3 hours ago, ich777 said:

Sorry but I can't help with Unifi since I don't have such hardware.

 

Second post in this thread.

All good now, thanks again.

37 minutes ago, papaHoff said:

Well, the most common error i see is that the CPU gets overloaded and all cores go to 100% and stays there.

And you are sure that it's not the audio stream that get's transcoded?

 

As you can clearly see here the transcoding is working:

 

41 minutes ago, papaHoff said:

Playback on phone:

Because most likely your phone can play the file natively...?

 

42 minutes ago, papaHoff said:

Playback on chromecast:

Because your Chromecast can't play the file natively and needs to be transcoded.

 

You can also see here that cuda is used so it is definitely working:

 

What I can't tell you is why you don't get any output from nvidia-smi but in none of your screesnhots above it's visible if you are issuing the command from the container or from the host.

 

I really don't know what to tell you here, please also don't forget as mentioned above that most of the times the audio stream needs to be transcoded too and that is entirely done on the CPU.

 

However from my perspective everything is working and you should ask in the support thread from your Jellyfin container however as said above everything is working as it should from what I can see.

Security flaws found in all Nvidia GeForce GPUs. Update drivers ASAP!

https://www.pcworld.com/article/2504035/security-flaws-found-in-all-nvidia-geforce-gpus-update-drivers-asap.html

 

 

is this will effects unraid with nvidia GPU? 

thank you 

24 minutes ago, brutalmethod said:

Security flaws found in all Nvidia GeForce GPUs.

You're a bit late to the game, already reported here:

 

and answered one post below.

 

Sure, this vulnerability is real bad but I assume you don't have your server exposed to the open Internet since a attacker first would need first remote access to the container that the GPU is passed through or your server.

I added an older Nvidia GT 630 GPU to my Unraid server with the goal of using it for video transcoding in Plex. I installed the "Nvidia Drivers" plugin and tested all available driver versions, up to v470.239.06, which, according to Nvidia’s website, should support this model. However, within the plugin, I receive the following message:

 

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

 

How can I resolve this?

1 hour ago, francispb said:

How can I resolve this?

I can't help without Diagnostics.

I'm sorry, here is the diagnostic file.

 

 

servidor-diagnostics-20241105-1328.zip

2 minutes ago, francispb said:

I'm sorry, here is the diagnostic file.

Nov  5 12:01:32 servidor kernel: NVRM: The NVIDIA GeForce GT 630 GPU installed in this system is
Nov  5 12:01:32 servidor kernel: NVRM:  supported through the NVIDIA 390.xx Legacy drivers. Please
Nov  5 12:01:32 servidor kernel: NVRM:  visit http://www.nvidia.com/object/unix.html for more
Nov  5 12:01:32 servidor kernel: NVRM:  information.  The 470.239.06 NVIDIA driver will ignore
Nov  5 12:01:32 servidor kernel: NVRM:  this GPU.  Continuing probe...
Nov  5 12:01:32 servidor kernel: NVRM: No NVIDIA GPU found.
Nov  5 12:01:32 servidor kernel: nvidia-nvlink: Unregistered the Nvlink Core, major device number 244
Nov  5 12:01:32 servidor kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 244
Nov  5 12:01:32 servidor kernel: NVRM: The NVIDIA GeForce GT 630 GPU installed in this system is
Nov  5 12:01:32 servidor kernel: NVRM:  supported through the NVIDIA 390.xx Legacy drivers. Please
Nov  5 12:01:32 servidor kernel: NVRM:  visit http://www.nvidia.com/object/unix.html for more
Nov  5 12:01:32 servidor kernel: NVRM:  information.  The 470.239.06 NVIDIA driver will ignore
Nov  5 12:01:32 servidor kernel: NVRM:  this GPU.  Continuing probe...
Nov  5 12:01:32 servidor kernel: NVRM: No NVIDIA GPU found.

 

As you can clearly see this card is not supported, even if the docs say that it's supported.

 

This card is a horrible choice for transcoding because it is based on the Fermi architecture and I think no about 12 years old. This card will even struggle with transcoding h264 (AVC) and won't transcode any h265 (HEVC) content and will fall back to the CPU.

 

I would recommend that you look out fro a new card eg: a Nvidia T400, it is fairly recent, based on the Turing architecture, low power (depending on the variant), don't need external power, you should get it for pretty cheap and the 2GB variant is good enough for at least 3x simultaneous 4K HDR transcodes (depending on the media even more).

1 hour ago, francispb said:

I added an older Nvidia GT 630 GPU to my Unraid server with the goal of using it for video transcoding in Plex. I installed the "Nvidia Drivers" plugin and tested all available driver versions, up to v470.239.06, which, according to Nvidia’s website, should support this model. However, within the plugin, I receive the following message:

 

may rather look for another card, this one is not supported anyhow, here you have a nice overview ... nvenc is what you need to be supported, on the right side you see which features are supported ...

 

https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new

Can not install，what should I do，please

 

 

diskstation-diagnostics-20241108-1414.zip

1 minute ago, yuntong said:

Can not install，what should I do，please

If you are located in China please note that you have to use a Proxy to be able to download the driver package since the Nvidia Driver plugin is using GitHub Releases which seems to be not working reliable over there.

 

You could also install everything manually but that is a really tedious process...

1 minute ago, ich777 said:

If you are located in China please note that you have to use a Proxy to be able to download the driver package since the Nvidia Driver plugin is using GitHub Releases which seems to be not working reliable over there.

 

You could also install everything manually but that is a really tedious process...

I have already download by proxy net，and access your github repo successful

1 minute ago, yuntong said:

I have already download by proxy net，and access your github repo successful

So is it now working?

If not try the following:

1. Remove the plugin
2. Reboot your server
3. Make sure the proxy is enabled
4. Install the plugin (this might take some time because the Nvidia drivers are about 200MB in size)

2 minutes ago, ich777 said:

So is it now working?

If not try the following:

1. Remove the plugin
2. Reboot your server
3. Make sure the proxy is enabled
4. Install the plugin (this might take some time because the Nvidia drivers are about 200MB in size)

Thanks,I will try.

I'm located in Dallas, TX, and have the same issue as Yuntong. Some history to help with analysis:
Until a few weeks ago, I'd been blissfully using the driver with Plex for 9-10 months with no issues. I upgraded Unraid to 6.12.13 from 6.12.10 but missed the note on ensuring that this driver was updated to the latest version. After the reboot, the driver disappeared. I tried an install from CA but kept getting this message. I went through the recommended steps (remove the plugin, reboot server, install the plugin), but I got the same result. I'm using a GTX 3060 card and verified that Unraid sees the card (GPU Stats). My BIOS is updated to the latest version. Diagnostics attached. 
Any help is GREATLY appreciated.
 

chilltower-diagnostics-20241110-1539.zip

7 hours ago, dwgtx said:

I'm located in Dallas, TX, and have the same issue as Yuntong. Some history to help with analysis:
Until a few weeks ago, I'd been blissfully using the driver with Plex for 9-10 months with no issues. I upgraded Unraid to 6.12.13 from 6.12.10 but missed the note on ensuring that this driver was updated to the latest version. After the reboot, the driver disappeared. I tried an install from CA but kept getting this message. I went through the recommended steps (remove the plugin, reboot server, install the plugin), but I got the same result. I'm using a GTX 3060 card and verified that Unraid sees the card (GPU Stats). My BIOS is updated to the latest version. Diagnostics attached. 
Any help is GREATLY appreciated.
 

chilltower-diagnostics-20241110-1539.zip 238.18 kB · 2 downloads

you may need to uninstall the NVIDIA plugin and reinstall it from the app store.

then in the NVIDIA plugin

setting> nvidia> click on production and hit the downlaod and update button restart.

8 hours ago, dwgtx said:

I upgraded Unraid to 6.12.13 from 6.12.10 but missed the note on ensuring that this driver was updated to the latest version.

Did you notice the message that you should wait for third party drivers to update and to tell you that it is safe to reboot?

 

8 hours ago, dwgtx said:

After the reboot, the driver disappeared.

My comment above is maybe why it disappeared.

 

8 hours ago, dwgtx said:

I tried an install from CA but kept getting this message.

I just tried it and it is working over here:

 

May I ask what Network gear that you have and if you do some AdBlocking on your Network?

 

The Plugin is using the GitHub API and some people reported here that Unifi network gear at times blocks access to the API or to GitHub, I'm not 100% sure since I don't own any Unifi gear, I assume it has something to do with IDS.

 

However it seems that you are running into a Checksum error which is very weird, are you sure that your USB Boot Device is okay? But I still assume that this is some kind of network error.

I appreciate the replies from the community. Here are answers to questions thus far:

• while the message during install indicates that the "Package nvidia-driver-2024.07.10.txz installed", I see no evidence that it actually did (I'm assuming the initial install was deleted when I tried to reinstall after the 6.12.13 update). The plugin does not appear under Installed Apps/Plugins in CA; there is no nvidia-driver folder on my flash config folder nor in the plugins-error or plugins-removed folders and nothing in the /boot/config/plugins/ folder. So it's unclear how I would go about uninstalling a plugin I can't find
• Regarding reboot after upgrading to 6.12.13, unfortunately, I did not notice the message to wait for third-party drivers to update before reboot. My bad, and now I'm paying for it
• I'm using Sodola 2.5G network switches with an Eero Pro 6 gateway that handles DHCP and connects to my cable modem. My Unraid server has a fixed IP. There is no AdBlocking at the network level (browser only) and no AB plugins on the Unraid server
• I'm in the process of replacing my USB Boot device to see if that is the culprit
• Note: the current network equipment and Flash Boot device were utilized during the successful initial driver install. However, upgrading the USB device was already on a maintenance checklist. I'll provide an update when that's complete
   

Well, well, well... This was a sight for sore eyes. Or, as we say in Texas: "I'm happier than a jackrabbit at a speed dating social."
It was, in fact, the USB drive. To assist others who may encounter this issue, here are the before and after USB specs:

• Was Using: Sandisk 32G Ultra 3.1 (in use for 2+ years and apparently wore out)
• Replaced with: SAMSUNG BAR Plus 32G 3.1 (Spaceinvader One recommendation)

As Joan Baez aptly put it: "You don't know what you have till it's gone." I had no idea just how vital this driver was to my daily existence until I rushed through a server update without reading the (crucial-and-will-never-skip-again) installation notes. I'm grateful to the community for your time and expertise and incredibly thankful for ich777's work and commitment to his community engagement. I'm headed to his donation link and encourage everyone to do the same.
https://paypal.me/chips777

1 hour ago, dwgtx said:

Replaced with: SAMSUNG BAR Plus 32G 3.1 (Spaceinvader One recommendation)

Maybe a recommendation from my side: Transcend 32GB USB 2.0

This is a really good device and served my server for years without any issue, as a bonus you should be still be able to get a USB 2.0 version which runs way cooler than a 3.0+ USB device, sure the boot times take a bit longer (about 10 to 20 seconds) but that is negligible.

 

Just a recommendation, if you are satisfied with your new USB Boot device no need to change.

 

Anyway glad that everything is solved now and you are up and running again.

I am trying to do some troubleshooting with a newly installed nVidia GPU(RTX4070 Super OC) since installing its been hanging on the USB boot process at random points. not sure at what point these drivers are loaded. but i located the directory that the drivers are in and wanted to change the version on the USB boot device, to see if this is an issue but since im unable to get to the gui atm there is no easy way, is this possible and if so, where would i get the previous version driver package built for this?

14 minutes ago, J_Dub said:

I am trying to do some troubleshooting with a newly installed nVidia GPU(RTX4070 Super OC) since installing its been hanging on the USB boot process at random points. not sure at what point these drivers are loaded. but i located the directory that the drivers are in and wanted to change the version on the USB boot device, to see if this is an issue but since im unable to get to the gui atm there is no easy way, is this possible and if so, where would i get the previous version driver package built for this?

Please don't try to do this manually since it is really complicated and you maybe have to change multiple files in that process and find the right driver, I would recommend that you delete the nvidia-driver.plg and nvidia-driver folder from the directory .../config/plugins/

 

After that the server should boot without the driver you then can install the Driver again from the CA App and it should download the driver just fine, you then can choose another driver on the plugin page.

 

Can you please explain what was your issue in the first place and maybe where your server did hang in detail so that I can maybe tell more about the issue?

From time to time (maybe every 2-3 days) I will get the following errors in my syslog. They then happen every few seconds and fill up the syslog so that the gui becomes unresponsive and eventually I have to do a dirty restart.  

 

I reached out to Limetech but was told to consult this mega post.  Can anyone advise what I should do?  I have a Nvidia 3080 RTX.  Only really use it for re-encoding with ffmpeg via a Tdarr container. Stopping docker doesn't stop the errors happening once they start happening and it isn't related to activity that I can see as Tdarr only runs most days for 30 mins at most as it only reencodes new stuff in my library.

 

2024-11-14 13:31:48    Kernel.Warning    192.168.100.78    Nov 14 13:31:48 DavoUnraid kernel: NVRM: GPU 0000:10:00.0: RmInitAdapter failed! (0x62:0x65:2468)
2024-11-14 13:31:48    Kernel.Warning    192.168.100.78    Nov 14 13:31:48 DavoUnraid kernel: NVRM: GPU 0000:10:00.0: rm_init_adapter failed, device minor number 0

8 hours ago, Davo67 said:

Kernel.Warning

Please post your Diagnostics.

 

Do you maybe also have Diagnostics when this happens?

Hoping for some help with this.

 

Subject: nv_open_q causing CPU spikes
 

Unraid with the Nvidia driver was running fine for me on the 8GB (2x4GB) of RAM I had in there but I figured I'd replace with 16GB (2x8GB). Replaced the RAM and then rebooted. I have always ran Unraid headless but this time I happened to have a monitor hooked up so I could check to make sure that no XMP or any overclocking was set for the new RAM. After confirming that and running memtest (passed for 5+ hours) I booted and noticed that it said something about installing the nvidia open source driver? I just pressed enter and moved on and didn't think about it. 

I had already had an nvidia card installed and plugin installed with the "latest branch" nvidia driver running fine with no issues so was surprised to see the reference to the open source driver install popping up. 

Since then I keep getting CPU spiking. Everythign works as expected but CPU is always fluctuating and spiking. I run only a single Plex docker on this (binhex-plex) server, along with gpu-stats and Fix Common Problems and a couple other popular plugins. No VMs. Before, since I only run a plex server on this Unraid server and nothing else, I would never see more than 1 or 2% activity unless i was actively watching a plex stream. Now, it is constantly fluctuating CPU and spiking. When I run the top cmd I see that two processes are causing it shfs and nv_open_q, but priimarily one is causing the spiking: nv_open_q 

 

Here is what I have tried to resolve this:

1. Uninstalled and reinstalled the nvidia driver (countless times, rebooting after each install).

2. Removed the 2x4GB of RAM from slots A2 and A4 where they always were before when it was running fine, and moved the new 2x8GB into those slots.

3. Yesterday I opened top and did a kill cmd on the nv_open_q. Checked back a minute or two later and the CPU was no longer spiking. So definitely that process is the cause.

 

Attaching diagnostics and screenshots. Any help is appreciated. Only other thing I can think to try is to remove the new RAM and go back to the old RAM but I can't see how that would fix this. I think I need to remove the nv_open_q  somehow but not sure how. Uninstalling the nvidia driver hasn't accomplished that.

unraid-diagnostics-20241117-1232.zip