[ksec]

This is quite amazing.

On my Intel Mac it still requires 4 thread with close to 100% CPU usage to play a 4K AV1 using latest VLC. ( Which I assume comes with latest dav1d ) This is easily 20W usage.

But on M1 this takes less than 1W.

This basically eliminate any argument against AV1. Even with software decoding. You could watch AV1 4K video with passive cooling on a MacBook Air without it ever getting warm.

It may be a good question to ask, with these kind of processing power, what can we do to further push the next gen video codec.

[brokenmachine]

> It may be a good question to ask, with these kind of processing power, what can we do to further push the next gen video codec.

I love it.

I assume you saw no problem with quality of the 4k video on a laptop screen, but you automatically want it to do more.

Of course, I'm not against the idea of improvement, it's just funny how the goalposts continually move.

[ksec]

>Of course, I'm not against the idea of improvement, it's just funny how the goalposts continually move.

LOL Yes because we want to lower bitrate. Taking Youtube or Netflix aside, the vast majority of video on the internet haven't seen any improvement in quality. They moved to newer codec for bandwidth reduction at the same quality so they can serve more content. Until Bandwidth becomes cheap enough where like Audio we no longer want better quality at 64Kbps or lower bitrate, people instead are simply using higher bitrate. I mean 20+ years later, we still dont have what they promised was mp3 128Kbps quality at half the bitrate.

There is a question of how far we can push codec development within the video codec industry. Not because they have ran out of idea but the complexity of codec means to both encoding and decoding. That we may be edging towards a point where Software decoding is not a sustainable solution and does requires dedicated hardware, which in itself is a topic of discussion with different hardware manufacturers. And Video Encoding or Decoding Engine is already getting quite big in terms of Die area.

So what the M1 shown here is a door to a whole new world of possibilities. For example there are half dedicated AV1 hardware decoder and hybrid GPU decoder that uses more power than M1 in pure software mode.

And this is on a M1 with A14 Core, not A15, or the coming A16. We should be able to see another 30% reduction in energy usage within the next 2-3 years.

We could also question and ask, may be we could finally do Wavelet transform where previously was thought as too computational expensive.

We have a roadmap on 5G with 3GPP Rel 18, Massive MIMO is far from finished, we also have roadmap on Ethernet and Cost reduction on SerDes and Controller. Our Server is finally getting higher Core count with reduction in cost / performance. By 2030 vast majority of the world, even in developing nation could enjoy 1080P Video with very good quality at under 1Mbps.

And then the next frontier will be, how do we further improve Live Streaming Video quality with ultra low latency and bitrate. Actually that is what I am partly pushing for right now. That may be a task for another 10 years.

I love continually moving the Goalpost :)