NVIDIA Provides Clarity on Open Linux Nucleus Modules for GPU Systems
Andy Ritger, who leads the development of drivers for unix-like OS in Nvidia, recently addressed queries regarding the Translation of NVIDIA Proprietary Drivers to the use of open linux nucleus modules on GPUs based on microarchitectures, starting with tring. During the discussion, several key points were highlighted.
- When asked about achieving functionality parity between open and proprietary drivers, NVIDIA mentioned that the upcoming release of drivers 560 will align approximately with proprietary modules. However, limitations such as the inability to use RTD3 (Run Time D3) dynamic power management mechanism with open modules on older GPU generations before AMPERE were noted. Work to achieve complete parity is ongoing. Additionally, long-standing issues related to VRR on laptops will be addressed in the release of 560.
- NVIDIA clarified that they do not plan to integrate open modules into the main Linux nucleus.
- In response to inquiries about supporting open Nouveau and NVK drivers, NVIDIA recommended using proprietary drivers for optimal performance. They also provided separate Open nuclei modules for use. While some documentation and patches have been shared with the developers, significant support is not currently available.
- Regarding the support for Nouveau open driver with closed components like CUDA, AI, RT/PT, DLSS, and OPTIX, NVIDIA stated that compatibility is not feasible at present. The use of Nvapi/Nvidia-SMI features with Nouveau is also limited. However, changes in the API Nvapi license have enabled projects like Wine and Proton to create their own implementations for gaming purposes.
- NVIDIA acknowledged minimal employee involvement in Nouveau development but indicated that some employees are contributing to discussions and development efforts.
/Reports, release notes, official announcements.