In May 2022, NVIDIA released open-source Linux GPU kernel modules with dual GPL and MIT licensing through the R515 driver, initially targeting datacenter compute GPUs while GeForce and Workstation GPUs remained in alpha state. NVIDIA committed at that time to providing more robust and fully-featured GeForce and Workstation Linux support in subsequent releases, with the intention for the open-source modules to eventually supplant the closed-source driver. The unified driver architecture across all NVIDIA GPUs—from desktop and laptop systems to cloud-based AI workloads—was identified as critical to getting the transition right.

After two years of development, NVIDIA has achieved equivalent or better application performance with the open-source GPU kernel modules while adding substantial new capabilities including heterogeneous memory management support, confidential computing, and coherent memory architectures. Based on this progress, NVIDIA is making the transition fully to open-source GPU kernel modules the default in the upcoming R560 driver release.

Compatibility varies by GPU architecture. Cutting-edge platforms such as NVIDIA Grace Hopper and NVIDIA Blackwell require the open-source GPU kernel modules, with proprietary drivers unsupported on these platforms. For newer GPUs from the Turing, Ampere, Ada Lovelace, or Hopper architectures, NVIDIA recommends switching to the open-source modules. Conversely, older GPUs from the Maxwell, Pascal, or Volta architectures remain incompatible with the open-source modules and should continue using the proprietary driver. Systems with mixed deployments containing both older and newer GPUs should also continue using the proprietary driver. NVIDIA provides a detection helper script (nvidia-driver-assistant) to guide users in selecting the appropriate driver for their hardware.

Installation workflows have been updated to reflect this transition. Beginning with the CUDA 12.6 release, installing CUDA or the NVIDIA drivers via package manager or .run file now defaults to the open-source driver for compatible GPUs, while still offering UI toggles and command-line options to select between proprietary and open-source drivers as needed. Users installing through the .run file interface can select the kernel module type via an ncurses menu, or specify preferences using command-line overrides for automation tools such as Ansible.