Nvidia + Bumblebee + CUDA for Optimus

Hello all,

Not sure if anyone can help me here, but I will try and be as precise as possible in exposing my problem…

I’m trying to install Nvidia 331.67 + Bumblebee 3.2.1 and CUDA 5.5, in Kali Linux 1.0.7 amd64, in order to use the computational power of my GeForce GT 630M to run pyrit and hashcat for pen testing purposes. I have a Toshiba Satellite P850 with 2 video cards, Intel HD4000 and a Geforce GT 630M.

I’m following this guide https://forums.kali.org/showthread.php?21956-%28Guide%29-Installing-Nvidia-Bumblebee-CUDA-for-Optimus-enabled-Laptops (and it does work for others but not for me).

Everything worked well, until…

my return on that last command is:

root@kali:~# modprobe nvidia-uvm
ERROR: could not insert 'nvidia_uvm': No such device

If any of you have ideas as to what might be causing this problem, I would greatly appreciate your input.

On some laptops nvidia driver cannot initialize the GPU after it was powered down by bbswitch. Usually you need to use optirun for nvidia-related activities:

optirun --no-xorg sudo modprobe nvidia-uvm

(although after the above command the GPU will not be powered down automatically unless you manually unload the module and invoke optirun again). You should be able to use “optirun --no-xorg” to run CUDA applications, but note that you need some further manual configuration to make nvidia-uvm unloaded automatically, otherwise the GPU won’t be powered down after the application terminates. See Bumblebee bug 565 for details.

The Bumblebee issue tracker might be a better place for such queries.

Thank you monoid,

I was just dropping by to say that my problem is solved, then I saw your post…

A new Howto(post #25) was added in the first Howto I was following(see link in first post here), and that last Howto(post #25) worked beautifully.

(Now working with Nvidia 340.24 + Bumblebee 3.2.1 and cuda_6.0.37_linux_64)

Still not sure what was my prb, but after a week of intense search on how to do that procedure(Nvidia + Bumblebee + CUDA for Optimus), I don’t really care at this point ;) it w o r k s.