[Plugin] Nvidia-Driver

[ich777]

8 hours ago, lincolnliu said:

I am not sure what you mean by using the P4 as console GPU because it doesn’t have any IO ports. Which is why I put in a separate GPU for BIOS

Ah, my bad, I thought that the P4 has outputs too.

 

8 hours ago, lincolnliu said:

I am going to try disable c state, will also try passing it through to my windows 11 VM to see if I can get it working there to rule out a hardware issue.  Will report back.

Hope it isn't a hardware fault, but AMD platforms and Nvidia cards can be a bit "difficult" at least.

[TDA]

hello,

I noticed that upgrading to Unraid 6.12.0-rc6, the power consumption of my tesla p4 (which was in idle 7W) is now 24W.

I don't know if it's a bug of the plugin, or of unraid itself.

[ConnerVT]

Do you have "nvidia-persistenced" in a script on your system?  And did it execute since upgrading to rc6?

[TDA]

1 minute ago, ConnerVT said:

Do you have "nvidia-persistenced" in a script on your system?  And did it execute since upgrading to rc6?

Never configured any script as I'm aware of - since I'm using the GPU only for PLEX.

[ich777]

6 hours ago, TDA said:

the power consumption of my tesla p4 (which was in idle 7W) is now 24W

Is this before anything transcoded on Plex or after something transcoded there?

 

5 hours ago, TDA said:

Never configured any script as I'm aware of - since I'm using the GPU only for PLEX.

The described issue above was most certainly the case also on anything below the beta series because if you don't have nvidia-persistenced run on boot (in the go file or as @ConnerVT mentioned in a user script) with this line:

nvidia-persistenced

 

The cause of the issue is the card doesn't know that it is already in a state where the OS is fully booted since there is no desktop environment on Unraid and if nothing ever transcoded after the boot process the card will always draw more power. However after something used the card or nvidia-persistenced is used the card will drop tho a lower power state than P0 <- highest power state (something like P8 <- lowest power state) and it will draw less power.

[TDA]

On 5/30/2023 at 7:42 AM, ich777 said:

Is this before anything transcoded on Plex or after something transcoded there?

 

The described issue above was most certainly the case also on anything below the beta series because if you don't have nvidia-persistenced run on boot (in the go file or as @ConnerVT mentioned in a user script) with this line:

nvidia-persistenced

 

The cause of the issue is the card doesn't know that it is already in a state where the OS is fully booted since there is no desktop environment on Unraid and if nothing ever transcoded after the boot process the card will always draw more power. However after something used the card or nvidia-persistenced is used the card will drop tho a lower power state than P0 <- highest power state (something like P8 <- lowest power state) and it will draw less power.

Hello,

 

Previously I didn't had the issue though.

So at the moment, I'm using the card only with dockers, so I could add it to the go file.

But I read that if you are using it also for VMs, that's not good and you should avoid it.

So in a multi scenario (VM+Docker) how should it be configured?

 

 

[ich777]

20 minutes ago, TDA said:

So at the moment, I'm using the card only with dockers, so I could add it to the go file.

You've wrote above that you are using the card only with Docker container so I assumed this will be the case anyways.

 

21 minutes ago, TDA said:

So in a multi scenario (VM+Docker) how should it be configured?

Not with nvidia-persistenced since this will most likely causes a hard crash from your server.

 

If you are using nvidia-persistenced you have to make sure that you've stop nvidia-persistenced before you start the VM.

 

The easiest way would be that you start a transcode (of course where the dGPU is used) after you've rebooted and wait a bit so that it goes to a higher power state.

 

21 minutes ago, TDA said:

Previously I didn't had the issue though.

I'm pretty sure it was the same because if you read back a couple of pages (20 or 30) you will find multiple reports about the same behaviour and that is caused because of the above explained reasons.

 

You can safely downgrade and try it again, it will be the same as on the current build. There is nothing changin inside the driver and the driver makes the card even work.

[doctortt]

May I request the author make driver version 535.43.02 available (released on May 30, 2023)? Thanks

[ich777]

32 minutes ago, doctortt said:

May I request the author make driver version 535.43.02 available (released on May 30, 2023)? Thanks

It is available for Unraid 6.11.5 but I assume you are on Unraid 6.12.0-rc6 correct?

 

I've stopped auto building new drivers for RC releases only those which are avilable at the release from the  Unraid RC version will be compiled and there fore listed in the plugin.

 

Please wait for a new RC release and as said above the newest version will be compiled. As said above for Unraid stable all driver versions are usually compiled when they are released by Nvidia.

 

May I ask why do you need the latest driver version and for what do you use the card?

 

EDIT: I almost forgot to say that the driver which you are mentioned is from Nvidias Beta branch.

[doctortt]

29 minutes ago, ich777 said:

It is available for Unraid 6.11.5 but I assume you are on Unraid 6.12.0-rc6 correct?

 

I've stopped auto building new drivers for RC releases only those which are avilable at the release from the  Unraid RC version will be compiled and there fore listed in the plugin.

 

Please wait for a new RC release and as said above the newest version will be compiled. As said above for Unraid stable all driver versions are usually compiled when they are released by Nvidia.

 

May I ask why do you need the latest driver version and for what do you use the card?

 

EDIT: I almost forgot to say that the driver which you are mentioned is from Nvidias Beta branch.

Yes, I'm on 6.12 rc6. I'm a new Unraid user from Truenas Scale and I need ZFS so that's why I went with 6.12 rc6. I guess I will have to wait for Lime to make 6.12 stable in order to have the latest Nvidia driver.

 

The reason I'm asking for the latest driver because I notice the 535.43.02 has a long list of issues fixed and I have a 4060ti founder edition (came out not long ago) installed in my Unraid box.

 

I have been researching on the fan behavior of my 4060ti FE and I just wanted to have the latest driver package to cross out any driver related issue but understanding that fan behaviors are driven by the default vbios. 

 

To provide more context, I learned from the official Nvidia forum that Founder Editions in general don't have advanced cooling like the AIBs, as a result, the fans of FE cards including my 4060ti run at 30% constant by default at all time and this cannot be changed. The issue Im experiencing is that the fans dont ramp up to higher % and it stays at 30%. 

 

I did another test by passing through the gpu to a vm and was able to make the fans to ramp up to higher than 30% using a custom fan curve with msi afterburner , but as you know, once the vm / afterburner is shut, the custom fan curve setting goes away.

Edited June 4, 2023 by doctortt

[ich777]

4 minutes ago, doctortt said:

I have been researching on the fan behavior of my 4060ti FE and I just wanted to have the latest driver package to cross out any driver related issue but understanding that fan behaviors are driven by the default vbios. 

You can always use:

nvidia-persistenced

if you are expiriencing any issues related to the fan or power draw since nvidia-persistnenced is in place to make the card think that it is running a dektop environment even when it's not running a desktop environment.

 

Sadly enough you can't controll the fan from the Nvidia cards from the command line at least not on Unraid.

 

8 minutes ago, doctortt said:

535.43.02

Please give my server a bit, I've started the build for RC6 for the latest driver revision, but please keep in mind usually I'll compile the driver for RC versions only one time. Hope this helps, the build should be done right now and you should be able to see it if you visit the plugin page.

[doctortt]

16 hours ago, ich777 said:

You can always use:

nvidia-persistenced

if you are expiriencing any issues related to the fan or power draw since nvidia-persistnenced is in place to make the card think that it is running a dektop environment even when it's not running a desktop environment.

 

Sadly enough you can't controll the fan from the Nvidia cards from the command line at least not on Unraid.

 

Please give my server a bit, I've started the build for RC6 for the latest driver revision, but please keep in mind usually I'll compile the driver for RC versions only one time. Hope this helps, the build should be done right now and you should be able to see it if you visit the plugin page.

 

Thank you so much. I see that 535.43.02 is now available on RC6 and I installed it - much appreciate it for making it available.  Yes, ``` nvidia-persistenced``` is what I needed. It's only when the driver is loaded (either linux version or windows version), then the default vbios fan curve will come into effect. 

[Handyman]

@ich777 I searched for this but couldn't find proper solution, the plugin only sees one of the GPUs.

 

I have Quadro M2000 and GTX 1650, but only M2000 is picked up by the plugin/driver.




1650 & M2000  in IMMOU

Edited June 10, 2023 by Handyman

[ich777]

27 minutes ago, Handyman said:

I have Quadro M2000 and GTX 1650, but only M2000 is picked up by the plugin/driver.

Please post your Diagnostics.

[Handyman]

Hey, I think I figured out the issue in Diagnostics....But I still need help/suggestion wrt why this happens. I mean it's only 75W card, it should work without power cable just fine.
 

Jun 10 15:43:57 Graviton kernel: NVRM: GPU 0000:04:00.0: GPU does not have the necessary power cables connected.
Jun 10 15:43:57 Graviton kernel: NVRM: GPU 0000:04:00.0: RmInitAdapter failed! (0x24:0x1c:1211)
Jun 10 15:43:57 Graviton kernel: NVRM: GPU 0000:04:00.0: rm_init_adapter failed, device minor number 0

 

graviton-diagnostics-20230610-1556.zip

[ich777]

5 minutes ago, Handyman said:

I mean it's only 75W card, it should work without power cable just fine.

No, if it has a plug for external power you have to connect it.

[Handyman]

1 minute ago, ich777 said:

No, if it has a plug for external power you have to connect it.

Oh! Wasn't sure about that. Will have to buy it then, thank you.

[Handyman]

1 hour ago, ich777 said:

No, if it has a plug for external power you have to connect it.

That worked! Thank you.

 

GPU 0 - 1650

GPU 1 - M2000
Extra Qs just in case if you have any idea, plex is using wrong GPU now (GPU 0), even though I am adding correct GUID variable (For M2000). Any idea why this happens if you're aware.

[alturismo]

4 hours ago, Handyman said:

Extra Qs just in case if you have any idea, plex is using wrong GPU now (GPU 0), even though I am adding correct GUID variable (For M2000). Any idea why this happens if you're aware.

thats pretty sure more a plex related question, but may as hint what you could try

 

there is a preferences.xml file

 

where you can specify the render device (search for it or add it)

 

depending on your system its either D128, D129, D130, ...

 

a simple terminal command will show you all available cards

root@AlsServer:~# ls -la /dev/dri/ren*
crwxrwxrwx 1 nobody users 226, 128 Apr 22 08:00 /dev/dri/renderD128
crwxrwxrwx 1 nobody users 226, 129 Apr 22 08:00 /dev/dri/renderD129
crwxrwxrwx 1 nobody users 226, 130 Apr 22 08:00 /dev/dri/renderD130
root@AlsServer:~# 

 

i have 3 (intel iGPU, NV 3080ti, NV 1060), jest test it through ... be aware, a unraid update can switch the card(s) and Plex aint smart enough to ... so may a manual adjustment would be necessary, Plex Forums would be the place to look for this

[ich777]

8 hours ago, Handyman said:

Any idea why this happens if you're aware.

I‘m assuming you are using the linuxserver.io container correct?

 

Did you follow the second post in this thread to add you GPU to the container?

[Handyman]

Thanks both of you, just seeing this now. But since I haven't been able to use my 1650 in windows VM no matter what I tried, I moved around the GPUs, will use 1650 with Plex and moved M2000 to VM, so it doesn't matter even if Plex doesn't use specified GPU.

[QuantumRetrofit]

I think this may be the right place to ask this.  My self and it seems a couple others are experiencing an issue where a transcode process starts then almost immediately stops (as seen from nvidia-smi).  Sometimes I'll get an error popup on plex, other times the video just never loads.  I can sometimes 'trick' it into holding on to a process and working, but not always.  I'm running unraid 6.12.0 and I have tried several things including using the plexinc/plex, linuxserver/plex, and binhex/plexpass images, using nvidia driver versions 535.54.03 and 535.43.02, and switching GPU PCIe slots.  I don't know if this is any thing, but GPU Statistics indicates power state-throttling: p8 yes(gpu_idle).  I also saw no change in either the plex logs or the unraid log.

[ich777]

16 minutes ago, QuantumRetrofit said:

I also saw no change in either the plex logs or the unraid log.

Have you yet tried the official Plex container?

 

I‘ve also gone throught the other topic but without any further information I can‘t help, at least the Diagnostics would be helpful.

 

From the OP from the other thread, is this the web client from Plex, it is a known issue if you set a transcoding profile there that it won‘t work.

 

What is the source material HEVC, AVC, HDR,…?

[QuantumRetrofit]

You know, I think it is just with the web client...  I appreciate the rapid response.  Thanks

Edited June 16, 2023 by QuantumRetrofit

[Omen]

plugin: installing: nvidia-driver.plg
Executing hook script: pre_plugin_checks
plugin: downloading: nvidia-driver.plg ... done

plugin: downloading: nvidia-driver-2023.05.16.txz ... done


+==============================================================================
| Skipping package nvidia-driver-2023.05.16 (already installed)
+==============================================================================


-----ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR------
---Can't get latest Nvidia driver version and found no installed local driver---
plugin: run failed: '/bin/bash' returned 1
Executing hook script: post_plugin_checks

 

Facing issues with nvidia drivers after upgrading to 6.12.0. I can't find "nvidia-driver-2023.05.16.txz" anywhere, nvidia plugin is not installed. Any ideas?

 

Update 1:

Found at: ./var/lib/pkgtools/packages/nvidia-driver-2023.05.16

Deleted it, the error is almost the same:

 

plugin: installing: nvidia-driver.plg
Executing hook script: pre_plugin_checks
plugin: downloading: nvidia-driver.plg ... done

plugin: downloading: nvidia-driver-2023.05.16.txz ... done


+==============================================================================
| Installing new package /boot/config/plugins/nvidia-driver/nvidia-driver-2023.05.16.txz
+==============================================================================

Verifying package nvidia-driver-2023.05.16.txz.
Installing package nvidia-driver-2023.05.16.txz:
PACKAGE DESCRIPTION:
Package nvidia-driver-2023.05.16.txz installed.

-----ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR------
---Can't get latest Nvidia driver version and found no installed local driver---
plugin: run failed: '/bin/bash' returned 1
Executing hook script: post_plugin_checks

 

Update 2:

 

Seems like deletion on the last step helped, but after reboot. Solved.

Edited June 19, 2023 by Omen