Folding Forum

Drivers which switch between Intel and nVidia add a significant level of complexity and have been known to cause a problem, though I vaguely remember somebody might have solved that.

If the client sees the Intel chipset rather than the nVidia hardware& driver, it's obviously not going to work, but don't ask me for any other details. That's the absolute limit of my understanding.

@monok: This guide recommends to use nvidia-prime for dual intel/nvidia optimus GPU.
There you can switch to make nvidia active and folding can use it.
http://askubuntu.com/questions/412452/g ... ime-gt650m

Moved off-topic thread on troubleshooting a FAH install on a Linux system to - viewtopic.php?f=96&t=29352.

FYI: No need to edit /etc/default/grub as long as GrUB is working properly.

For either init or systemd Linux systems all* you have to do is:

* When the GrUB menu appears at boot, navigate to the Linux OS you want to modify and press e
* Navigate down to the linux command line and place a 3 at the end of it
* Press F10 to boot

It boots into command line interface(CLI) which is actually init level 3.
Then do the nVidia install steps.

Note that " 3" works any time one needs to boot to 'text mode', i.e., the CLI(init 3). This is very handy for updating core parts of the OS, especially if SystemD is one of those parts, .

{ *all assumes a std GrUB install that is not hidden, write or password protected. If any of those are true then at least password is needed ... but for the people who use such a GrUB, this guide is prbly not needed anyway. }

This sticky might benefit from an edit. Please see my post, viewtopic.php?f=80&t=35249

Long story short: opencl does not work in Linux, unless you are using the "correct" GCC compiler. If you have been upgraded to GCC-9, FAH will fail because it cannot find an opencl device, or similar message. You must rollback your compiler to GCC-8 for the time being. If/when nVidia updates their drivers to use GCC-9 and G++-9, then we can forget about checking the compiler versions.

Hello Community
Since I also had some difficulties to get the (latest) .run-file for (any) NVIDIA card installed, also on Debian releases, I just want to share my experiences.
1st) make sure you are (definetely) blacklisting any NVIDIA related (open-source) modules like nouveau (and maybe optimus/bumblebee)
2nd) don't forget to deactivate those modules also in initram BEFORE rebooting (by doing an 'update-initramfs -u -k all')
3rd) make sure to reboot your system with absolutely only access via console/not any display manager related modules started (boot options are 'vga=off text console 3', just to be really, really sure)
4th) check twice steps 1st) - 3rd)
5th) from CLI (console) do a 'sudo (ba)sh ./<cuda-file>.run' or: '.. ./<NVIDIA-driver-xxx>.run'. This should extract, compile and install your driver file
6th) (not necessary but helpful): do compile your samples (only available if complete CUDA-xxx.run file was installed); then you can use the 'deviceQuery' command to show infos about/show connection to available NVIDIA cards
(remember to add its (complete) path to your env variable ${PATH}, eg. in .bashrc in this case)
7th) in case of error: check
- if there (really) are no modules loaded, related to NVIDIA, optimus, nouveau or bumlbebee in 'etc/modules', '/etc/initramfs/modules', '/lib(32|64)/modules', '/usr/lib(32|64)/modules'; do also a 'lsmod | grep -i nvidia|nouveau|optimus|bumblebee'; there should no module be found at this step
- if still there are: check if you did blacklisting correctly AND did update initramfs
- if you did blacklisting correctly, update initramfs, make sure no nvidia related modules are still loaded, BUT could still not install driver/compile CUDA env, try to do an 'invidia-driver-uninstall', or search manuall for any loaded, nvidia-related modules (kind of more intensive/experienced way)
8th) if all those steps do not help, check if you really install THE CORRECT linux-headerfiles and sources;
9th) there may be a complete, new installation necessary of Debian AND immedaitely a fresg install of the nvidia driver.
10th) in case of 9th):
- do also add i386 architecture (needed for some nvidia 32bit-libs): 'sudo dpkg --add-architecture i386' (this step MUST BE done at first; next step depends on this)
- do also install linux-headers(-amd64), linux-source, firmware-linux, build-essential, ( not really necessary: g++, gcc)
- try now to install the nvidia driver as mentioned above, hope it will do now.
Cheers, good luck and in any lucky case, have fun.
Roger

thanks.

I've had to rebuild several installations by not blacklisting correctly. Often I have been able to find install packages that include the blacklisting.

I have a problem with my Nvidia card, who can help me?

Marcostone wrote:I have a problem with my Nvidia card, who can help me?

Welcome to the forums.

That is a fairly broad question !!! ... if you could expand with some explanation as to what type of problem we might be able to help?

The only issue I have faced with .run installation is that I also need to turn off the Linux header automatic updates. Sometimes these auto updates have caused GPU driver updates as well resulting in software incompatibility issues.

You can just Google to know how to stop linux kernel auto updates. Hope that helps!