OP
mike_linux
Active Member
Ok, but I don't believe that this driver supports cuda
.
None of the chosen Nvidia drivers work
- Thread starter mike_linux
- Start date
.
D
Deleted member 108694
Guest
If you are just wanting to use CUDA have a look here - https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html
Oh, no... No, it definitely doesn't. But you can still get CUDA running.
If you want supported drivers, you're gonna have to fork out for a supported piece of hardware - or return to a state where your hardware was supported. Given the lengths of support out there in the Linux world, that's probably your second best solution. Like, 18.04 ends up with like 10 years of support with ESM enabled. I think it's 10 years, something like that.
Yeah, it's 10 years... I checked.
I'd have to check how long the kernel is supported, but it's an LTS kernel as memory serves. So, you can keep it going for quite a while, so long as you don't change the kernel's major version to an unsupported kernel.
If you want supported drivers, you're gonna have to fork out for a supported piece of hardware - or return to a state where your hardware was supported. Given the lengths of support out there in the Linux world, that's probably your second best solution. Like, 18.04 ends up with like 10 years of support with ESM enabled. I think it's 10 years, something like that.
Yeah, it's 10 years... I checked.
I'd have to check how long the kernel is supported, but it's an LTS kernel as memory serves. So, you can keep it going for quite a while, so long as you don't change the kernel's major version to an unsupported kernel.
OP
mike_linux
Active Member
I have tried to look which versions where saved when I did my Timeshift backup (everything worked well until then) and got this output after issuing
Output
As I can see there is also the kernel version 5.11 listed and 460 nvidia driver version. Maybe the number after the kernel version e.g. 5.11.0-38-generic is also important. Hope somebody can give some solution based on the output.
OP
mike_linux
Active Member
@Lord Boltar I am trying to install cuda using the tutorial you mentioned but at the last step:
I get the following error message in my output
What are unmet dependencies and how can I fix this?
Last edited:
Unable to correct problems, you have held broken packages
After upgrading from 10.04 to 12.04 I am trying to install different packages. For instance ia32-libs and skype (4.0). When trying to install these, I am getting the 'Unable to correct problems, ...
askubuntu.com
D
Deleted member 108694
Guest
you can install freeglut3-dev and see if that helps
once installed then look at instructions here - https://developer.nvidia.com/cuda-d...tu&target_version=20.04&target_type=deb_local
once installed then look at instructions here - https://developer.nvidia.com/cuda-d...tu&target_version=20.04&target_type=deb_local
Last edited by a moderator:
D
Deleted member 137406
Guest
@mike_linux
Not meaning to PO anyone etc but after all of this and no joy I'd just go ahead and install Linux Mint 20.x since it comes with the 5.4 kernel.
After installing the updates go to the Additional Drivers Manager and see if the desired Nvidia Proprietary driver is available I bet it is.
My 2 bits worth.
Not meaning to PO anyone etc but after all of this and no joy I'd just go ahead and install Linux Mint 20.x since it comes with the 5.4 kernel.
After installing the updates go to the Additional Drivers Manager and see if the desired Nvidia Proprietary driver is available I bet it is.
My 2 bits worth.
OP
mike_linux
Active Member
I have installing freeglut based on this tutorial and also restarted the system.
but the problem still remains the same
The funny thing ist that here at Hardware for Linux they mention that the GTX 860m graphics card supports the driver versions from 340.107 to 460.32.03. but in reality this does not work.
EDIT
Finally, I have managed to install cuda-11 (I did not do anything els but suddenly the installation process proceeded). Nevetheless, you cannot obtain any information about cuda for example by issuing nvcc --version because you need to install the nvidia-cuda-toolkit which also installs the latest Nvidia driver and you get the same problems as bef
My current kernel version that works is 5.13.0-51-generic.
Based on that I condude as mentioned before that the GPU drivers for old graphic cards do not work properly on newer versions of Ubuntu (e.g. Ubuntu 20.04 LTS). I have also checked to install the server versions of the GPU using this guide but is also didn't worked.
How can I obtain a boot log while booting to see why the system hangs? Maybe this will provide me suitable information to fix the problem.
Last edited:
D
Deleted member 108694
Guest
To see information from your most recent boot, run
To see the details that were logged during a previous boot, just in case you need to compare them with your most recent, run
Code:
journalctl -b
Code:
journalctl --list-boots
OP
mike_linux
Active Member
THX but is there any flag which I can set to this command to list the GPU access? Or even better when the GPU gets the first time accessed (i.e. the GPU driver gets loaded)? The list is huge and you get confused with all these displayed services.
D
Deleted member 108694
Guest
If there is I do not know maybe someone else can chime in on this
Last edited by a moderator:
Maybe something with grep? 'journalctl -b | grep <some word that narrows down the results>' kinda deal?
I don't use Nvidia for anything, as a general rule. So, I can't narrow it down from there. OP would have to figure out the keyword themselves.
I don't use Nvidia for anything, as a general rule. So, I can't narrow it down from there. OP would have to figure out the keyword themselves.
OP
mike_linux
Active Member
Ok guys, thanks fpr your helps but I'm slowly running out of ideas and I don't know ich after the update old graphic cards of nvidia work (my case is one proof). I need GPU support for AI development so I will connect my visual studio code compiler to google colab to leverage the power online. I don't know at the time what else to try. If you have any ideas in future please pm me. Thank you again for your extraordinary help. Now let's have a beer 
.
.
OP
mike_linux
Active Member
Hi guys, finally I have managed to install the nvidia driver 510 without freezing my system. The steps I had executed to fix this issue are:
1. At first there were none of the nvidia drivers installed. Then I had cleaned up all my broken and unused packages with
3. To get the grub menu shown I pressed ESC while the system started
4. Here I selected the 5.4.0 kernel (kernels greater then 5.4. did not work e.g. 5.13)
5. The system boots up without freezing anymore.
Despite the fix, $ nvidia-smi does not work and the follogwing error message gets displayed
Nevetheless, there is a workaround to get the supported nvidia drivers (the weired thing is that version 515 gets listed also)
If you know a better command to get all the GPU specifications pls post it.
1. At first there were none of the nvidia drivers installed. Then I had cleaned up all my broken and unused packages with
2. Then I installed the nvidia driver 510 via the software & update dialog in Ubuntu and rebooted the system
3. To get the grub menu shown I pressed ESC while the system started
4. Here I selected the 5.4.0 kernel (kernels greater then 5.4. did not work e.g. 5.13)
5. The system boots up without freezing anymore.
Despite the fix, $ nvidia-smi does not work and the follogwing error message gets displayed
Trying to install the latest nvidia driver 515 stopped the system again. So I reinstalled the nvidia driver 510 which does not stop the system. As you can see in the following screenshot the system works with the driver version 510.
Nevetheless, there is a workaround to get the supported nvidia drivers (the weired thing is that version 515 gets listed also)
If you know a better command to get all the GPU specifications pls post it.
Similar threads
