[Plugin] Nvidia-Driver

[ich777 773]

48 minutes ago, rainfactor said:

I wouldn't mind however the R420 Pcie slot can only give 30W and that is the reason why I went with the 730 as it draws only25W

I think the P400 is the perfect choice it only draws 30W (but please be sure to get a P400 and not a 400 those two cards look almost the same, a 400 is not capable of hw transcoding).

 

Here are the specs:

https://www.pny.eu/consumer/explore-all-products/nvidia-quadro/895-nvidia-quadro-p400

 

2 hours ago, Spucoly2 said:

This is what I get from inside the docker container terminal.

This seems perfectly fine to me.

Is this the first time you try hw transcoding?

[jbartlett 195]

Apologies if it has been covered but I've been over half of the pages and haven't seen anything related.

 

I have two nVidia cards installed on my system.

 

[10de:1c30] 41:00.0 VGA compatible controller: NVIDIA Corporation GP106GL [Quadro P2000] (rev a1)

[10de:10f1] 41:00.1 Audio device: NVIDIA Corporation GP106 High Definition Audio Controller (rev a1)

 

[10de:1d01] 42:00.0 VGA compatible controller: NVIDIA Corporation GP108 [GeForce GT 1030] (rev a1)

[10de:0fb8] 42:00.1 Audio device: NVIDIA Corporation GP108 High Definition Audio Controller (rev a1)

 

The first card, the Quadro P2000, is bound via VFIO on the System Devices page and is attached to a VM. When the VM is not in use, I'd like to use it in a Folding@Home Docker container. From what I read, it was OK to use a graphics card on a VM or Container, but not at the same time and that at least one card had to NOT be bound to VFIO.

 

The GT 1030 card is the boot card and is seen by this plugin. However, the Quadro P2000 is not.

 

Any advice?

 

root@VM1:~# nvidia-smi
Thu Jan 21 22:56:49 2021       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 455.45.01    Driver Version: 455.45.01    CUDA Version: 11.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce GT 1030     Off  | 00000000:42:00.0 Off |                  N/A |
| 42%   34C    P0    N/A /  30W |      0MiB /  1999MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

 

[itimpi 875]

3 minutes ago, jbartlett said:

The first card, the Quadro P2000, is bound via VFIO on the System Devices page and is attached to a VM. When the VM is not in use, I'd like to use it in a Folding@Home Docker container. From what I read, it was OK to use a graphics card on a VM or Container, but not at the same time and that at least one card had to NOT be bound to VFIO.

 

Once a card is bound to VFIO it is not available to be used either by Unraid or by any Docker container.  

[ich777 773]

6 minutes ago, jbartlett said:

From what I read, it was OK to use a graphics card on a VM or Container, but not at the same time and that at least one card had to NOT be bound to VFIO.

As @itimpi said a card that you bind to VFIO is "not available" to Unraid and is only visible or usable in VM's.

 

If you want to use it that way don't bind it to VFIO. But keep in mind that this could lead to serious problems (hard lockups,...).

[jbartlett 195]

Tried using pci-stub instead of VFIO but that netted the same result. Looks like if I want to use a graphics card with both Docker and a VM (but of course not at the same time), I'll need to create separate boot configs - one with the graphics card stubbed out for the VM and one without for the Docker.

[ich777 773]

2 hours ago, jbartlett said:

Tried using pci-stub instead of VFIO but that netted the same result.

Yes, that's also the same because you are "hiding" it from Unriad, that doesn't make sense that you stub it.

 

2 hours ago, jbartlett said:

I'll need to create separate boot configs - one with the graphics card stubbed out for the VM and one without for the Docker.

No, just don't stub it or bind it to VFIO and you are good to go.

 

Why are you stubbing the card?

 

You can also assign the card to the VM when it's not stubbed or bound to VFIO

[Spucoly2 1]

13 hours ago, Spucoly2 said:

This is what I get from inside the docker container terminal.

I got it. I had to go on the actual unraid server in GUI mode and sign in to plex with my plex pass account. That is strange as I have never had to do that before.

 

[jbartlett 195]

8 hours ago, ich777 said:

No, just don't stub it or bind it to VFIO and you are good to go.

 

Why are you stubbing the card?

 

Because at one point in time, a video card wouldn't even show up in the VM editor unless it was stubbed out. I missed that this restriction was removed.

[jbartlett 195]

Is it possible to add more than one GPUID in the "Nvidia Visible Devices" variable? If so, what is the delimiter?

[ich777 773]

9 hours ago, jbartlett said:

Is it possible to add more than one GPUID in the "Nvidia Visible Devices" variable? If so, what is the delimiter?

Yes that is possible, you can enter the ID of the GPU: '0,1,2,3,4' or you can simply type in 'all'

[Proxykn 2]

hey,

I installed the nvidia driver from Community Applications, download and install was ok so i press the done button, then went to check nvidia driver in the setting tab and open it from there but nothing happens, it's just a white screen.

in system log i get this error

 

Jan 23 18:37:13 ServerTower nginx: 2021/01/23 18:37:13 [error] 5348#5348: *739 upstream timed out (110: Connection timed out) while reading upstream, client: 192.168.0.221, server: , request: "GET /Settings/nvidia-driver HTTP/1.1", upstream: "fastcgi://unix:/var/run/php5-fpm.sock:", host: "192.168.0.51", referrer: "http://192.168.0.51/Plugins"

 

i'm stuck at this point and dont know what to do about it

Edited January 23 by Proxykn

[ich777 773]

51 minutes ago, Proxykn said:

I installed the nvidia driver from Community Applications, download and install was ok so i press the done button, then went to check nvidia driver in the setting tab and open it from there but nothing happens, it's just a white screen.

in system log i get this error

What card have you installed, have you also read the Troubleshooting und how to report problems section on the first page (first post on the bottom)?

Have you restarted and followed the guid in the first post?

Which card do you have? Have you bound a card to VFIO? Is this the first time you try transcoding or was it running before?

[Proxykn 2]

27 minutes ago, ich777 said:

What card have you installed, have you also read the Troubleshooting und how to report problems section on the first page (first post on the bottom)?

Have you restarted and followed the guid in the first post?

Which card do you have? Have you bound a card to VFIO? Is this the first time you try transcoding or was it running before?

sorry for the lack of info in my first post

 

i put a gtx 1060 in my unraid rig that came from my older gaming pc, first time to try transcoding with gpu in my unraid server.

 

when i use nvidia-smi i get this

 

when i try to run the nvidia-smi again in terminal nothing happens

 

this is the driver plugin page it just keeps loading and then it stops

 

i have not bound my card to VFIO

i did everything in first post until step nr4

nr4 does not show up for me.

Edited January 23 by Proxykn

[ich777 773]

29 minutes ago, Proxykn said:

when i try to run the nvidia-smi again in terminal nothing happens

I get this once with a GTX960 the main problem was that it was not getting enough power.

 

30 minutes ago, Proxykn said:

this is the driver plugin page it just keeps loading and then it stops

The driver plugin page waits for the output of the command 'nvidia-smi' that's why you get nothing displayed because it simply times out and no result from 'nvidia-smi' is received.

 

Please check if you GTX1060 is getting enough power from the power supply and also try to reseat the card in a differnt PCIe slot if possible.

 

But the card is working fine I think? Have you got a spare box where you can test the card on Windows (with graphics drivers installed - you have to install the graphics drivers to see if the card is fully working, the basic display output works always fine).

[Proxykn 2]

5 minutes ago, ich777 said:

Please check if you GTX1060 is getting enough power from the power supply and also try to reseat the card in a differnt PCIe slot if possible..

 

I got a 550W power supply only 3 HDD, 2 SSD connected and Network card. CPU is i5-9600k so don't think that GPU is not getting enough power.

i only have 1 pcie 16 slot and never used anything in that slot.

 

GPU worked fine when i pulled it of old rig and put in its original box with packing but i wil test to make sure.

 

[ich777 773]

9 minutes ago, Proxykn said:

I got a 550W power supply only 3 HDD, 2 SSD connected and Network card. CPU is i5-9600k so don't think that GPU is not getting enough power.

I think that should be enough, are you interested in transcoding? That is also possible with your i5, just download the Plugin Intel-GPU-TOP from me and add a device '/dev/dri' in your template of the Docker Container and it should transcode just fine (no need to edit the go file or create any other files, this is all done by the plugin - just be sure to set the Primary Graphics Card to your 'iGPU' and not to 'PCIe' in your BIOS).

You can also install the Plugin GPU Statistics from @b3rs3rk to watch the utilisazion.

 

Please not the Intel transcoding is maybe more power efficient than the Nvidia one and Intel has no artificial restrictions of parallel transcodes.

 

9 minutes ago, Proxykn said:

GPU worked fine when i pulled it of old rig and put in its original box with packing but i wil test to make sure.

Please report back about your findings.

[Proxykn 2]

2 hours ago, ich777 said:

I think that should be enough, are you interested in transcoding? That is also possible with your i5, just download the Plugin Intel-GPU-TOP from me and add a device '/dev/dri' in your template of the Docker Container and it should transcode just fine (no need to edit the go file or create any other files, this is all done by the plugin - just be sure to set the Primary Graphics Card to your 'iGPU' and not to 'PCIe' in your BIOS).

You can also install the Plugin GPU Statistics from @b3rs3rk to watch the utilisazion.

 

Please not the Intel transcoding is maybe more power efficient than the Nvidia one and Intel has no artificial restrictions of parallel transcodes.

 

Please report back about your findings.

GPU works fine, did stress test and all is good, fans ramp up when temp go up and then are steady at 100% use

 

i had this gpu laying around and wanted to use it instead of laying in box getting old and doing nothing:)

 

i got the Intel-GPU-TOP working but not sure how powerfull it is at transcoding but i wil work with it for now and see, but i would like the 1060 to work also.

 

Edited January 23 by Proxykn

[SeeGee 8]

Are there any power management features available? Possible methods of reducing power usage during GPU idle, etc?

 

[ich777 773]

9 hours ago, Proxykn said:

GPU works fine, did stress test and all is good, fans ramp up when temp go up and then are steady at 100% use

 

i had this gpu laying around and wanted to use it instead of laying in box getting old and doing nothing

This is really strange...

Can you try it once more in Unraid and let the default primary gpu in the BIOS at iGPU please.

 

You can install actually both plugins, the Intel one and also the Nvidia-Driver.

 

9 hours ago, Proxykn said:

i got the Intel-GPU-TOP working but not sure how powerfull it is at transcoding but i wil work with it for now and see, but i would like the 1060 to work also.

The Intel transcoder is very capable of multiple 4K streams at once (depending on the quality and bitrate) and some say that it transcodes even better than NVENC (but I only heard of that ), uses way less power and in idle it can set it's power to 0W if that's important to you.

 

Also @mgutt wrote a thread about that here:

 

 

[ich777 773]

4 hours ago, SeeGee said:

Are there any power management features available? Possible methods of reducing power usage during GPU idle, etc?

The GPU should go to P8 if it's idle, but keep in mind that the readings of 'nvidia-smi' are often wrong since most features are broken...

For example my GTX1050 displays most of the time the wrong power state, sometimes when it's transcoding P0 but most of the time P8 and when it's in idle it displays most of the time also P0 but it's actually in P8 and at last the power reading of the GTX1050's are mostly broken for all cards.

 

On my GTX1060 this is a completely different story, the reading is completely right and even reports the right power draw...

 

If you wan't to limit the card's power usage you can do that by:

nvidia-smi -i 0 -pl 50

This will set the card 0 to a power limit of 50W.

[Proxykn 2]

5 hours ago, ich777 said:

This is really strange...

Can you try it once more in Unraid and let the default primary gpu in the BIOS at iGPU please.

 

 

i wil try for sure, might not get it done today got other things to attend to.

 

5 hours ago, ich777 said:

You can install actually both plugins, the Intel one and also the Nvidia-Driver.

 

The Intel transcoder is very capable of multiple 4K streams at once (depending on the quality and bitrate) and some say that it transcodes even better than NVENC (but I only heard of that ), uses way less power and in idle it can set it's power to 0W if that's important to you.

 

Also @mgutt wrote a thread about that here:

 

 

 

if the iGPU is this good then it wil cover my needs perfect and might put the 1060 back in it's box (poor thing)  but i wil see if i can get the 1060 to work first before i do that  

 

thank you for pointing me to this option had no clue this was possible, stil a newbie in the unraid world  

[Proxykn 2]

19 minutes ago, Proxykn said:

 

i wil try for sure, might not get it done today got other things to attend to.

 

 

if the iGPU is this good then it wil cover my needs perfect and might put the 1060 back in it's box (poor thing)  but i wil see if i can get the 1060 to work first before i do that  

 

thank you for pointing me to this option had no clue this was possible, stil a newbie in the unraid world  

 

1 more question i had about the iGPU, i have small screen where my server sit but when i downloaded the Intel-GPU-TOP the screen went out, i guess you can't have both giving picture to my screen and transcoding?

 

i'm sorry if this is not the right place to ask.

Edited January 24 by Proxykn

[ich777 773]

40 minutes ago, Proxykn said:

1 more question i had about the iGPU, i have small screen where my server sit but when i downloaded the Intel-GPU-TOP the screen went out, i guess you can't have both giving picture to my screen and transcoding?

You should have a display output, eventually it's the monitor that can't display the resolution or the connection type, hdmi, dp,... I also know another user that has a 9900 and he sees the terminal output just fine.

 

Or are you running Unraid in GUI mode?

[Proxykn 2]

42 minutes ago, ich777 said:

You should have a display output, eventually it's the monitor that can't display the resolution or the connection type, hdmi, dp,... I also know another user that has a 9900 and he sees the terminal output just fine.

 

Or are you running Unraid in GUI mode?

 

yes it was in GUI mode

[ich777 773]

2 hours ago, Proxykn said:

yes it was in GUI mode

Eventually this is the problem, please try to boot the server in console mode and see if you got a output.