[Plugin] Nvidia-Driver

[Howboys]

On 7/15/2022 at 10:03 PM, ich777 said:

These are two times the same things that you‘ve wrote…

 

This is the first time that I hear from such an issue that is related to TPM in general.

I know that CSM disabled and booting with UEFI can cause issues with the Nvidia driver but never heard of TPM causing such issues.

 

EDIT: From what I saw from your syslog it seems like it never did load the module, but I think the Diagnostics from a boot with UEFI instead of a Legacy boot (CSM)?

 

Yup. The first time I tried it, I may not have deleted the driver entirely. The next time I tried it with a fresh install, it ended up working.

 

However, then I took my new build and tried to boot up with my main server's boot USB and it wouldn't do it. So I had to enable CSM Support (Secure Boot, TPM, fTPM disabled though, didn't try enabling them).

[big_boss]

After updating to 6.10.3 my t600 no longer can be used to transcode or in vms. It is detected in the nvidia plugin. I have VT-D Enabled but IOMMU is showing as disabled in Unraid. I am pretty confused. The nvidia-smi command shows there are no processes, which is true as plex transcoding us being done on the cpu, but as this image shows, the gpu statistics says it is doing something. It all worked fine untill I added my raid card and updated to 6.10.3. I think it could possibly be the motherboard as 2 of the ram slots stopped working this morning as well. Any suggestions?

terabyte-diagnostics-20220719-1409.zip

[ich777]

26 minutes ago, big_boss said:

The nvidia-smi command shows there are no processes

Don't use this command use it in combination with 'watch' like:

watch nvidia-smi

(to exit press CTRL + C)

 

27 minutes ago, big_boss said:

which is true as plex transcoding us being done on the cpu, but as this image shows, the gpu statistics says it is doing something.

Doesn't have to be true, you have to undestand that Plex transcodes in chunks and ahead of time and also does throttle the transcode and maybe you've issued the command where it was actually throttling, that's why I recommend using the command from above.

Also transcoding is not done entirely on the CPU, don't forget that if you transcode it is most likely that the Audio has to be transcoded to and if you have embedded subtitles in the file Plex can use almost 100% of the CPU, in your case, if the transcode is done entirely on the CPU it would use 100% of your CPU since this is a fairly old CPU nowadays.

 

Please double check that it isn't working, my Nvidia T400 is working just fine on 6.10.3 with the latest Nvidia driver and from what I see, yours should also work fine too.

[big_boss]

2 minutes ago, ich777 said:

Don't use this command use it in combination with 'watch' like:

watch nvidia-smi

(to exit press CTRL + C)

 

Doesn't have to be true, you have to undestand that Plex transcodes in chunks and ahead of time and also does throttle the transcode and maybe you've issued the command where it was actually throttling, that's why I recommend using the command from above.

Also transcoding is not done entirely on the CPU, don't forget that if you transcode it is most likely that the Audio has to be transcoded to and if you have embedded subtitles in the file Plex can use almost 100% of the CPU, in your case, if the transcode is done entirely on the CPU it would use 100% of your CPU since this is a fairly old CPU nowadays.

 

Please double check that it isn't working, my Nvidia T400 is working just fine on 6.10.3 with the latest Nvidia driver and from what I see, yours should also work fine too.

Thanks for speedy reply and for all the awesome work you do for this community. 

It is reporting this back. I think it may be working but not correctly, does this mean it is an issue with the plex implementation of the gpu?

[big_boss]

Interestingly, when I disable HW accelerated transcoding on plex, the gpu still shows the same as the image above, suggesting it is being incorrectly utilised

 

EDIT - When I disabled both checkboxes for hardware acceleration No running proccess was found with watch nvidia-smi

Edited July 19, 2022 by big_boss
more info

[ich777]

44 minutes ago, big_boss said:

EDIT - When I disabled both checkboxes for hardware acceleration No running proccess was found with watch nvidia-smi

So this means everything is working in terms of the Nvidia Driver or am I wrong?

[big_boss]

Just now, ich777 said:

So this means everything is working in terms of the Nvidia Driver or am I wrong?

Yes I beleive so.

 

I think I have found the root of the issue. The 4670k only has 16 pcie lanes, when adding the raid card to the second pcie slot this will have halved the pcie bandwith of the second slot, resulting in the gpu only getting 8 lanes instead of the 16 and I think this would account for the drop in perfomance.

 

Thanks again for your commitment to the community and for your time spent on noobs like myself.

[ich777]

3 minutes ago, big_boss said:

I think I have found the root of the issue. The 4670k only has 16 pcie lanes, when adding the raid card to the second pcie slot this will have halved the pcie bandwith of the second slot, resulting in the gpu only getting 8 lanes instead of the 16 and I think this would account for the drop in perfomance.

Definitely not.

8 lanes are plenty of lanes to do the transcode, even one lane of PCIe 3.0 is enough to do it.

[big_boss]

1 minute ago, ich777 said:

Definitely not.

8 lanes are plenty of lanes to do the transcode, even one lane of PCIe 3.0 is enough to do it.

I am completely out of ideas then

[big_boss]

5 hours ago, ich777 said:

So this means everything is working in terms of the Nvidia Driver or am I wrong?

After some more investigation, I can see that the nvidia gpu is only being used to decode and not to encode

[ich777]

Just now, big_boss said:

After some more investigation, I can see that the nvidia gpu is only being used to decode and not to encode

Maybe you are transcoding to a format that it can't handle or something like that, but that seems more of an issue with the container/settings than the plugin.

[tqb333]

 Please, could you assist. Is there any possible way to install legacy driver for NVS300?

 

 

I was trying to 340.108 (the latest for my card) by 

sh NVIDIA-Linux-x86_64-310.40.run

But I got "binutils" error, so I installed a Dev Pack plugin and get all packages from there but anyway I got error:

You do not appear to have libc header files installed on your system. Please install your distribution's libc development package

 

I dont know how to get libc and would it help me? 

 

 

Edited July 20, 2022 by tqb333

[ich777]

6 hours ago, tqb333 said:

 Please, could you assist. Is there any possible way to install legacy driver for NVS300?

What do you want to do with this driver?

 

This drivers are not suitable for usage in Docker contianers, only driver versions 418.81.07 and up are supported, see here.

[tqb333]

12 hours ago, ich777 said:

What do you want to do with this driver?

 

This drivers are not suitable for usage in Docker contianers, only driver versions 418.81.07 and up are supported, see here.

Using in Jellyfin.

So, no way to use it? I dont understand why. Another windows and linux systems can use it. Why docker cant?

 

p.s. oh, sry, ok I am starting to understand after I read it https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#platform-requirements
Before I didnt see any info about NVIDIA doest support containers, I thought it a docker or plugin problem

Edited July 21, 2022 by tqb333

[tqb333]

Ok, I have another Card:  GeForce GTX 670 -- its is not easy to install (i need rizer, power etc) but i will.
I hope in future it wont be unsupported, because this card at the last string in the driver...

[ich777]

54 minutes ago, tqb333 said:

p.s. oh, sry, ok I am starting to understand after I read it https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#platform-requirements

Exactly.

 

Also this card doesn't support h265 or better speaking HEVC, I think you only can use it for decode and encode of certain h264 files if I'm not mistaken.

 

If you are willing to buy something new and recent, look into a Nvidia T400, this card is Turing based supports h265, has a maximum power consumption from 35W and doesn't need external power. You can get those cards for about $120,- brand new.

[tqb333]

1 hour ago, ich777 said:

Exactly.

 

Also this card doesn't support h265 or better speaking HEVC, I think you only can use it for decode and encode of certain h264 files if I'm not mistaken.

 

If you are willing to buy something new and recent, look into a Nvidia T400, this card is Turing based supports h265, has a maximum power consumption from 35W and doesn't need external power. You can get those cards for about $120,- brand new.

Thanks for Advice, but I already have my old gtx 670 and better to use it if I have it already.

Any recommendation how to low the power of this card?

[tqb333]

21 minutes ago, tqb333 said:

Thanks for Advice, but I already have my old gtx 670 and better to use it if I have it already.

Any recommendation how to low the power of this card?

Ok, I googled and was surprised my gtx670 doesnt support HEVC too... How can I check do I need card for decoding at all? May be sofware is enought for my Jellyfin? 

 

And i cant understand how it works. If the GPU doesnt have encoder for HEVC, the Jellyfin have to use software decoding, so CPU, right? Why cant it use GPU clock to do it?

 

At all, Can I use H.264 (AVC) and switch off H.265 (HEVS)?

I dont need 8k, only 4k sometimes. Usselly fulllHD, and have big-enough band to watch it without best compression

 

 

p.s. BTW as i see only EDGE support h.265 codec-support

Edited July 21, 2022 by tqb333

[ich777]

1 hour ago, tqb333 said:

Thanks for Advice, but I already have my old gtx 670 and better to use it if I have it already.

As you already figured out this card is not working for HVEC...

 

1 hour ago, tqb333 said:

Any recommendation how to low the power of this card?

This card draws even in Idle more than most newer cards when they are not in idle...

 

1 hour ago, tqb333 said:

How can I check do I need card for decoding at all? May be sofware is enought for my Jellyfin? 

Here: Click

 

1 hour ago, tqb333 said:

And i cant understand how it works. If the GPU doesnt have encoder for HEVC, the Jellyfin have to use software decoding, so CPU, right?

If you GPU isn't capable of transcoding a HEVC file it should fall back to CPU transcoding but note that his will hit your CPU real hard and older CPUs or low power CPUs won't be able to keep up and you got stuttering video playback while transcoding.

 

1 hour ago, tqb333 said:

Why cant it use GPU clock to do it?

I don't fully understand this question but you have to understand that in the Nvidia GPUs there is a small portion of the chip which is the NVENC/NVDEC (Nvidia Encoder/Nvidia Decoder) and if that doesn't support a specific codec then you are out of luck.

 

1 hour ago, tqb333 said:

At all, Can I use H.264 (AVC) and switch off H.265 (HEVS)?

I dont need 8k, only 4k sometimes. Usselly fulllHD, and have big-enough band to watch it without best compression

If you have a h265 file and want to transcode it to h264 you can usually use your GPU for Encoding the file but not for Decoding, that is then done on the CPU and so on, and that can hit your CPU pretty hard.

 

1 hour ago, tqb333 said:

p.s. BTW as i see only EDGE support h.265 codec-support

That means native playback and has nothing to do with the issue that you are trying to solve...

 

 

Sorry I can't help any further, these are all questions that belong in the Jellyfin or whatever container you are using support thread and not here. This is not an issue of the driver, these are hardware limitations where I can't help.

[KnobleOutlaw]

last night I was streaming from Plex, suddenly the stream freezes. went to check my unraid dashboard, timeout error... reboot the server, unraid boots up fine, everything else is working correctly EXCEPT my plex docker container is stopped, errors on start, the error was a bad param on --runtime=nvidia, ok so I remove that param and all the other gpu flags, plex container works fine now on cpu enc.
I notice that the driver plugin is missing, and go to reinstall it, I get this error: 
 

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

+==============================================================================
| Skipping package nvidia-driver-2022.05.06 (already installed)
+==============================================================================


-----ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR------
----Can't get Production Branch version and found no installed local driver-----
-----Please wait for an hour and try it again, if it then also fails please-----
------go to the Support Thread on the unRAID forums and make a post there!------
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

I'm assuming I need to delete the driver installation so that the plugin can be reinstalled? 
the card still does appear in hardware devices
 

[1022:1604]00:18.4 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 15h Processor Function 4

[1022:1605]00:18.5 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 15h Processor Function 5

[10de:1c82]01:00.0 VGA compatible controller: NVIDIA Corporation GP107 [GeForce GTX 1050 Ti] (rev a1)

[10de:0fb9]01:00.1 Audio device: NVIDIA Corporation GP107GL High Definition Audio Controller (rev a1)

[1b21:0612]02:00.0 SATA controller: ASMedia Technology Inc. ASM1062 Serial ATA Controller (rev 02)

[ich777]

1 hour ago, KnobleOutlaw said:

I'm assuming I need to delete the driver installation so that the plugin can be reinstalled? 

Please post your Diagnostics.

A crash shouldn't wipe the Nvidia Driver plugin at least I don't know how that would be possible.

Can you open this file on your local computer: Click

 

Do you have any AdBlocking software installed on your network?

Please double check on the plugin page that the plugin is first uninstalled from the Plugins Error tab if it is in there, then reboot and then try again to redownload the driver plugin from the CA App.

 

EDIT: I think GitHub has a bit of an issue currently, I'm downloading between 50-100 KB/s currently...

[KnobleOutlaw]

1 hour ago, ich777 said:

Can you open this file on your local computer

yes I can open it just fine, I do not have any network adblockers, (just browser plugins which shouldn't affect the unraid box)

 

1 hour ago, ich777 said:

double check on the plugin page that the plugin is first uninstalled from the Plugins Error tab if it is in there

removed from errored plugins, rebooted, reinstalled and got this slightly different error message 
 

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

+==============================================================================
| Installing new package /boot/config/plugins/nvidia-driver/nvidia-driver-2022.05.06.txz
+==============================================================================

Verifying package nvidia-driver-2022.05.06.txz.
Installing package nvidia-driver-2022.05.06.txz:
PACKAGE DESCRIPTION:
Package nvidia-driver-2022.05.06.txz installed.

-----ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR------
----Can't get Production Branch version and found no installed local driver-----
-----Please wait for an hour and try it again, if it then also fails please-----
------go to the Support Thread on the unRAID forums and make a post there!------
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

attached diagnotics (unfortunate did it after reboot)

tower-diagnostics-20220721-1342.zip

[ich777]

2 minutes ago, KnobleOutlaw said:

removed from errored plugins, rebooted, reinstalled and got this slightly different error message 

This is basically the same message, are you sure your server uses your default DNS server?

 

It seems it can't download the file? Maybe try to change the DNS server from your Server to 8.8.8.8 for testing purposes.

May I ask where are you located in the world?

 

Also please upgrade to 6.10.3, you are still on an outdated version of Unraid.

[KnobleOutlaw]

I am located on east coast United States, have never had any issue with any other downloads or anything, the unraid server is using my router as a DNS, my router was using 1.0.0.1, I changed it to 8.8.8.8 and am still unable to install. will updating unraid OS version maybe fix the issue?

[ich777]

26 minutes ago, KnobleOutlaw said:

will updating unraid OS version maybe fix the issue?

Good question, I would always recommend to stay on the latest version from Unraid...

 

I will spin up my test server real quick and see if I'm experiencing the same issue.

 

EDIT: Everything working fine over here:

 

 

Do you maybe have a Firewall and blocked some things there?