Jump to content

Emby server detecting Intel GPU as VAAPi not as Quicksync


bgeelen

Recommended Posts

bgeelen

Hi,

 

I have the following setup.

ESXI 6.7 running an ubuntu server 20.04 VM.

Emby is running in a docker container and i passed trough 

I Passed trough the intel 630 iGPU to the ubuntu device.

I made sure the there are no permission issues:

root@host:~# ls -l /dev/dri
total 0
drwxrwxrwx 2 root root         80 Feb 19 20:36 by-path
crwxrwxrwx 1 root video  226,   0 Feb 19 20:36 card0
crwxrwxrwx 1 root render 226, 128 Feb 19 20:36 renderD128

Emby shows this at the hardware transcoding page:

image.thumb.png.f4acc931f59a23e7d6440823b037cb29.png

 

Attached are the logs.

Now i have a these of questions:

1) Is there a difference between VAAPi and Quicksync ?

2) If so, any clue how to make the device work as quicksync device ?

 

On the hardware setup page it is not realy clear if there is a difference between them. 

The main page shows this:

image.thumb.png.2669e4e5860bab0c945eeef32d76eba5.png

 

But on the actual linux hardware page the Quicksync part looks like it belongs under the same part as the VAAPi.

image.thumb.png.375882790d2bdea3f1ea5bf3343daab7.pngimage.thumb.png.e8c34be018bee1a85d9f8df3b4f68102.png

 

Please help me understand what to do.

hardware_detection-63749401649.txt embyserver (2).txt

Link to post
Share on other sites
softworkz

Yes, there are differences but also similarities. QuickSync works on top of VAAPI and provides more settings, better quality and sometimes also better performance.

At this time, it's best to wait until the next or subsequent Emby beta, where we will support QuickSync out of the box (if all goes well).

 

Link to post
Share on other sites
Ratatouille

Great news should we expect a big difference in terms of performance vaapi vs quick sync?

Edited by Ratatouille
Link to post
Share on other sites
softworkz

Normally it was more about QSV (Intel Media SDK) providing more options and somewhat better quality, but recently I had compared both on a Tiger Lake CPU and QSV was 1.5 times faster than VAAPI doing the same task.

Now, that was just a single test on a single system and that doesn't say much. But I'm curious to see what users will report in this regard.

  • Thanks 1
Link to post
Share on other sites
  • 3 weeks later...
Ratatouille

Well  i am using 4.6.0.33 I still do not see any difference .

Regarding this are you gonna implement "quick sync always on" supposedly available with Rocket Lake chips. This technology is supposed to work together with an external GPU?

Link to post
Share on other sites
softworkz
1 hour ago, Ratatouille said:

Well  i am using 4.6.0.33 I still do not see any difference .

It's still pending.

1 hour ago, Ratatouille said:

Regarding this are you gonna implement "quick sync always on"

What do you mean by "always on"?

Link to post
Share on other sites
Ratatouille

from https://www.tomshardware.com/news/intel-rocket-lake-release-date-specifications-performance-all-we-know

"Intel's Xe graphics also bring in-built AV1 decode acceleration, which is helpful because the codec reduces bandwidth up to 20% for video streaming (such as 4K and 8K content). Rocket Lake also features Intel's 'new' Always-On QuickSync Video. This feature allows Xe Graphics to run concurrently with a discrete GPU so you can offload some workloads, like video streaming with QuickSync, to alleviate the burden on your discrete GPU. This feature has already been around for several years, but it required BIOS support, and you had to manually enable the integrated graphics unit after you installed a video card. Intel says the technique is now a supported and validated configuration that's enabled by default in the BIOS."

So I suppose we will be able to use quick sync in same time as an Nvidia GPU card or at least  one of the other with automatic switching

Link to post
Share on other sites
softworkz

Thanks for the reference. That article is somewhat inaccurate.

It is correct about the feature for GPU switching, having an integrated GPU for low-power use, combined with a discrete powerful GPU. This has existed in fact for quite a while.
But it's never been about video acceleration. There's no automatic switching possible at all for this.

Even in the combination Intel-onboard + Intel-discrete GPU, I don't think that there will be a "hot-switching". Actually the article doesn't even say that precisely. It just says that both GPUs can be used - which is nothing really spectacular at all. I mean, you have two GPUs - and then....you will be able to use both... what a sensation! LOL

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...