NVIDIA has released R310.14 dirvers for Linux. An interesting feature is a "OpenGL threaded optimizations". __GL_THREADED_OPTIMIZATIONS environment variable controls offloading of the driver's CPU computation to a worker thread. It should boost performance of the computation intensive applications, but decreases performance when using state-reading commands. In any case very interesting new feature.
Is there any similar option on Windows, or is it planned for the foreseeable future?