10-bit HEVC Decoding For RadeonSI Gallium3D Appears Fit

[andreano]

As long as HEVC is incompatible with free software (like its predecessor AVC), Firefox will probably take it upon themselves to fight it – they will refuse to support it as long as they can, just as they did with AVC. Other free software browsers are in the same boat, but may not care. I, for one, really hope VP9/AV1 can win the web this time.

That's actually a bit important for the interoperability of the web, folks!

[liam]

Hardware solutions or even hybrid solutions cut a lot of corners to get the more speed and hardware costs down (limited buffers, limited search, leaving out some functionality - sometimes even basic things like b-frames). The main purpose of the hardware encoders is real-time (streaming, video-conference stuff) encoding, so they don't / can't achieve the best quality / size ratios.

I know the arguments, but i was looking for a shoot-out (i.e., a site that has encoded some scenes with both methods and offers an a/b test).
Also, please link to the paper which proves that an asic can't encode video with quality as good as software while still being faster.
It's not my intent to be a jerk about this but my feeling about this area is that you seem to see similar sorts of arguments as you get from folks who are against digital audio encoding (of at least digital encoding that is less than 24 bits sample depth). The difference, i think, is that i care most about effective transparency once use case is taken into account rather than absolute quality that is really only useful for ideal viewing situations (thx-ish types of parameters).

[andreano]

prove that an asic can't encode with quality as good as software

It's not a matter of can't, but software will always have more features for less money. Encoding is a hard problem (in contrast to decoding) – you have to leave something out if you have a time or cost budget to meet.

So when disregarding time and energy consumption, only considering quality per bitrate (aka efficiency), common knowledge is that software should win.

You can buy a broadcast encoder if you need to approach software quality/bitrate in real-time. That will cost you a little fortune, and it comes with datacenter grade cooling fans – you can't expect consumer grade encoders (your mobile phone) to match that. Yes, consumer grade hardware encoders are really as bad as quikee describes.

[pal666]

I, for one, really hope VP9/AV1 can win the web this time.

that's obvious outcome considering it is supported by everyone from hardware vendors to content producers to browser vendors

[liam]

It's not a matter of can't, but software will always be years ahead of the asics, and have more features for less money.

When disregarding time and energy consumption, only considering quality per bitrate (aka efficiency), common knowledge is that software should win.

Though you say "it's not a matter of can't" the rest of your sentence seems to suggest exactly that (specifically the "always" part). To provide a little bit of push-back to that idea consider that you are disregarding (very, very likely) paradigm shifts (trivially, much cheaper FPGAs, since at this moment, we are seeing them used in more places than ever which, all else being equal, implies lower production costs). If we look a bit to recent history we have relied on special purpose hardware to make concepts that were/are infeasible in "software" (possibly not ever actually implemented due to complexity and issues of practicality) actually useful (gpus, obviously, but also the dsps, hardware codecs, and dedicated hardware instructions in a cpu isa).
Lastly, if you disregard time and energy, you are using a different definition of efficiency than I'm familiar
If you look to the mobile hardware space you'll see this really interesting cycle of: next gen processor (could be a gpgpu) -> introduction of previously too expensive algorithm -> new algorithm becomes essential -> introduction of equivalent functionality to the algorithm in hardware added to soc.

[andreano]

obvious outcome

I don't feel too certain.

One problem is that HEVC is used in broadcast (e.g. DVB-T2), and when businesses with roots in this market try to deliver on the internet, they come with the wrong expectations (in their filter bubble, the internet is called "OTT", is full of HEVC, and browsers are replaced by "devices").

Another is that HEVC seems to be more hyped in the general population, steal more headlines and be more prevalent in tech specs – why is this article not about 10-bit VP9, for example?

Thirdly, 99% of the world's population don't get why implementability might matter to a web standard: when they read headlines like "Netflix found HEVC 20% better than VP9" for example, they really think that's all there is to consider! If you want to cry, Safari users recently unable to play 4K videos on Youtube are flaming hatefully right now.

[Ansla]

Safari users recently unable to play 4K videos on Youtube are flaming hatefully right now.

And so they should, the problem is where their anger is directed. Somehow I think they consider Google responsible for that, not their beloved Apple.

[pal666]

I don't feel too certain.

due to lack of information

One problem is that HEVC is used in broadcast (e.g. DVB-T2)
....
"Netflix found HEVC 20% better than VP9"

in reality netflix found hevc unsafe to use and joined aomedia along with other internet broadcasters

https://youtu.be/lzPaldsmJbk?t=4m32s

Another is that HEVC seems to be more hyped in the general population, steal more headlines and be more prevalent in tech specs – why is this article not about 10-bit VP9, for example?

because vp9 is not backed by every hardware manufacturer, unlike av1

now i think i missed vp9 part of your original post and replied to av1 only

[andreano]

different definition of efficiency than I'm familiar with

Jargon for coding efficiency (yes, that's the E in HEVC).

https://youtu.be/lzPaldsmJbk?t=4m32s

I love that video — packed with previously undisclosed info. He says Netflix was an early adopter of HEVC but bailed when they saw the price tag.

Here is something for both of you: Guest editor at streamingmedia.com predicting rise of HEVC and software encoding in 2017

This guy is representative of what I'm talking about: The online video delivery industry (except Netflix) is heading in the opposite direction of the rest of the world! How can you argue about the success of HEVC without a word on VP9 or AV1? How can you predict the future of online video based on smart TVs and phones, disregarding browsers? His company wants HEVC to succeed, thus the peculiar choice of facts. There is the occasional mention of VP9 and AV1 on streamingmedia.com, but generally, I see too much unfounded ignorance. Welcome to the splinternet.