Intel Gen12/Xe Graphics Have AV1 Accelerated Decode - Linux Support Lands

[mikkl]

AMD! Hurry up!
Look! Intel is beating you! What if they eventually end up bringing AV1 encode too?! (on top of 4:4:4 and their high-quality encoder)
...yeah, while you are stuck in:

Just to make it clear, Tigerlake-U is going to support the main profile of AV1 which means 8/10 bit 4:2:0. This might change with RKL-S next year or future Gen12 versions but TGL-U won't support 12 Bit 4:4:4 unlike VP9. However the main profile will be enough for the likes of youtube, this is a potentially massive feature for any mobile device.

[tildearrow]

Just to make it clear, Tigerlake-U is going to support the main profile of AV1 which means 8/10 bit 4:2:0. This might change with RKL-S next year or future Gen12 versions but TGL-U won't support 12 Bit 4:4:4 unlike VP9. However the main profile will be enough for the likes of youtube, this is a potentially massive feature for any mobile device.

I can understand that for now, considering AV1 is still fresh...

What I can't understand is the lack of 4:4:4 H.264 (at least!) encoding on AMD cards...

[andreano]

Id guess that cameras will pick one format, and TVs will follow. Id further guess that this will be H266 or EVC, as HW for realtime encoding is a design criteria for those.

Network effects and path dependence: I would agree about the principles, but cameras – hardware encoders in price sensitive consumer devices – is hardly the forefront of development. I would rather bet that software will always come first and more or less decide the game. This is where browsers come in.

VVC vs EVC in terms of risk: One is a superset of an ongoing licensing fiasco; the other is more defensively built than even AV1.

[starshipeleven]

Really? This must be why the slow framerate on my mom's 12-year-old computer...

To be fair, a 12 year old computer isn't going to have much in the way of hardware encoding acceleration.

[tildearrow]

To be fair, a 12 year old computer isn't going to have much in the way of hardware encoding acceleration.

...with a card from 2016.

[starshipeleven]

...with a card from 2016.

sneaky

[Guest]

So... No H266 for at least a couple more years. Not surprised at all.

Well that got released just now. So yeah, it's going to take time. Won't be surprising if Apple and a few others introduce something next year.

[bug77]

Well that got released just now. So yeah, it's going to take time. Won't be surprising if Apple and a few others introduce something next year.

Right. But it's not about a few implementations. We need a lot of them, otherwise content will be scarce.
Oh well, nothing new, we've had new codecs before, we know it takes time. I was hoping we learned something and we'll move quicker this round. Apparently not.

[Guest]

Right. But it's not about a few implementations. We need a lot of them, otherwise content will be scarce.
Oh well, nothing new, we've had new codecs before, we know it takes time. I was hoping we learned something and we'll move quicker this round. Apparently not.

Well it's not just implementations, patents are a problem too. Plus it takes time to figure out how to encode good quality videos with new codecs.

Companies' main aim with using new codecs is to reduce their network and storage costs.

[bug77]

Well it's not just implementations, patents are a problem too.

If you mean it's a problem for content providers, yes, it is, but I don't care about that.
As an end-user, I'll consume whatever is available, I suspect I'm not alone.

Plus it takes time to figure out how to encode good quality videos with new codecs.

It's all in the spec, the only thing you need to do is see which settings work for you.

Companies' main aim with using new codecs is to reduce their network and storage costs.

That's why I don;t think anyone balks at the patent cost after all. Free is better, but proprietary you can have today is better than open you can't.