[Plugin] Linuxserver.io - Unraid Nvidia


Recommended Posts

1 hour ago, IamSpartacus said:

Has anyone tested background transcoding (mobile sync or optimized versions) performance on a P2000?  Trying to determine if I'd see a significant perforboost over my E5-2680v3.

I've found my p600 do well with sync, incredibly fast compared to CPU

Link to comment
5 minutes ago, IamSpartacus said:

I appreciate that sentiment but that doesnt really give me any hard numbers or frame of reference to go off of.  What kind of CPU are you co.paring it to and how much faster is it in comparison?

You talking emby or plex?

 

Latest emby beta now does a movie at 500 fps on my p400.  From original bitrate 1080p to 3 Mbps 1080p

 

Paired with a 1920x I only see a 2-4 percent cpu usage.

Link to comment
5 minutes ago, Dazog said:

You talking emby or plex?

 

Latest emby beta now does a movie at 500 fps on my p400.  From original bitrate 1080p to 3 Mbps 1080p

 

Paired with a 1920x I only see a 2-4 percent cpu usage.

I'm interested in both.  I realize Emby will perform better since Plex can't do both encode and decode yet in Linux.

 

Can you elaborate a little more?  What exactly does 500fps equal time wise and what's the original bitrate of the file you're giving as an example here?

Edited by IamSpartacus
Link to comment
4 hours ago, IamSpartacus said:

Can you elaborate a little more?  What exactly does 500fps equal time wise and what's the original bitrate of the file you're giving as an example here?

*Very* rough calculation.  Assuming a 2hr movie at 25fps means you've got 120x60x25 = 180,000 frames. 180,000 / 500 = 360 seconds = 6 minutes.

Link to comment

Any ideas as to why my 1050ti isn't being populated in the plugin? All I see is 

Quote

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

 

My system seems to recognize the card.

Quote

 *-display UNCLAIMED
       description: VGA compatible controller
       product: GP107 [GeForce GTX 1050 Ti]
       vendor: NVIDIA Corporation
       physical id: 0
       bus info: pci@0000:03:00.0
       version: a1
       width: 64 bits
       clock: 33MHz
       capabilities: pm msi pciexpress vga_controller cap_list
       configuration: latency=0
       resources: memory:f4000000-f5ffffff memory:e0000000-e7ffffff memory:e8000000-ebffffff ioport:cc80(size=128) memory:c0000-dffff

 

Edited by dr_jared88
Link to comment
1 hour ago, hus2020 said:

Hi I run the official Plex Docker under unraid. I see that every-time I patch the plex transcoder to plex transcoder2, the next day it reverts back to original. Is the official docker doing background update check every day? Can anyone provide inputs...

Yes, most likely you have a appdata backup being run that is stopping then restarting docker. Check your logs.

Link to comment
On 2/22/2019 at 9:06 AM, saarg said:

The vaapi errors might have something to do with it. Might be something in binhexs emby container. You don't pass through an igpu? 

 

Can you test installing the original emby container and see if that works? 

 

When you have encoding and decoding working for plex. Do you notice CPU still spiking? I have spikes upto 45% and hover around 15%. I see decoding and encoding working via nvidia-smi. 

You seem very knowledgeable and might have some insight. I thought it might be audio transcoding but after reading up nvdec should be handling this also?

Link to comment
On 2/9/2019 at 4:15 AM, linuxserver.io said:

<SNIP>

Feature suggestion for the Plex Docker (suggesting it here because it's relevant here)
Replace the /usr/lib/plexmediaserver/Plex Transcoder binary with a script that calls that binary like so:

#!/bin/sh
if [[ -n NVIDIA_VISIBLE_DEVICES ]]; then
     /usr/lib/plexmediaserver/Plex\ Transcoder2 -hwaccel nvdec "$@"
else
     /usr/lib/plexmediaserver/Plex\ Transcoder2 "$@"
fi



Plex versions > 1.15 have nvdec support enabled in their FFMPEG build - but it isn't invoked ever. Using this the GPU is successfully used for both encoding and decoding. This can be verified via `nvidia-smi dmon -s u` and watching the `dec` column.
Link to comment
20 minutes ago, sturmstar said:

A quick question - are there already any other dockers which are using the GPU? I wan't my GPU to work for it's Money! 😀

I've been attempting to get the folding@home docker working with it, but I'm not having any luck.  I've posted for support here and on the F@H docker support thread, but this area seems to be flooded with Plex and Emby only questions.  It would be nice to know if this addition was created solely to support plex and emby and no other docker applications will be supported.

 

Folding on a GPU in a docker sounds so much nicer than passing it through to a VM and eating up a ton of disk space and resource overhead to run a full fledged VM.

Link to comment
5 hours ago, lasewun said:

When you have encoding and decoding working for plex. Do you notice CPU still spiking? I have spikes upto 45% and hover around 15%. I see decoding and encoding working via nvidia-smi. 

You seem very knowledgeable and might have some insight. I thought it might be audio transcoding but after reading up nvdec should be handling this also?

 

I barely tested transcoding during development and some troubleshooting, so don't have too much experience about it. Never tried the decoding patch for plex. 

But audio is not handled by the GPU at all as far as I know. 

Link to comment
1 hour ago, Tiger770 said:

I've been attempting to get the folding@home docker working with it, but I'm not having any luck.  I've posted for support here and on the F@H docker support thread, but this area seems to be flooded with Plex and Emby only questions.  It would be nice to know if this addition was created solely to support plex and emby and no other docker applications will be supported.

 

Folding on a GPU in a docker sounds so much nicer than passing it through to a VM and eating up a ton of disk space and resource overhead to run a full fledged VM.

As long as the software supports Cuda it should be able to use the GPU. But if it uses some other api I don't think it will work. Then you might have to pass through /dev/dri and install the same driver that is installed on the host in the container. 

But this is something we have not tested at all. 

Link to comment
4 hours ago, Xaero said:

Feature suggestion for the Plex Docker (suggesting it here because it's relevant here)
Replace the /usr/lib/plexmediaserver/Plex Transcoder binary with a script that calls that binary like so:


#!/bin/sh
if [[ -n NVIDIA_VISIBLE_DEVICES ]]; then
     /usr/lib/plexmediaserver/Plex\ Transcoder2 -hwaccel nvdec "$@"
else
     /usr/lib/plexmediaserver/Plex\ Transcoder2 "$@"
fi

 

 


Plex versions > 1.15 have nvdec support enabled in their FFMPEG build - but it isn't invoked ever. Using this the GPU is successfully used for both encoding and decoding. This can be verified via `nvidia-smi dmon -s u` and watching the `dec` column.

 

We are not going to patch plex in our container. 

Link to comment

Something other strange occured to me - for the second time - ...

 

Is Plex "autoupdating" itself inside the Docker? (plexinc:plexpass) - reason why I ask is because it happend the second time that the "transcoder" script file is replaced by the original transcoder file. Same size and age than transcoder2 ...

 

hmmm.... any ideas?

Link to comment

image.png.43e9ce993e572c11f627c06ee6fe744f.png

 

Something else - not important overall - but I just stumpled upon.

 

If the VM with GPU Passthrough (1070) is not startet - both GPUs are showing here inside nvidia-smi. But why is nvidia-smi showing a workload of constantly 5% for this GPU? It's GUID is nowhere stated, and as I said - Passthrough is working fine (it disappears and reappears).

 

So is this a indication error or could "something" really use this GPU?

 

best regards

Stefan

 

Link to comment
10 hours ago, lasewun said:

When you have encoding and decoding working for plex. Do you notice CPU still spiking? I have spikes upto 45% and hover around 15%. I see decoding and encoding working via nvidia-smi. 

You seem very knowledgeable and might have some insight. I thought it might be audio transcoding but after reading up nvdec should be handling this also?

What type of files are you using for testing and what CPU? 

 

From my testing, some cpu spiking is normal. If you're converting the audio, especially down mixing (I think that is the correct term) from a TRUEHD 7.1 or a DTS file down to a stereo mix, it can put some load on the cpu.  On 2 Atmos TrueHD 7.1 & DTS-HD Master Audio 7.1--> stereo transcodes on a i7 - 4770 i was getting about 25% usage with encode and decode both on.  Cpu is handling that aspect.

Link to comment
  • trurl locked this topic
Guest
This topic is now closed to further replies.