- Setup last virtualgl (dl last stable version like 2.5 at http://sourceforge.net/projects/virtualgl/files/ )
dpkg -i virtualgl*.deb
- Setup bumblebee for nvidia :
sudo apt-get install linux-headers-$(uname -r) freeglut3-dev libxmu-dev libpcap-dev bumblebee-nvidia primus
-
Setup via additional drivers UI Nvidia latest/tested version (352.69 currently)
-
Configure bumble bee :
in
/etc/bumblebee/bumblebee.conf
:Driver=nvidia
and replace all lines withnvidia-current
with your nvidia version (e.gnvidia-352
)then
sudo service bumblebeed restart
-
Setup nvidia-modprobe
sudo apt-get install nvidia-modprobe
(auto module loader for non-root users) and nvidia-primesudo apt-get install nvidia-prime
-
Download cuda 7.5 (
.run
mode, and answerno
to "setup NVIDIA driver 3XX ...") and NEVER use nvidia driver for X11 server (usesudo prime-select intel
if you need to disable nvidia) -
Add to your PATH env variable :
/opt/VirtualGL/bin:/usr/local/cuda-7.5/bin
-
Add to
/etc/ld.so.conf
:/usr/local/cuda-7.5/lib64
then runsudo ldconfig
-
Add lines to file at startup (e.g /etc/rc.local)
NVIDIA_VERSION=352
update-alternatives --set x86_64-linux-gnu_gl_conf /usr/lib/nvidia-$NVIDIA_VERSION-prime/ld.so.conf
update-alternatives --set i386-linux-gnu_gl_conf /usr/lib/nvidia-$NVIDIA_VERSION-prime/alt_ld.so.conf
prime-select intel
- Now reboot your laptop
- Then you can test your CUDA setup :
cd ~/NVIDIA_CUDA-7.5_Samples/1_Utilities/deviceQuery
make
optirun ./deviceQuery
- How to use with cudaHashCat64 or another runnable compatible with CUDA ?
alias cuda='optirun --no-xorg'
#then:
cuda ./cudaHashCat64 -b
- search nvidia installed packages :
sudo dpgk --get-selections | grep 'install' | grep 'nvidia'
- search libcuda installed packages :
sudo dpgk --get-selections | grep 'install' | grep 'cuda'
- for each, mark hold in apt :
sudo apt-mark hold <package>