4.0 - Transcoding Question - AMD/Nvidia Graphics Cards

[davedick 19]

With the release of Emby 4.0, does hardware transcoding work well with AMD and Nvidia graphic cards?

 

I'm currently using an Nvidia 1060 3 GB, but was thinking of switching it to a AMD RX 580, as the Nvidia card can handle 2 streams, and the AMD card can handle more (based on what I read in multiple places)

 

Has anybody noticed a transcoding difference between using Nvidia compared to AMD ?

 

Last, if the graphics card is at the max transcodes (Nvidia) or at capacity (AMD), does Emby automatically switch to using the CPU for transcoding?

 

Thanks

Edited January 16, 2019 by davedick

[Luke 42077]

 

 

does Emby automatically switch to using the CPU for transcoding?

 

Yes it does.

 

@@softworkz may have more insight. Thanks.

[clollar 9]

Great questions.  I too am interested in some information about gpus vs cpus...and apus, lol.  I have an AMD apu...A10 7850k currently in the server, but my main system has an AMD fx 6120 six core, but it has a GTX 960.  So I'd like some clarification on which would best suit emby, vs which would best suit daily use, gaming, etc.  Plus I'm building a new PC for my main system, I'd like to know which combo of leftovers makes the best Emby server for my uses.  Thanks in advance!  Can't wait to sink my teeth into the new update fellas! 

[clollar 9]

Specifically if anyone has an idea of what my shared ram/vram setting should be for my a10 7850k?  Basically would I benefit from committing more ram to the cpu, or the integrated graphics?  The integrated gpu ram tops out at 2gb, I currently just set it back to auto, which seems to have it a 1gb now?

[softworkz 5066]

Has anybody noticed a transcoding difference between using Nvidia compared to AMD ?

 

Last, if the graphics card is at the max transcodes (Nvidia) or at capacity (AMD), does Emby automatically switch to using the CPU for transcoding?

 

Yes it does.

 

@@softworkz may have more insight. Thanks.

 

To be precise - yes for the first part and no for the second part.

 

What we are doing right now is "error fallback", which means that on any error due to hw acceleration, we're falling back to software transcoding.

 

What we're not doing yet is load balancing between multiple hw acceleration methods.

But we have started preparing for that:

• We can identify and address individual hw encoders and decoders, even when there are multiple GPU boards of the same type

  .

• We have priority ordering in the codec configuration

 

So stay tuned, there's more to come...

[davedick 19]

I assume you mean "yes to the first part" refers to the difference between Nvidia and AMD?  If so,  can you give some feedback on the difference?

 

Thanks!!

 

To be precise - yes for the first part and no for the second part.

 

What we are doing right now is "error fallback", which means that on any error due to hw acceleration, we're falling back to software transcoding.

 

What we're not doing yet is load balancing between multiple hw acceleration methods.

But we have started preparing for that:

• We can identify and address individual hw encoders and decoders, even when there are multiple GPU boards of the same type
  .
• We have priority ordering in the codec configuration

 

So stay tuned, there's more to come...