Downloader and installer for the nVidia driver package (331.104)
Installing these packages can take a long time. These packages will automatically download the NVIDIA driver (+-70MB) from
http://download.nvidia.com/XFree86/Linux-x86_64 or http://download.nvidia.com/XFree86/Linux-x86 and save it in /usr/src.
The driver and libraries will be installed after the download.
The installer will install the nVidia files the Bumblebee way. The libraries will NOT interfere with any other libraries of a distribution and will be installed in a seperate nvidia directory. Usage will be controlled by editing the ld.so.conf(.d) and/or using the "ModulePath" var in xorg.conf.
You can also download the NVIDIA package yourself and copy it to the /usr/src directory. The installer will then skip the download part and just install the driver and libraries the Bumblebee way.
These rpms can be used by normal and Optimus enabled nVidia systems.
Tips and tricks
You can use the /usr/lib/nvidia/switch-libGL.sh or /usr/lib64/nvidia/switch-libGL.sh to force the use of the nVidia libGL* libraries.
This is only necessary if an application is compiled with hard linked library paths ( the rpath option was used ), for example "xbmc" from Packman.
A better solution is to start the offending application with:
LD_LIBRARY_PATH=/usr/lib/nvidia:/usr/lib/nvidia/vdpau <application_name> [ <arguments> ]
LD_LIBRARY_PATH=/usr/lib64/nvidia:/usr/lib64/nvidia/vdpau <application_name> [ <arguments> ]
and force the application to use the nVidia libs.
- Update to 331.104
* Added support for the following GPU:
* Fixed a bug that could cause the CUDA debugger to fail after
exiting X on systems with persistence mode enabled.
* Fixed a bug that could cause silent and intermittent failures when a CUDA
application writes to a peer device's memory with GPUDirect.
* Fixed a bug that could prevent 32-bit GPU-based applications from running
correctly on 64-bit systems when using GPUs with very large memory-mapped
* Fixed a bug that could cause poor system performance and hangs when running
large compute workloads on IOMMU-enabled systems.
* Fixed a bug that could cause a system crash when two or more processes try
to load the nvidia-uvm driver.
* Fixed a bug that could cause framelock status to be incorrectly reported in
the nvidia-settings control panel on some multi-GPU framelock configurations.
* Improved the performance of OpenGL applications that request a large number
of GPU mappings via the GL_NV_shader_buffer_load extension.
* Fixed a bug that could cause a system crash when killing applications that use
NVIDIA GPUs on multiple-GPU systems with IOMMU functionality enabled.
* Fixed a bug that caused the NVIDIA kernel module to leak memory for DMA
mappings on some configurations.