[Plugin] Nvidia-Driver

[wills]

@ich777 I'm at work now and have the 3060 Ti back in, so I can't attach a diagnostics rn.  I was using driver v470.63.01.  It's an EVGA RTX 3060 Ti XC.  On the GPU Statistics page, the pulldown where the UUID would be doesn't show anything with the 3060 installed.  There were a few times I ran nvidia-smi and nothing showed, but whenever it did show and I went back to the driver plugin it disappeared.  I'm mostly asking because I haven't seen any great success stories searching around and a few pages back someone with two 3060s couldn't get theirs working either, said they didn't show up, so I don't know if it's a problem with these cards right now or what.  The motherboard is a Z590 Aorus Elite with an i9-11900K CPU if that helps.

[wills]

I just got the card from EVGA yesterday, so it's probably the latest revision of whatever they do to these things to make them less mineable, I have seen online there are V1 and V2 of some of these LHR cards.  It works fine in Windows.

[wills]

20 hours ago, ich777 said:

Can you attach the Diagnostics?

Here's the diagnostics.  I've now also tried disabling the iGPU along with having a dummy plug and display plugged into the 3060, along with updating Unraid to 6.10.0, nothing works.  The diagnostics file was prior to updating to the latest Unraid, if I can get it working and it matters I'll roll-back.

unraid-diagnostics-20210818-0654.zip

[wills]

 

 

[wills]

This user has the same issue with two 3060s, one EVGA, one Gigabyte.  I see he also posted on this thread back on page 43 and was asked to provide a log, but I don't see that he did.
 

 

[ich777]

24 minutes ago, wills said:

This user has the same issue with two 3060s, one EVGA, one Gigabyte.  I see he also posted on this thread back on page 43 and was asked to provide a log, but I don't see that he did.

Yes exactly, never got a response from him.

 

33 minutes ago, wills said:

Here's the diagnostics.

It looks like you are booting with UEFI or am I wrong?

Can you try to boot with CSM/Legacy mode?

 

Also please make sure that Secure Boot is Disabled.

 

What you can also try is to attach a HDMI monitor to the card and try to boot again.

 

 

EDIT: I've investigated a little further and see a lot of posts in the Nvidia support forums lately about RTX3060's not working but no solution yet I think for most people.

[wills]

2 hours ago, ich777 said:

It looks like you are booting with UEFI or am I wrong?

Can you try to boot with CSM/Legacy mode?

 

Also please make sure that Secure Boot is Disabled.

 

What you can also try is to attach a HDMI monitor to the card and try to boot again.

 

 

EDIT: I've investigated a little further and see a lot of posts in the Nvidia support forums lately about RTX3060's not working but no solution yet I think for most people.

It is in UEFI, I can try legacy later today.  I think secure boot is disabled, will check that also.  I've tried an HDMI display and dummy plug, and that hasn't worked.  You mentioned yesterday a 3060 user having it working a few pages back.  I looked and wasn't able to find that, but I may have missed it.  So far in searching I haven't found anybody with a 3060 conclusively working yet, but it's hard to pin down with 3060 Tis coming up in search results also.  

[wills]

Most of the posts on the Nvidia forum I've seen from people that have 3060 working are running a 5.8+ kernel.

[ich777]

42 minutes ago, wills said:

You mentioned yesterday a 3060 user having it working a few pages back.  I looked and wasn't able to find that, but I may have missed it.

Sorry that was a user with a 3060Ti, looked also again yesterday...

 

6.10.0-rc1 is Kernel version 5.13.8

 

I don't think it's a Kernel thing since the drivers are loaded from what I see in your lsmod output also I have reports that 3070's and 3080's are working...

[wills]

@ich777 for legacy boot mode on my Gigabyte motherboard, I think it's CSM Support -> Enabled, and then Storage Boot Option Control -> Legacy.  There is also an Other PCI Devices option that allows UEFI/Legacy, will that one matter?

Also, could changing any of this affect Unraid adversely since it's setup to UEFI now?

[ich777]

5 minutes ago, wills said:

There is also an Other PCI Devices option that allows UEFI/Legacy, will that one matter?

Yes.

 

5 minutes ago, wills said:

Also, could changing any of this affect Unraid adversely since it's setup to UEFI now?

Rename the EFI folder on your USB Boot device (/boot) to "-EFI"

[wills]

@ich777 I got it work - I think.  Changed boot/PCI devices to legacy and it then showed up for the first time under the GPU Statistics plugin, but it kept dropping out for no apparent reason.  I could get it to drop just by doing different things to access the card, such as going to the Dashboard page where the the load monitor is, going to the Nvidia driver plugin page, running nvidia-smi, it would fall out after doing a couple/three of those things reliably.  I then tried different configs of the iGPU being Auto/Enabled/Disabled with it being primary vs PEG1 (3060).  None of that seems to matter.

If I had a monitor connected and on, it wouldn't drop out that I could see, but if the display shut off, upon a reboot, the 3060 would drop out like before.  So then I tried an HDMI dummy plug on the 3060, that didn't work, then tried a DP dummy plug on it and that didn't work, but for some reason having both the HDMI/DP dummies in it hasn't fallen off yet.  I've tried the normal stuff to break it and it hasn't and rebooted/cold-booted and it's staying active.

I don't really understand why it's working like that, but I'm not going to breathe on it.    Thanks for your help.  If you want me to guinea-pig some other stuff let me know.

[ich777]

3 hours ago, wills said:

If you want me to guinea-pig some other stuff let me know.

Can you send me the Diagnostics after it dropped feom the system, that would be really helpful since I can see what error it throws in the syslog.

 

Glad you got it working now, hopefully it's stable now and doesn't drop off.

[wills]

8 hours ago, ich777 said:

Can you send me the Diagnostics after it dropped feom the system, that would be really helpful since I can see what error it throws in the syslog.

The zip i uploaded yesterday was after it dropped out.  Looking in the syslog.txt file, at the bottom there are a bunch of these, repeating every few seconds:

Aug 18 06:52:12 Unraid kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x23:0xffff:1204)
Aug 18 06:52:12 Unraid kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0

I just ran one now after the system has been running all night, and I don't see any errors like that now.

[wills]

Is it possible the GPU was entering a low-power mode and that caused it to drop off?  Is there a command that can be ran to prevent it from doing that?

[ich777]

8 minutes ago, wills said:

Is it possible the GPU was entering a low-power mode and that caused it to drop off?  Is there a command that can be ran to prevent it from doing that?

nvidia-persistenced

[wills]

Watching the GPU Statistics output, it looks as though the card is not entering any alternate power state (staying in P0) now.  I don't know if that is the result of legacy mode or because I have dummy plugs inserted or what, but I'm fine with that.  The plugin never worked for me in UEFI mode with the 3060, so I don't really know what it was doing, and if it dropped out, I wouldn't be able to see the power state anyway.  It only draws around 50w of power even with the GPU not dropping its core clock speed, so nbd.

 

Edited August 19, 2021 by wills

[ich777]

1 hour ago, wills said:

It only draws around 50w of power even with the GPU not dropping its core clock speed, so nbd.

Have you tried to issue the above command?

This should also help with power draw, but I think it may be related to the dummy plugs.

[wills]

2 hours ago, ich777 said:

Have you tried to issue the above command?

This should also help with power draw, but I think it may be related to the dummy plugs.

I enabled it and the idle usage went to this:



Running Emby, transcoding a 4K HDR -> 1080p SDR title it showed this:



This was with Emby's throttling off so I could more accurately access load, but I normally run with throttling on the help the CPU.  I couldn't figure out how to turn off persistence mode with that command, I tried nvidia-persistenced --no-persistence-mode but it threw an error, so I turned it off with nvidia-smi -pm DISABLED.

But having it on didn't seem to affect the GPU, so I turned it back on to save a few watts, thanks.  Do you know a recommended way to enable this either with nvidia-persistenced or nvidia-smi -pm ENABLED upon system boot after the Nvidia driver plugin has been loaded?

EDIT:  I added a script to User Scripts to be loaded at startup of array.  If that's wrong let me know, otherwise I'll roll with that.

Edited August 19, 2021 by wills

[ich777]

1 hour ago, wills said:

I added a script to User Scripts to be loaded at startup of array.  If that's wrong let me know, otherwise I'll roll with that.

You can also add this to the go script that is located in /boot/config/go, at least I do it this way.

[Borbosch]

Today i tried to reinstall the Plugin because i got some errors on Startup:

 

modprobe: FATAL: Module nvidia not found in directory /lib/modules/5.10.28-Unraid

During the reinstall the download of the driver fails. Anyone experiencing similar problems?

 

plugin: installing: https://github.com/ich777/unraid-nvidia-driver/raw/master/nvidia-driver.plg
plugin: downloading https://github.com/ich777/unraid-nvidia-driver/raw/master/nvidia-driver.plg
plugin: downloading: https://github.com/ich777/unraid-nvidia-driver/raw/master/nvidia-driver.plg ... done
plugin: downloading: https://github.com/ich777/unraid-nvidia-driver/raw/master/packages/nvidia-driver-2021.07.30.txz ... done

+==============================================================================
| Skipping package nvidia-driver-2021.07.30 (already installed)
+==============================================================================


+==============================================================================
| WARNING - WARNING - WARNING - WARNING - WARNING - WARNING - WARNING - WARNING
|
| Don't close this window with the red 'X' in the top right corner until the 'DONE' button is displayed!
|
| WARNING - WARNING - WARNING - WARNING - WARNING - WARNING - WARNING - WARNING
+==============================================================================

-----------------Downloading Nvidia Driver Package v470.63.01------------------
----------This could take some time, please don't close this window!------------

--------------Can't download Nvidia Driver Package v470.63.01-----------------
plugin: run failed: /bin/bash retval: 1

Updating Support Links



Finished Installing. If the DONE button did not appear, then you will need to click the red X in the top right corner

 

[ich777]

34 minutes ago, Borbosch said:

During the reinstall the download of the driver fails. Anyone experiencing similar problems?

Yes such a message pops up now and then, the message says that the plugin can't download the driver itself.

Have you installed any ad blocking software in your network like PiHole or AdGuard?

 

Can you download this archive: Click?

If you can't download this archive something is blocking the connection to Github.

 

From here I can download the file just fine.

 

Please try to uninstall the plugin and reinstall it again.

[Borbosch]

8 hours ago, ich777 said:

Yes such a message pops up now and then, the message says that the plugin can't download the driver itself.

Have you installed any ad blocking software in your network like PiHole or AdGuard?

 

Can you download this archive: Click?

If you can't download this archive something is blocking the connection to Github.

 

From here I can download the file just fine.

 

Please try to uninstall the plugin and reinstall it again.

Okay, seems like a temporary problem. Today the installation and download just works. Yesterday i gave up after 6 tries.
Thank you very much for your time.

[stormshaker]

Hardware tone mapping isn’t working for me - Plex drops back to CPU transcoding.

 

GTX-1650, Unraid 6.9.2

pms official docker - on the latest branch

latest production nvidia-driver - 470.63.1

hardware transcoding working fine

 

Do I have to use the linuxserver.io container? Looks like other people have HDR tonemapping working with the official docker.

 

J

 

[ich777]

10 minutes ago, stormshaker said:

Hardware tone mapping isn’t working for me - Plex drops back to CPU transcoding.

I would try to create a forum entry in the Plex forums because when hardware transcoding is working just fine and HDR not then there must be something wrong with the transcoder itself.

 

Hope that makes sense to you.

 

You can try the linuxserver.io Plex container if it fixes the issue for you.