8

got a rtx 2080ti 2 days ago, previous was using two gtx 1080, and run my tensorflow program with no problems, after replaced with rtx 2080ti, the system cannot find driver for this device, and the lshw shows me the following

  *-display UNCLAIMED       
       description: VGA compatible controller
       product: NVIDIA Corporation
       vendor: NVIDIA Corporation
       physical id: 0
       bus info: pci@0000:01:00.0
       version: a1
       width: 64 bits
       clock: 33MHz
       capabilities: pm msi pciexpress vga_controller cap_list
       configuration: latency=0
       resources: memory:de000000-deffffff memory:c0000000-cfffffff memory:d0000000-d1ffffff ioport:e000(size=128) memory:c0000-dffff

i tried manually install some drivers from PPA, but nothing happens.

Is that my computer problem or it is indeed havent been supported yet.

my system version is 18.04

thank you

  • Which PPA have you tried? Are you trying to use the nvidia proprietary drivers or the open source nouveau driver? – Kristopher Ives Oct 1 '18 at 0:50
  • Tried ppa:graphics-drivers/pap with the 390 and 384 driver not nouveau I also noticed when I tried to reinstall the system nouveau says unknow chipset – stepstone Oct 1 '18 at 1:19
  • 1
    This article says that driver version 410 is needed for Turing cards. It also tells how to install a CUDA package that comes with that driver. – Chai T. Rex Oct 1 '18 at 1:23
  • Thank you man. U saved my day thought ppa is synchronized with official drivers. Downloaded driver from Nvidia websit and with anaconda everything is perfect right now – stepstone Oct 1 '18 at 2:01
6

I wrote a complete article on how to install drivers and cuda for RTX 2080 series GPUs on Ubuntu, mainly because it can be a time consuming task for some(most?) users. You are advised to follow the article along with the following answer.

I observed the same issue with both Ubuntu-16.04 and 18.04. In Ubuntu-16.04 the machine hadn't found the correct Nvidia driver and it had been using open source Nouveau display driver. Hence you're required to download and install the proper driver from Nvidia.

Now install the drivers as follows.

1) cd Downloads/

2) ls
#(NVIDIA-Linux-x86_64–410.57.run #output of ls)

3) chmod +x NVIDIA-Linux-x86_64–410.57.run
#(to get permission to execute the run file)

4) sudo ./NVIDIA-Linux-x86_64–410.57.run --no-x-check

Note: --no-x-check flag is important. Otherwise you will have to disable the x-server and log out from the GUI.

Once the drivers are installed reboot the machine and verify the installation with nvidia-smi command on a terminal. It will show you GPU and installed driver information.

The next part is installing cuda which can be a pain at times. You will have to log out from the GUI and perform all the actions on a terminal.

Download your relevant cuda run file from this link. Then create a file called blacklist-nouveau.conf in your home directory and add following lines to this file.

blacklist nouveau
options nouveau modeset=0

Now we have to log out from the GUI by pressing (ctrl+alt+f1) to install cuda. Enter your username and password to go to the terminal.

Now follow the steps below.

1) sudo service lightdm stop

2) sudo -i

3) sudo cp /home/avin/blacklist-nouveau.conf /etc/modprobe.d
#change avin with your user name

4) sudo update-initramfs -u

5) exit

6) cd Downloads/

7) md5sum cuda_10.0.130_410.48_linux.run #(Optional)

8) sudo sh cuda_10.0.130_410.48_linux.run

Cuda installation wizard will appear. Follow apparent steps, along with the article. Once the installation is finished run reboot on this terminal.

Once the machine is rebooted add cuda path to the .bashrc.

1) sudo gedit ~/.bashrc

2) Add following 2 lines at the end of the .bashrc file

export PATH=/usr/local/cuda-10.0/bin${PATH:+:${PATH}}

export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

3) source ~/.bashrc

Now run nvcc -V on a terminal which should yield an output as follows.

installation message

Now that the drivers and cuda are installed you might want to install tensorflow gpu version. Here is a good tutorial for the very same task.

  • 1
    The older version (link supplied) did not work with Ubuntu 18.10 latest updates. The most recent .run file worked fine after rebooting. I recommend everyone get the latest build for the drivers when you do this. – will Jun 2 at 6:21
1

i just had the exact same problem, and right now i am running a tensor flow model to test it and it seems fine

all i needed to do is to install the driver from Nvidia website and run it with sudo.

here you go the link https://www.geforce.com/drivers

i haven't done any benchmarking for the performance yet but everything seems fine

and a very bad downside for this approach is that everytime you will have a kernel update you will have to reinstall the driver!

  • 1
    Can you tell me your ubuntu version, driver version and cuda version? – Mohsin Bukhari Nov 6 '18 at 10:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.