Intel GPU hung in my NVIDIA Optimus laptop with Debian Stable KDE

[Vincent Cheng]

On Wed, Jul 17, 2013 at 6:35 PM, Shervin Emami <shervin.emami at gmail.com> wrote:
> Thanks Vincent, I am new to Debian so actually I hadn't installed any
> backports because I thought it was difficult and cause instabilities, but
> now I realize backports are actually really easy and potentially more stable
> than my current situation.

While there's no guarantee of stability if you use backported packages
(you may have noticed that backports are not enabled or even mentioned
in sources.list by default), they're typically more stable and less
likely to cause trouble compared to 3rd-party packages/repositories,
or compiling from source. After all, backported packages (from
backports.d.o) are prepared and/or uploaded by Debian
Developers/Maintainers.

> I was using the instructions at "http://wiki.debian.org/Bumblebee" for
> Wheezy so I was getting bumblebee from "http://suwako.nomanga.net/", and
> using the Wheezy versions of Linux kernel and nvidia-current, nvidia-glx,
> nvidia-kernel-dkms, freeglut3, libxmu6, nvidia-cuda-toolkit packages, and I
> had set glx to "/usr/lib/mesa-diverted".

Just wanted to point out that "nvidia-current" is an Ubuntu-specific
metapackage, and if you have that installed, you may want to check to
see if you've installed any other Ubuntu packages by accident.

> So now I should uninstall bumblebee, bbswitch-dkms, virtualgl &
> virtualgl-libs (and remove it from sources.list), then install them from
> wheezy-backports while still using the Wheezy versions of everything else?
> Or should I also install the nvidia packages from backports or download
> latest from NVIDIA website?

Yes, stick with packages from wheezy and cherrypick what you need from
wheezy-backports (you can obviously still fetch packages from other
sources, but then you're less likely to receive support from official
Debian support channels).

There currently aren't any nvidia packages in wheezy-backports, but
the version in wheezy (304.xx) is new enough to work with bumblebee
and with your Quadro 1000M (which according to nvidia only requires a
driver >= 270.41 [1]), so you'll get all you need for Optimus support
with:

# apt-get install nvidia-glx bumblebee-nvidia primus

And if you're using your nvidia card purely for CUDA, not for running
any opengl applications with your nvidia card, you don't even need
virtualgl/primus.

Hope that helps!

Regards,
Vincent

[1] http://www.nvidia.com/object/linux-display-ia32-270.41.06-driver