LinuxMint 17.1 rebbeca, OpenGL Nvidia-card, Blender 2.72

Support Technical Support
21

So for resume if you use a blender release from BlenderFoundatiuon.org:
=================================================== To analyze
1 - Open the terminal and copy here the result of the next commands to give us information about your system:

To obtain the setup of your graphic card:
lspci | grep -i vga

To obtain the version of your driver Nvidia:
cat /proc/driver/nvidia/version

To obtain the list of packages to installed with your synaptic setup:
sudo dpkg -l | grep -i nvidia

===================================================== To actions
1 - Remove all “nouveau firmware” drivers from your linux:
nouveau firmware
xserver-xorg-video-nouveau-dbg

2 - Add the Xorg Edgers PPA repository to your System->administration>Applications sources or:
https://launchpad.net/~xorg-edgers/+archive/ubuntu/ppa

3 - With Terminal:
$ sudo apt-get update

4 - install your nvidia driver “nvidia xxx” + " libcuda1" + “nvidia-libopencl1” + “nvidia-opencl-icd” for your current release of Nvidia driver with your synaptic manager

Once all installed, disable the Xorg Edgers repository in your system->administration>Applications sources.

5 - download and install the “nvidia-modprobe” deb package:
https://launchpad.net/ubuntu/+source/nvidia-modprobe

6 - Go to your applications manager and remove all others release of Nvidia that isn’t xxx ok for your graphic card!

7 - Last: Reboot the system.

===============================================
If you want to have Blender from a PPA always updated:
My friend Thomas Schiex has made a Trusty version of his Blender compilation now (he want to replace IRIE PPA stoped since 6 months):
sudo add-apt-repository ppa:thomas-schiex/blender
sudo apt-get update
sudo apt-get install blender

and then you must download/install the CUDA Toolkit 6.5.14 if you have a Nvidia graphic card+CUDA cores:
https://developer.nvidia.com/cuda-downloads

All is ok since blender 2.73 RC (with OSL in OpenCL).
A great Thanks to Thomas!

EDIT: So minimal config on AMD64:


ii  libcuda1-340                  340.65-0ubuntu1~xedgers14.04.1        
ii  nvidia-340                     340.65-0ubuntu1~xedgers14.04.1        
ii  nvidia-340-uvm             340.65-0ubuntu1~xedgers14.04.1        
ii  nvidia-libopencl1-340     340.65-0ubuntu1~xedgers14.04.1        
ii  nvidia-modprobe            340.24-1                             
ii  nvidia-prime                                   
ii  nvidia-settings                340.65-0ubuntu1~xedgers14.04.1 
+
ii CUDA-Toolkit                   6.5 mini
22

Good summary, surely this will be useful for other users too.
But in reality this should be easier and work fine automatically installing drivers from the driver manager or “additional drivers”. At least in kubuntu, I do not need to do anything else, just install the “nvidia-modprobe” package after installing the driver.

23

Yes YAFU but if someone make a mistake like me…I have changed libs or others I don’t know…
Your explanations offer rationally to see clearly the solution.
…to reset the compass on time!

And I am on Linuxmint, that’s different as kubuntu.

24

@spirou4D

I did exactly what you do on mint 17 but I got black screen. Totally black no nothing.

25

I’m running Linux Mint 17.1 and have gotten the official Nvidia drivers working. Several people were having the same issue over in the Octane forums, here is what I posted over there to fix the problem. Now I want to use the official drivers and not what is in the repo. Mostly because the repo is always behind and appears to have other issues, like missing CUDA, etc.

I am now running Linux Mint 17.1 and it’s working with version 343.36 of the Nvidia driver, also I think I have a good method of installing the driver straight from Nvidia and not using the edgers ppa. Here is what I did:

This is from Michel on the Linux Mint forums (with my modifications):

  1. Get a copy of the driver install file from the Nvidia driver download web page.

http://www.nvidia.com/Download/index.aspx?lang=en-us

  1. You need to edit/create the /etc/modprobe.d/disable-nouveau.conf file. I use vi in a terminal or you can use gedit if you are more comfortable with it:

Code:
sudo vi /etc/modprobe.d/disable-nouveau.conf

or
Code:
sudo gedit /etc/modprobe.d/disable-nouveau.conf

  1. Add the following lines to the file and save it:

Code:
blacklist nouveau
options nouveau modeset=0

  1. Then run this command:

Code:
sudo update-initramfs -u

  1. reboot system
  2. Once the system has rebooted, do CTRL + ALT+ F1 to go to one of the virtual terminals and log in.
  3. Stop the current display manager which also stops X:

Code:
sudo service mdm stop

  1. cd to where the Nvidia installer is and run it with the following command:

Code:
sudo sh NVIDIA-Linux-x86_64-343.36.run

  1. Select 'yes" on every question
  2. Either reboot from there or run ‘sudo service mdm start’ and you should be good.

This has survived a reboot of my system and Octane can see both of my cards, including the 980. Everything appears to be working fine. :smiley:

When you need to install a new version of the driver, you should be able to log out and start on step 6 and go from there. I think the biggest problem was with the Nouveau driver conflicting with the Nvidia driver, you definitely need to blacklist it. :evil:

I can use my 980 with Cycles no problem, so it should work for you as well.

Jason

26

@Kevin If you saw the branch on which you sit, in fact you fall!

In #6 phase -> Go to your applications manager and remove all others release of Nvidia that isn’t 340!

…after to install the 340 driver necessarily!!!

27

@ Grimm

I failed to restart x server after reboot. Do you havr any idea?

28

Application manager do you mean synaptic?

29

No just Application manager (Gestionnaire d’applications in french).

For restart x server ; in Terminal ctrl +alt + T
sudo service lightdm start

take a look here:

30

I will stick with driver 304 for now. I found it strange that reverting to driver 304 in driver manager enable cuda while newer driver doesn’t. Anybody have any idea?

Btw does older driver have less cuda performance than new one?

31

Kevin, what desktop are you using? I’m using Cinnamon which uses mdm. If you are using Gnome shell then it’s gdm, other desktops use other display managers. Once you know which one to bring down and then bring up it shouldn’t be a problem, or you can just reboot to restart the dm.

32

@Grimm
I am using mint 17 cinnamon. I install my driver using driver manager. The driver 304 works just fine, but on newer driver version(331), blender won’t detect cuda.

@Spirou4D
I might be stupid but I can’t seems to find “application manager” I only have software manager. I using mint 17.

I don’t think I will try to update my driver, it is working and I don’t want to broke my os. I have worst nightmares when it come to installing driver.

33

Spirou4D and I are using two different methods for getting the drivers installed. He is using the repositories, I’m using the official driver installs from Nvidia. I find the drivers from the repositories to be mostly out of date so I like to stick to the official Nvidia releases. As it looks like you are installing the driver using the Driver Manager, you should install the “nvidia-331-updates” version, it has CUDA. The “nvidia-331” version of the driver does not.

The main reason I use the official driver install is because I have a 980 and support for it came in the 34x series of drivers. I’m currently running version 343.36 of the driver.

34

@Grimm
I never knew that update version have cuda, I will try it out. But does driver version effect cycles performance?

When I update my system by update manager, blender doesn’t detect cuda for nvidia-304 anymore. I don’t know why but I will update to nvidia-331-updates and try it out if cuda is detect.

35

@Grimm
I install nvidia-331-updates and cuda tool kit but blender doesn’t detect cuda. Should I install nvidia modprobe 340? I think I might run into trouble because it is version 340 not 331.

Are there modprobe for driver 331?

Edit: I install nvidia modprobe 340 on driver 331 just fine, reboot survive blankscreen issue etc. and blender now detect cuda. Problem solved! Wooooooo!

36

Excellent! :slight_smile:

37

On mint17qiana, ,GeForce gtx 660, Nvidia-331-updates, blender2.73, this instruction works. Thank you so much!

38

Just did a fresh install of Mint 17.2 Rafaela which comes with 346.72 installed with the driver manager and the CUDA tool kit with Synaptic from the main repos
blender picked it up and it was available on the system tab of User Preferences using blender 2.75.

Enjoy the Choice :slight_smile:

39

…solved…

40

[SOLVED] at least for me… i tried the nvidia 304-updates-driver now with nvidia-modprobe 340. Can use cuda now in blender, but after some tests within my material library cpu seems to be nearly 3 times faster. So, much ado about nothing. Sorry for disturbing…