None of the chosen Nvidia drivers work

[mike_linux]

Hi guys, long time ago . Hope you're healthy and doing fine.

I am facing the following problem:
At first my system is the following
- Ubuntu 20.04.04 LTS 64bit
- Nvidia GTX860m (I know it's an old one)
- Intel® Core™ i7-4710HQ CPU @ 2.50GHz × 8
- RAM 8GB

Now to my problem I am facing. I am doing AI development and need so I need to install Tensorflow with GPU usage. For that I have to install also CUDA 11.2 and cuDNN 8.1 (it's specific for neural networks designed to leverage better GPU performance). After I had installed cuda and wanted to install cuDNN I needed to update my system via sudo apt update mentioned by the nvidia side. But then the problem occured. At first the system didn't booted anymore. Because of that I have tried the following steps and fixed the broken GPU driver.

I got into the Grub bootloader menu while starting the system by pressing the ESC key.
Then I droped to the terminal window, removed the driver and listed the supported drivers by issuing

sudo apt-get purge ^nvidia-
sudo apt list nvidia-driver-*

Which listed me the following

Listing... Done
nvidia-driver-390/focal-updates,focal-security,focal 390.151-0ubuntu0.20.04.1 amd64
nvidia-driver-390/focal-updates,focal-security,focal 390.151-0ubuntu0.20.04.1 i386
nvidia-driver-418-server/focal-updates,focal-security 418.226.00-0ubuntu0.20.04.2 amd64
nvidia-driver-418/focal 430.50-0ubuntu3 amd64
nvidia-driver-430/focal-updates,focal-security 440.100-0ubuntu0.20.04.1 amd64
nvidia-driver-435/focal-updates 455.45.01-0ubuntu0.20.04.1 amd64
nvidia-driver-440-server/focal-updates,focal-security 450.191.01-0ubuntu0.20.04.1 amd64
nvidia-driver-440/focal-updates,focal-security,focal 450.119.03-0ubuntu0.20.04.1 amd64
nvidia-driver-450-server/focal-updates,focal-security 450.191.01-0ubuntu0.20.04.1 amd64
nvidia-driver-450/focal-updates,focal-security,focal 460.91.03-0ubuntu0.20.04.1 amd64
nvidia-driver-455/focal-updates,focal-security,focal 460.91.03-0ubuntu0.20.04.1 amd64
nvidia-driver-460-server/focal-updates,focal-security 470.129.06-0ubuntu0.20.04.1 amd64
nvidia-driver-460/focal-updates,focal-security,focal 470.129.06-0ubuntu0.20.04.1 amd64
nvidia-driver-465/focal-updates,focal-security,focal 470.129.06-0ubuntu0.20.04.1 amd64
nvidia-driver-470-server/focal-updates,focal-security 470.129.06-0ubuntu0.20.04.1 amd64
nvidia-driver-470/focal-updates,focal-security,focal 470.129.06-0ubuntu0.20.04.1 amd64
nvidia-driver-495/focal 515.48.07-0ubuntu0.20.04.1 amd64
nvidia-driver-510-server/focal-updates,focal-security 510.73.08-0ubuntu0.20.04.1 amd64
nvidia-driver-510/focal-updates,focal-security,focal 510.73.05-0ubuntu0.20.04.1 amd64
nvidia-driver-515/focal 515.48.07-0ubuntu0.20.04.1 amd64

I have tried to install all of these drivers by issuing

sudo apt-get install nvidia-driver-xxx

Where xxx stand for the driver version e.g. 390, 460 etc. Also, I have tried to install each driver by using the Software updater GUI.

In addition, I have followed these tutorials to try fixing the problem
Which graphics driver in the Additional Drivers list should I choose?
How To Install Nvidia Drivers On Ubuntu 20.04 / Ubuntu 18.04
Ubuntu will not boot normally after nvidia driver install
Despite of all these steps I didn't figure out which of the drivers works, because everytime I installed one the system didn't boot and I had to got again and again into the grub menu and follow the steps above.

Hope you can help me to fix this problem. I have heard that nvidia driver problems are a common problem on Linux .

 

[Deleted member 137406]

Unfortunately you are SOL as the 5.4 kernel was the last to support the 340 and 390 series Nvidia drivers based on my own experience.

Nvidia and Linux is just a hit and miss and has mostly been a miss for users with older Nvidia cards.

The open source Nouveau graphics driver is it for those of us who have older Nvidia graphics cards.

Grab an AMD/ATI graphics card they work well with Linux based on my own experience.

Have a look at Linux Mint 20.x series it still uses the 5.4 series kernel.

Home - Linux Mint

Linux Mint is an elegant, easy to use, up to date and comfortable desktop operating system.

linuxmint.com

 

[mike_linux]

Unfortunately you are SOL as the 5.4 kernel was the last to support the 340 and 390 series Nvidia drivers based on my own experience.

Nvidia and Linux is just a hit and miss and has mostly been a miss for users with older Nvidia cards.

The open source Nouveau graphics driver is it for those of us who have older Nvidia graphics cards.

Grab an AMD/ATI graphics card they work well with Linux based on my own experience.

Thank you @Bartman I will keep this in mind and switch the next time to ATI cards if they support CUDA and cuDNN. After I issued this here

$ dkms status

I got this output which is that what you mentioned

rtl8192eu, 1.0, 5.13.0-48-generic, x86_64: installed
rtl8192eu, 1.0, 5.13.0-51-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-100-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-104-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-105-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-107-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-109-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-110-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-113-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-117-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-120-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-77-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-80-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-81-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-84-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-86-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-88-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-89-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-90-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-91-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-92-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-94-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-96-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-97-generic, x86_64: installed
rtl8192eu, 1.0, 5.4.0-99-generic, x86_64: installed

So based on your observation I conclude that with kernel version 5.4.0-109 the system worked well, but with the current version installed which is the 5.13.0-48. How can I downgrade this?

 

[Deleted member 108694]

Yep - Agree with @Bartman on this those older Nvidia cards are not supported with kernels later then 5.4
to install the 5.4 series

sudo apt install linux-generic

Then reboot and select a 5.4 kernel in advanced options of grub menu. Once booted to the 5.4 kernel then you can remove the 5.13 kernel

 

[Deleted member 137406]

Thank you @Bartman I will keep this in mind and switch the next time to ATI cards if they support CUDA and cuDNN. After I issued this here

I got this output which is that what you mentioned

So based on your observation I conclude that with kernel version 5.4.0-109 the system worked well, but with the current version installed which is the 5.13.0-48. How can I downgrade this?

Correct.

 

[mike_linux]

Yep - Agree with @Bartman on this those older Nvidia cards are not supported with kernels later then 5.4
to install the 5.4 series

sudo apt install linux-generic

Then reboot and select a 5.4 kernel in advanced options of grub menu. Once booted to the 5.4 kernel then you can remove the 5.13 kernel

THX @Lord Boltar I will try this and reply if this has fixed my problem. If this will work how can I remove the 5.13 kernel then and install the compatible GPU driver version (i.e. 390, 470, etc)? Is it better to let this do the system by itself by issuing

$ sudo ubuntu-drivers autoinstall
$ sudo reboot

 

[Deleted member 108694]

Also forgot
You have to remove the HWE meta packages (hardware enablement)if installed as well with this command, changing «{image, headers}» by the numbering of what we want to eliminate (dpkg --list | grep linux-image in the terminal to see them all). This must be done with everyone except Linux 5.4: this will keep the kernel at 5.4

sudo apt remove linux-{image,headers}-generic-hwe-20.04

 

[Deleted member 108694]

Is it better to let this do the system by itself by issuing

Not really if you have HWE installed it will install 5.13 again use the

sudo apt install linux-generic

instead

the 390 drivers is probably your best bet

 

[mike_linux]

@Lord Boltar I have done all what you have mentioned but unfortunately the 390 version doesn't work. I don't know which version has worked before. Issuing uname -r gives me only the 5.4 kernel version which is installed.

5.4.0-120-generic

I don't know how I can find again the proper driver version. Maybe you or others have addional solutions? BTW what does this OP on my logo mean?

 

[Deleted member 108694]

OP=Original Poster

@Lord Boltar I have done all what you have mentioned but unfortunately the 390 version doesn't work. I don't know which version has worked before. Issuing uname -r gives me only the 5.4 kernel version which is installed.

I don't know how I can find again the proper driver version. Maybe you or others have addional solutions? BTW what does this OP on my logo mean?

I was looking at the Nvidia web site and it is showing the latest drivers will work on your GTX860m - so looking at your screenshot on your post # 1 select the NVIDIA driver metapackage nividia-driver-515 (proprietary) once installed reboot - and also if that is the case your 5.13 kernel should work as well

 

[mike_linux]

OP=Original Poster

I was looking at the Nvidia web site and it is showing the latest drivers will work on your GTX860m - so looking at your screenshot on your post # 1 select the NVIDIA driver metapackage nividia-driver-515 (proprietary) once installed reboot - and also if that is the case your 5.13 kernel should work as well

@Lord Boltar I also have tried this version as I can and I had no success but I will try it again. I don't know why Nvidia such a big company is not capable to fix this problem which is their own fault.

 

[Deleted member 108694]

@Lord Boltar I also have tried this version as I can and I had no success but I will try it again. I don't know why Nvidia such a big company is not capable to fix this problem which is their own fault.

Because they do not care about 5 percent of the market share - when Windows is at 85 percent, Mac 10, Linux 5 give or take a bit depending on who you look at - I had an issue with my Nvidia 940MX and found that the 460/470 drivers were the best some of the other versions were a bit glitchy - Right now I am not even using the Nvidia drivers just using Nouveau drivers - my inxi -Fxz output
Graphics:
Device-1: Intel HD Graphics 620
vendor: Acer Incorporated ALI Aspire E5-575G driver: i915 v: kernel
bus ID: 00:02.0
Device-2: NVIDIA GM108M [GeForce 940MX] vendor: Acer Incorporated ALI
driver: nouveau v: kernel bus ID: 01:00.0

 

[mike_linux]

Because they do not care about 5 percent of the market share - when Windows is at 85 percent, Mac 10, Linux 5 give or take a bit depending on who you look at - I had an issue with my Nvidia 940MX and found that the 460/470 drivers were the best some of the other versions were a bit glitchy - Right now I am not even using the Nvidia drivers just using Nouveau drivers - my inxi -Fxz output
Graphics:
Device-1: Intel HD Graphics 620
vendor: Acer Incorporated ALI Aspire E5-575G driver: i915 v: kernel
bus ID: 00:02.0
Device-2: NVIDIA GM108M [GeForce 940MX] vendor: Acer Incorporated ALI
driver: nouveau v: kernel bus ID: 01:00.0

Sure my laptop is not the newest but before updating the system everything worked fine. I have tried installing kernel 5.13 and the gpu driver nividia-driver-515 (proprietary) but the problem still remains. The system does not boot and get stuck at boot logo. After issuing in tty3 console nvidia-smi I got saw that the gpu was installed. After removing the nividia-driver-515 (proprietary) driver but let the kernel 5.13, the system boots but without any nvidia driver. This shows me that the problem is only the nvidia drivers. I do not know what else to try .

 

[KGIII]

Nvidia doesn't support old graphics cards on any operating system. Nobody supports old hardware on any system. Hardware, like software, comes with a lifespan.

 

[mike_linux]

Nvidia doesn't support old graphics cards on any operating system. Nobody supports old hardware on any system. Hardware, like software, comes with a lifespan.

Ok I totaly agree with you, but why did my system work before the upgrade? Isn't this wired? Are there any open source drivers not especially from nvidia which I can install?

 

[KGIII]

Ok I totaly agree with you, but why did my system work before the upgrade?

If I understand correctly, you were still using a kernel where the OEM supported it. (Or there's some other problem, as yet unidentified.)

The opensource drivers may work just fine. (I should have been more clear and said no OEMs really support old hardware, they pretty much all have cutoff dates as tech sunsets or whatnot.) But, the opensource drivers may work well enough for you and the authors of that may support your graphics card.

"nouveau" is the opensource drivers for Nvidia graphics cards.

 

[mike_linux]

If I understand correctly, you were still using a kernel where the OEM supported it. (Or there's some other problem, as yet unidentified.)

The opensource drivers may work just fine. (I should have been more clear and said no OEMs really support old hardware, they pretty much all have cutoff dates as tech sunsets or whatnot.) But, the opensource drivers may work well enough for you and the authors of that may support your graphics card.

"nouveau" is the opensource drivers for Nvidia graphics cards.

Yes, that is right I used an OEM driver.

Where can I find the open source driver you mentioned?

 

[KGIII]

Where can I find the open source driver you mentioned?

It's in your default repositories, I'm sure. If you uninstall the proprietary drivers, it will quite likely default to the opensource drivers by default. You used the opensource drivers before you used the OEM drivers, even if just for the duration of your installation process.

 

[mike_linux]

It's in your default repositories, I'm sure. If you uninstall the proprietary drivers, it will quite likely default to the opensource drivers by default. You used the opensource drivers before you used the OEM drivers, even if just for the duration of your installation process.

Is this here the driver you mean?

I am not sure if this (X.Org X Server - Nouveau ...) is an open source nvidia driver.

 

[KGIII]

Is this here the driver you mean?

That is indeed the opensource driver, or looks like it is.