[Plugin] Nvidia-Driver


ich777

Recommended Posts

2 minutes ago, ich777 said:

Please also note that switching the branches in the Nvidia Driver plugin does nothing else then change the version in the config file from the driver plugin which is located on the USB Boot device.

I had assumed this. That’s why I was surprised when my system reacted the way it did. The timing may have been (probably was) completely coincidental though.

 

5 minutes ago, ich777 said:

Have you access to a display that is connected directly to the server. I'm really curious what causes this...

I tried that too. It just showed “no signal”. Same for iKVM via ipmi.

 

I’ve got the system log server running so it’s possible that something is logged there. I’ll dig into it later this week and see if anything was captured.

  • Thanks 1
Link to comment
31 minutes ago, nilsf2000 said:

Here are the Diagnostics. Thanks.

From what I see you are running unRAID as a VM on some kind of Hypervisor or am I wrong?

 

Can you try to boot with Legacy (CSM) boot and not in UEFI mode?

I usually don't give support when using virtualized unRAID instances because this can also be an issue related to the Hypervisor or even if a option is not set, set wrong,... on the Hypervisor for the guest.

 

I would recommend to look first here if you want to run unRAID virtualized:

https://forums.unraid.net/forum/46-virtualizing-unraid/

Link to comment
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

 

Edit: updated with latest diagnostics

 

I'm running into the NVIDIA-SMI has failed issue with a used GTX 1060 I just bought off ebay. The thing appears to work. I can get a display out of it for unRaid. I'd like to use it with Plex and tdarr is my goal. Both Docker containers. I looked it up and it should be compatible. I'm currently on driver v470.86 because I wanted to try something different but I'm about to go to the latest again and reboot with nothing (including dummy plug) in the back of this card. Thinking maybe it's something in my BIOS? I'm not sure I haven't used a GPU in 2 decades. I appreciate all the help I can get.

 

03:00.0 VGA compatible controller: NVIDIA Corporation GP106 [GeForce GTX 1060 3GB] (rev a1)
03:00.1 Audio device: NVIDIA Corporation GF116 High Definition Audio Controller (rev a1)

 

M/B: GIGABYTE GA-7PESH2 Version 00000001 - s/n: 0123456789AB

BIOS: GIGABYTE Version R17. Dated: 06/26/2018

CPU: Intel® Xeon® CPU E5-2660 0 @ 2.20GHz

 

 

IMG_20211120_174056.jpg

IMG_20211120_174108.jpg

unraid-diagnostics-20211120-1805.zip

unraid-diagnostics-20211120-1846.zip

Edited by pinion
Link to comment
5 hours ago, pinion said:

03:00.0 VGA compatible controller: NVIDIA Corporation GP106 [GeForce GTX 1060 3GB] (rev a1)
03:00.1 Audio device: NVIDIA Corporation GF116 High Definition Audio Controller (rev a1)

This seems like a counterfeit card to me since the audio controller is a GF116 (Fermi Architecture - GTX 500 series).

Does this card also have VGA on it? Not 100% sure from the pictures but this is also a indicator that this is not a Pascal (GTX 1000 series) based card.

 

I would recommemd that you watch out for a Nvidia T400 you can get them brand new for about Eur. 130,-

In terms of transcoding it is as good, or even better than a GTX 1060 and more efficient.

  • Like 1
Link to comment
On 7/21/2021 at 1:07 AM, ich777 said:

Put these lines in your /boot/config/go file on the bottom and reboot to enable it:

# Enable persistenced mode for Nvidia cards
nvidia-persistenced

 

That's really strange but eventually it's doing it over your iGPU (if you have one).

 

So I finally fixed my RTX 3060 Ti disappearing by just buying a damn new computer :D Went with an Asrock X570 and an AMD 5950X and now it all works.. Maybe it was just some hardware issue. but it's nice to be able to use this now without it disappearing :)

  • Like 1
Link to comment

OK so it happened again and I think I know why. When the setup runs headless, this happens.
 

I just had my HDMI cable in my video card and it was working fine for a few days. Today I'm building a pfsense router and i took out the HDMI cable to plug it in to my pfsense box. That's when unraid starts giving the can't init adapter error and the nvidia card disappears.

 

I might just buy one of those HDMI ghost plugs to see if that works, but I think if there's nothing plugged into the RTX 3060 Ti,  after a little it'll just fail to be detected.

 

Nope... it's happening even with it plugged in.. i give up. :)

image.png.e58ae2988ad03ebb3a70850e16844c75.png

 

OK Maybe it's because I turned off the monitor.. I'm keeping the  monitor on to see if it stays working....  So far so good..

Edited by HellraiserOSU
Link to comment

OK, so this seems to be it.

As long as I have the monitor plugged in and on, it's fine.

After a minute or so when I turn off the monitor, those RmInitAdapter errolr happens.. and when I turn the monitor back on, nothing comes back on the screen.

So maybe it has something to do with how it goes to sleep and can't wake itself up? Maybe I will order that headless ghost plugin and try.

  • Like 1
Link to comment
2 hours ago, HellraiserOSU said:

Maybe I will order that headless ghost plugin and try.

Yes, maybe buy a dummy HDMI plug and please report back if this is working.

Really curious to see if this solves the issue. :)

Maybe you have a card with a mining limiter and something needs to be plugged in to a monitor output port and that causes the issue.

 

Thank you for the updates! :)

Link to comment
7 hours ago, ich777 said:

Yes, maybe buy a dummy HDMI plug and please report back if this is working.

Really curious to see if this solves the issue. :)

Maybe you have a card with a mining limiter and something needs to be plugged in to a monitor output port and that causes the issue.

 

Thank you for the updates! :)

 

It is mining limited and after overnight leaving it plugged in with the monitor on, it's still working. I'll let you know on the ghost plug outcome.

  • Like 1
Link to comment

I'm running a 1050ti and an HD7770 on my server. The 1050ti is working great for HW transcoding in Plex using this plugin, but I'm having some issues passing through the HD7770 to a macos VM. Going to try swapping the PCIE slots the GPUs are plugged into. Is that going to mess up the nvidia driver config? Will I have to re-configure anything?

Link to comment
14 minutes ago, HellraiserOSU said:

I plugged in both an HDMI and DisplayPort headless ghost dongle.. both failed.. back to having the video card disappear ..

What Dummy Device did you buy?

There are different models out there, some even transmit EDID and I think that's such a device that you actually need for your card to get it working or at least to keep it working.

Link to comment
On 11/23/2021 at 2:21 PM, ich777 said:

That depends on your motherboard but I don't think so.

At least you can swap the cards back if something isn't working anymore.

So I finally got a chance to switch the cards around. The 1050ti still works perfectly for plex transcoding, and the HD7770 is now getting passed through to my macOS VM

  • Like 1
Link to comment

I tried one of the headless ghosts that was made for miners. Took about 10 mins and the GPU disappeared..  Removed GPUstats AND I turned on IOMMU and now it seems... stable? Well it's lasted longer then 5 minutes.. it's going on a half hour now. So usually when it fails it fails in the first 5 minutes.  Fingers crossed I don't have to reboot for a while. :)

  • Like 1
Link to comment

Hello Community, hello Ich777,

 

I've switched to using this plugin since running my card within a Windows vm was a huge struggle, because it kept crashing unraid or did not work on reboot etc. This is much more comfortable, efficient and stable. But there is one issue I can't get my head around. Why is nvidia-smi not able to alter the memory clock of my 3080? I've tried all settings in nvidia-smi also tried nvidia-settings with a DISPLAY and also tried the docker image DebianBuster-Nvidia with nvidia-settings and even GreenWithEnvy but had no luck so far.

Is there something essential that I'am missing? and why is MSI Afterburner able to set settings the nvidia driver in Linux is not able to?

Thank you :D

Link to comment
13 minutes ago, CK98 said:

Why is nvidia-smi not able to alter the memory clock of my 3080?

13 minutes ago, CK98 said:

Is there something essential that I'am missing?

Because you need to a X environment to change the graphic clocks that is basically not existent on unRAID, even in GUI mode.

 

What do you want to do exactly? Do you want to lower the clock for mining or do you simply want that the card enters P8 when it's idle?

 

If you want to change it for mining I would recommend that you install HiveOS or something similar that is based on Linux and pass the card through.

If you only want that the card enters P8 simply issue the command "nvidia-persistenced".

 

15 minutes ago, CK98 said:

and why is MSI Afterburner able to set settings the nvidia driver in Linux is not able to?

Basically the same as described above, the Docker container has a X environment but the X environment needs to be located on the host because the Docker container doesn't have "exclusive" access to the card.

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.