Jump to content
XPEnology Community

Nvidia Runtime Library


disone

Recommended Posts

This is really interesting work.  I have a Quadro P620 passed through to a Plex container on my Unraid box and works great - but I would much prefer to use Xpenology.

 

I followed this guide to get it working on Unraid where a lot of the hard work is done in one of their plugins.

 

Fingers crossed for us on the Xpenology side!

 

Edit:

..although looks like the plugin has been pulled as Unraid now supports Nvidia directly.. [ link ]

 

 

 

 

Edited by flannell
Link to comment
Share on other sites

В 12.01.2021 в 02:13, RedwinX сказал:

Hey All, 

 

Very intresting topic. I have installed a GT 1030 in a microserver gen 8, with DSM 6.2.3. I have compiled a custom version of ffmpeg, and still have the same error, cuda initialization failed. 

Driver is ok (I think) :

image.png.aa8215bdd5ce636cc4ba4bfed6d158da.png

image.png.23f1d62003afe1c3b35ca2773e656aa9.png

 

Trying to install NVIDIA Driver : 

image.png.defbcb2b91a0f43848ab442d2d7cc205.png

GT1030 does not support NVENC.

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

image.thumb.png.bf5f6d88c78d29d6815956448f7c70b3.png

 

I have a Quadro P400 rev 1.0 (HP). Can test, if someone can say how. 

nvidia-smi shows P400. How to test ffmpeg?

Edited by Demitsuri
Link to comment
Share on other sites

  • 4 weeks later...

if i change the spk info with an editor like Notepad++ i get errors with unkown fileformat 

has any one got a working spk file for the DS3617xs 

 

I see that the 3617xs has an broadwell arch and not the denverton

The 3617xs  has an Xeon D cpu and the dva3219 has an atom based cpu 

i geus this wont even happen 

Edited by sojab0on
found that ds3617xs and dva 3219 are differnt cpu types
Link to comment
Share on other sites

i got teh package to install but now when i start nvidia-smi it states 
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.
and i cant get it to do it 

installed the NVIDIARuntimeLibrary-dva3219-1.0.1-0014.spk with the INFO file edited to show no model 

Copied the .ko files from the nvidia_440.44_3615_623.zip that was given in a previous post to the /usr/lib/modules/ and ran the 
insmod on all the .ko files but it gave back errors 

 insmod /usr/lib/modules/nvidia.ko
insmod: ERROR: could not insert module /usr/lib/modules/nvidia.ko: Unknown symbol in module
ash-4.3# insmod /usr/lib/modules/nvidia*.ko
insmod: ERROR: could not insert module /usr/lib/modules/nvidia-drm.ko: Invalid parameters

After switching back to the 0011 spk smi started working and the insmod worked as well 
for some reason after a reboot the .ko files arent loaded and smi goes back to cant communicate with driver until i rerun insmod for the .ko files


but now its says 
 nvidia-smi
Unable to determine the device handle for GPU 0000:13:00.0: Unknown Error
Have and GTX690 to test with and in a newer amd build a RTX2070 in bouth i get this error 

quadro P2000 is underway 

 

any sugestions

after changeing a VM value hypervisor.cpuid.v0 = FALSE the nvdia-smi workend and showed 

 

ash-4.3# nvidia-smi
Sun Feb 14 12:55:15 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 381.22                 Driver Version: 381.22                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 680     Off  | 0000:13:00.0     N/A |                  N/A |
| 30%   38C    P0    N/A /  N/A |      0MiB /  1999MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|    0                  Not Supported                                         |
+-----------------------------------------------------------------------------+

on the xeon esxi system with teh gtx 680 

now its as most of us could not get to work the HW transcode to cuda 

 

Edited by sojab0on
Nvidia-smi started running after spk version cahnged
  • Like 1
Link to comment
Share on other sites

  • 6 months later...

what is the state of affairs? is a GTX 1060/70/80 (maybe lower?, GTX 580 in the best case) natively supported? Can xpenology be easily installed as a DVA3221 and use facial recognition for the surveillance station?

Edited by mr412
Link to comment
Share on other sites

@mr412I only had a brief look at Surveillance Station, but it's not something I was interested in. So I don't know how to get it working.

 

Getting an Nvidia card showing in Synology is pretty easy, just follow the above steps.

 

image.thumb.png.08ab2cdf0c42613234f54b2d25a28589.png

 

I am more interested in using it in Docker for something like CompreFace etc

The problem is that although we now have an up to date version of Docker ( 20.10.03 ), it's pretty much useless without getting nvidia-docker2 on there somehow.

 

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#docker

 

It's not something I have had time to investigate though.

Link to comment
Share on other sites

  • 2 weeks later...
On 8/19/2021 at 3:13 PM, wills106 said:

@mr412I only had a brief look at Surveillance Station, but it's not something I was interested in. So I don't know how to get it working.

 

Getting an Nvidia card showing in Synology is pretty easy, just follow the above steps.

 

image.thumb.png.08ab2cdf0c42613234f54b2d25a28589.png

 

I am more interested in using it in Docker for something like CompreFace etc

The problem is that although we now have an up to date version of Docker ( 20.10.03 ), it's pretty much useless without getting nvidia-docker2 on there somehow.

 

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#docker

 

It's not something I have had time to investigate though.

 

 

Can you make an "guide for stupids" how to install nvidia drivers?

I read all the post and I cant understand a ******* 😕

 

 

OK I will try to make a summary:

 

First, install the drivers in /usr/lib/modules/

nvidia-modeset.ko

nvidia-uvm.ko

nvidia.ko

 

Second, install the runtime library:

 

https://archive.synology.com/download/Package/NVIDIARuntimeLibrary

 

I can't download any of the driver files in this thread, my Syonology its an DS3617xs. Where I can find the drivers?

Edited by keteflips
  • Like 1
Link to comment
Share on other sites

As I am using a "DS918" I downloaded the pat file for DS1618 matching the same revision as mine ie  6.2.3-25426-3

 

Extracted:

nvidia-modeset.ko

nvidia-uvm.ko

nvidia.ko

 

Copied them into a folder in Volume1

 

Logged in via ssh and then did sudo -i

 

Copied them from the folder in Volume1 to /usr/lib/modules/

 

Insmod the 3 nvidia files.

 

Downloaded the runtime file and removed all reference to DVA and versions, so on installation it doesn't check it's a DVA machine.

 

Installed the runtime library.

 

Then nvidia-smi works. But need to work out how to get nvidia-docker2 on there so you can use containers in Docker with the GPU

  • Like 1
Link to comment
Share on other sites

  • 2 months later...
On 11/13/2021 at 3:19 AM, Zowlverein said:

Does anyone have the drivers package from 3615 - it looks like there were two versions posted, but both links are down.

 

I have tried with the 1618 package but I get error like this:

insmod: ERROR: could not insert module nvidia.ko: Package not installed
 

Thank you!

 

I case anyone might need it, I was able to build 381.22 for 3615 here:

https://gofile.io/d/Y7zXby

Link to comment
Share on other sites

  • 2 weeks later...

Got 440.44 working as well - actually it works fine using the standard Nvidia installer. Just need to extract to a different directory than default. But for convenience it is here: https://gofile.io/d/L3tRlg

 

Also created a custom build of ffmpeg using the SynoCommunity repo (https://github.com/SynoCommunity/spksrc) which adds nvenc support. This allows me to run the portable version of Jellyfin (after installing dotnet runtime) with nvenc encoding enabled (just need to install the ffmpeg spk and point Jellyfin to "/var/packages/ffmpeg/target/bin/ffmpeg" as the ffmpeg path). This build can be found here: https://gofile.io/d/iIKZaH. I had initially tried the ffmpeg build included in Surveillance Station, and while it has nvenc support and works for some basic encoding tasks it seems that the version is too old for some of the features used by Jellyfin.

 

Now just hoping to get docker working somehow with nvidia support. 

  • Thanks 1
Link to comment
Share on other sites

Have got cuda in docker working (at least nvidia-smi with the sample container: sudo docker run --gpus all nvidia/cuda:10.2-runtime nvidia-smi)

 

Required files here: https://gofile.io/d/zYTBCP

/usr/bin/nvidia-container-toolkit (v 1.5.1 from ubuntu 16.04 build)

/usr/bin/nvidia-container-cli (custom build with syscall security removed)

also need symlink (ln -s /usr/bin/nvidia-container-toolkit /usr/bin/nvidia-container-runtime-hook)

 

/lib/libseccomp.so.2.5.1 (from ubuntu 16.04 - also ln -s /lib/libseccomp.so.2.5.1 /lib/libseccomp.so.2)

/lib/libnvidia-container.so.1.5.1 (from same custom build - also ln -s /lib/libnvidia-container.so.1.5.1 /lib/libnvidia-container.so.1)

 

ldconfig - copied to /opt/bin/ from ubuntu installation (version in entware/optware didn't work)

 

/etc/nvidia-container-runtime/config.toml (need to update file to point to local ldconfig and nvidia drive paths)

 

 

Untitled.png

  • Thanks 1
Link to comment
Share on other sites

On 11/23/2021 at 8:31 PM, Zowlverein said:

Also created a custom build of ffmpeg using the SynoCommunity repo (https://github.com/SynoCommunity/spksrc) which adds nvenc support.

 

8 hours ago, Zowlverein said:

Have got cuda in docker working (at least nvidia-smi with the sample container: sudo docker run --gpus all nvidia/cuda:10.2-runtime nvidia-smi)

 

good that someone did it at last, might be useful to some if they look for a alternative to 918+ and intel qsv

maybe write some small post in

https://xpenology.com/forum/forum/91-additional-compiled-modules/

or

https://xpenology.com/forum/forum/36-tutorials-and-guides/

it should be placed more prominent, especially the docker support might be a big thing, i guess there are some docker images supporting it like jellyfin (free) or plex (not free with transcoding support)

 

btw, are we allowed to redistribute your files if they go offline and you are not around to re-upload?

 

Edited by IG-88
Link to comment
Share on other sites

I'll try to write up some steps in a post when I have a chance. I'm not super confident that I have all the prerequisites listed out above as I also have Entware installed with some additional libraries - so if there is someone with a Nvidia card that wants to try out docker support it might be helpful in getting a complete/accurate guide. And btw, I have now tested with Jellyfin with hardware encoding and handbrake-nvenc (only works up to version 1.3 due to the older driver) and both are working in docker.

 

Yeah, you can go ahead an redistribute if they go offline - I'll also try to fork the github projects I modified to get to the working builds of ffmpeg and libnvidia-container. I would imagine it should also be possible to package the docker addons as an spk - just haven't had a chance to look into it yet.

Link to comment
Share on other sites

  • 2 weeks later...
On 11/30/2021 at 4:40 PM, Zowlverein said:

I'll try to write up some steps in a post when I have a chance. I'm not super confident that I have all the prerequisites listed out above as I also have Entware installed with some additional libraries - so if there is someone with a Nvidia card that wants to try out docker support it might be helpful in getting a complete/accurate guide. And btw, I have now tested with Jellyfin with hardware encoding and handbrake-nvenc (only works up to version 1.3 due to the older driver) and both are working in docker.

 

Yeah, you can go ahead an redistribute if they go offline - I'll also try to fork the github projects I modified to get to the working builds of ffmpeg and libnvidia-container. I would imagine it should also be possible to package the docker addons as an spk - just haven't had a chance to look into it yet.

What a pleasant surprise to see when I checked back in on this. Would you be able to re-upload your files please? gofile is showing that the folder is empty

Link to comment
Share on other sites

  • 3 weeks later...
  • 3 weeks later...
On 1/20/2022 at 10:49 AM, K3K5 said:

link is down, can somebody reupload it?

 

On 11/30/2021 at 10:40 PM, Zowlverein said:

Yeah, you can go ahead an redistribute if they go offline -

 

ok, here we go (edit2: new links)

https://dailyuploads.net/mqt1gwqerzho

https://dailyuploads.net/axivw3phqyo0

https://dailyuploads.net/x3lhh25fqxhy

or

https://usersdrive.com/bzbq1k1zuz09.html
https://usersdrive.com/jx2vfqz52pwi.html
https://usersdrive.com/687a8og3x3aa.html

or

https://gofile.io/d/vSXOPM

or

https://www23.zippyshare.com/v/7bUZiDBj/file.html

https://www23.zippyshare.com/v/lM4C3bmS/file.html

https://www23.zippyshare.com/v/0d13K09h/file.html

 

nvidia440443615.tar.xz

docker.tar.xz

ffmpeg_x64-6.2.3_4.3.3-39+nvenc.spk

 

Edited by IG-88
Link to comment
Share on other sites

  • 1 month later...
  • 5 weeks later...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...