AMD Lands VCN 3.0 Video Encode Support For Navi 2 / Sienna Cichlid

[phoronix]

Phoronix: AMD Lands VCN 3.0 Video Encode Support For Navi 2 / Sienna Cichlid

The latest open-source/Linux bring-up for AMD's "Sienna Cichlid", a.k.a. Navi 2, is enabling VCN 3.0 video encode support...

http://www.phoronix.com/scan.php?pag...on-Mesa-Merged

[tildearrow]

Then what's the point of bumping the number if you did absolutely nothing?!

Ugh AMD, what a disappointment. Seriously, Intel and NVIDIA have 4:4:4 encode and you do not yet?! And your encoder is still the worst out of all of them?
And why do you have to cripple encoding support on every new iteration of your encoding block? Like I mean, VCE 3.x slowed down H.264 encoding just to allow HEVC encoding to exist...
Neither NVIDIA or Intel had to do that, ever.
And when you did implement 4:4:4, it was only for freaking I-frames (so not P-frames) and therefore too expensive on the bandwidth...
And even worse, this only lasted for the entirety of VCE 2.0's life... ...after that, 4:4:4 encoding support went away!
And it was so undocumented that I think only the PlayStation 4 ever used it for wireless display!

Yeah, I have typed too many "and"s but the truth is that it annoys me so much when AMD brings out a new version of their video encoder FOR NO FREAKING REASON AT ALL.

Like why is your encoder so horrible? Do you even have a team for it or do you rely upon external IP for it? Really?

I mean I know you may have the best open-source support (besides Intel), but sadly NVIDIA and Intel are still winning the hardware encoding race with their 8K HEVC 12-bit and even lossless predictive 4:4:4 support!
Yeah I mean they may not have AV1 yet or whatever, but at least they have 4:4:4!

And you know why do I want 4:4:4 encoding?
Because I record my screen at times and I want pixel color perfection when just recording the desktop.

Come on, could AMD speak on this matter? Why is documentation about new versions of VCE/VCN SO ELUSIVE?! D:

[[email protected]]

Then what's the point of bumping the number if you did absolutely nothing?!

Ugh AMD, what a disappointment. Seriously, Intel and NVIDIA have 4:4:4 encode and you do not yet?! And your encoder is still the worst out of all of them?
And why do you have to cripple encoding support on every new iteration of your encoding block? Like I mean, VCE 3.x slowed down H.264 encoding just to allow HEVC encoding to exist...
Neither NVIDIA or Intel had to do that, ever.
And when you did implement 4:4:4, it was only for freaking I-frames (so not P-frames) and therefore too expensive on the bandwidth...
And even worse, this only lasted for the entirety of VCE 2.0's life... ...after that, 4:4:4 encoding support went away!
And it was so undocumented that I think only the PlayStation 4 ever used it for wireless display!

Yeah, I have typed too many "and"s but the truth is that it annoys me so much when AMD brings out a new version of their video encoder FOR NO FREAKING REASON AT ALL.

Like why is your encoder so horrible? Do you even have a team for it or do you rely upon external IP for it? Really?

I mean I know you may have the best open-source support (besides Intel), but sadly NVIDIA and Intel are still winning the hardware encoding race with their 8K HEVC 12-bit and even lossless predictive 4:4:4 support!
Yeah I mean they may not have AV1 yet or whatever, but at least they have 4:4:4!

And you know why do I want 4:4:4 encoding?
Because I record my screen at times and I want pixel color perfection when just recording the desktop.

Come on, could AMD speak on this matter? Why is documentation about new versions of VCE/VCN SO ELUSIVE?! D:

Unfortunately, hardware accelerated E/D video is the weakest part of AMD's GPUs. They seems to always be behind Intel/Nvidia's competition, feature wise. A couple days ago I was blabbering about how a older Ivy Bridge CPU had 4k/60fps H264 decoding, while newer AMD silicon like the Kaveri only had 1080p/60fps. Intel also got VP9 decoding first.

Shenanigans like this still push people to Intel on the HTPC market, for example.

[AmericanLocomotive]

Limited resources. In 2014, 4K content was basically non-existent, so very little point to waste resources implementing a 4K video decode. When you pinching every possible penny you can, concessions have to be made. AMD in 2014 was not in a very good place at all.

It's going to be a couple of years before you start seeing all of this new Zen income materialize into tangible product differences.

[LoveRPi]

Hiring an team to build video encoders is expensive stuff. Large companies spend $10M+ annually on their encoder optimization teams.

[TemplarGR]

I think it is safe to assume that RDNA2 gpus will support AV1 decoding. I mean, they will arrive at the end of this year or even later, AV1 has been frozen for some time now, it would be silly not to include it.

[agd5f]

Come on, could AMD speak on this matter? Why is documentation about new versions of VCE/VCN SO ELUSIVE?! D:

Because the product hasn't launched yet?

[tildearrow]

Because the product hasn't launched yet?

Sadly, this is the case for already launched products as well.

The Wikipedia page for VCE does not show any useful information at all about VCE 4.0+:

VCE 4.0

The Video Code Engine 4.0 encoder and UVD 7.0 decoder are included in the Vega-based GPUs.^[10][11]

VCE 4.1

AMD's Vega20 GPU, present in the Instinct Mi50, Instinct Mi60 and Radeon VII cards, include VCE 4.1 and two UVD 7.2 instances.^[12][13]

And the page on VCN is only a bit more than a stub, with barely any info at all.

[Laughing1]

Sadly, this is the case for already launched products as well.

The Wikipedia page for VCE does not show any useful information at all about VCE 4.0+:



And the page on VCN is only a bit more than a stub, with barely any info at all.

https://www.x.org/wiki/RadeonFeature...renexthardware

Only has up to VCN2.0 as of yet though.

Needs AV1 decode added.

[tildearrow]

https://www.x.org/wiki/RadeonFeature...renexthardware

Only has up to VCN2.0 as of yet though.

Needs AV1 decode added.

Where is 4:4:4?