Qualcomm is Adding AV1 Support, Which Could Be Huge For Online Video (protocol.com) 24
Uptake of the open-video codec AV1 has been slow, with major video providers waiting for broader device support. Things could change over the coming months, as both consumer electronics companies and chipset providers are poised to introduce new hardware with native AV1 decoding capabilities. From a report: Chief among them is Qualcomm, which is planning to add support for AV1 to its upcoming flagship Snapdragon mobile processor, Protocol has learned from a source who has seen spec sheets for the chip. Internally known as SM8550, the chip is expected to be introduced at the end of this year at the earliest, which means we shouldn't expect any phones powered by it until 2023. The chip's Adreno video-processing unit will support native AV1 decode, something that none of Qualcomm's previous chips have offered. That's barring any major changes, with our source cautioning that things could shift before the chip actually enters production.
Flagship only? (Score:3)
To my understanding flagship ARM CPUs could already decode AV1 in software in real time as they had enough computational capability to do so. It was compute/energy intensive, but it was doable for reasonable resolutions and framerates.
It's the low and mid end models that really NEED the hardware acceleration. And according to the story, they're not getting it? In the world where overwhelming majority of devices are sold with low and mid end CPUs?
It's nice to have AV1 hardware decode in 2023 flagship models, sure. But that's not much of a breakthrough for the codec. Basically people who already have devices that could decode AV1 will be able to decode it with better energy efficiency. And everyone else is not getting it.
Re: (Score:2)
Is it more power efficient to have dedicated hardware for decoding?
Re: (Score:2)
I think this is commonly understood to be true in practically all applications, from BTC mining to digital media playback and capture. Playback suffers a hit when the codec has been tweaked or uses features not found in the hardware decoder. Fine example would be the way some anime media ripping groups some years back (decade?) used the 10-bit version of h264 to produce video that approached the quality of the newer h265.
Entirely anecdotal: but I think AV1 even in its initial beta state already produces bet
Re: (Score:2)
Re: (Score:2)
Sure. But if it takes many years for adoption to be anywhere near "universal", who's going to use AV1 to encode video instead of alternatives for anything that has a chance to need phone playback?
That's really the problem with video codecs. As long as they're not near universally adopted, if you encode in them, it's basically something you do in addition to encoding in universally accepted one. It's why h.264 remains industry standard in many ways in spite of being launched back in what, 2003? But ASICs for
Re: (Score:2)
Re: (Score:2)
MP3 is at least a good quality codec. But jesus fucking christ at SBC and bluetooth.
Re: (Score:2)
Re: (Score:2)
But it's not a meaningful improvement in most consumer usage scenarios. Few people care if a single piece of music at maximum quality they can realistically hear is 5 megabytes or 3.
Bluetooth audio being ubiquitous in modern mobile audio is typically encoded in SBC. Which is way, WAY worse than mp3. And that what most bluetooth devices use. Today. And that's what most bluetooth devices will default to if even one device in the chain just doesn't quite play nice, even if they support something significantly
Re: (Score:2)
And Vorbis is better than AAC, and Opus is better than Vorbis. Technology advanced, but just because there are better codecs available doesn't mean people will use them.
Re: (Score:2)
Lots of people use those codecs. Millions of Spotify users listen to music encoded with Vorbis although most neither know nor care. In the last few years, pretty much every YouTube video where I've checked the video codec on using "stats for nerds" has been using Opus as the audio codec and I presume that's the case for most people since audio decoding in software isn't going to strain any modern CPU.
Re: (Score:2)
And Discord uses Opus too. The codecs do have adoption, but they lack public awareness. Ask any random person about MP3 and they will know it. Most people looking to pirate some music casually will just google 'youtube save mp3.'
Re: Flagship only? (Score:1)
Re: (Score:2)
Re: (Score:2)
Last years Samsung S21 and this year's S22 sold in many non-US markets uses an Samsung Exynos CPU that already has hardware decoding. The Exynos 1280
CPU that will power this years Samsung midrange phones is also rumoured to have AV1 hardware decoding. Cheaper Chinese phones from brands like Oppo, Xiaomi and Realmi are using Mediatek CPUs which already have hardware decoding.
You are right that the hardware acceleration is needed in the midrange phones as well, but you are wrong if you think that stre
Re: (Score:2)
My point is that things like streaming services will use whatever is best for their bottom line. One of the criteria there is "should we bother wasting server compute to encode this in another codec for every resolution and frame rate we offer, for some, or for any at all?" Encoding isn't free. Storage isn't free. But bandwidth and customer satisfaction is also not free. So codec choice is a careful balancing act of those things. Especially now in the world where most customers are used to video "just worki
Article from the future: (Score:2)
Old YouTube Videos Draining Phone Batteries
This is going to be the story when they finally drop hardware accelerated H.26x codecs from mobile processors. Of course, video sites will probably start transcoding old video uploads which will be ugly.
Re: (Score:2)
I highly doubt H.264 is going to be leaving mobile processors within 20 years. It became the de-facto video codec for everything, and not even its own successor can oust it.
If only (Score:3)
web sites could get their videos to play without having to figure out which of the 50 scripts on the page need turned on.
This is the 21st century. If you need more than two scripts to get a video to play, you're royally fucking up.
Re: (Score:2)
Those scripts aren't there to benefit you, but the corporate overlords.
Why would they want you to know which ones are necessary? That's like expecting them to tell you which domains need to be allowed to set cookies.
Thank you EFF, Google, On2, Xiph, etc (Score:5, Interesting)
For this even being the prevailing conversation in video codec support. The fact that the news is about support for an open and free codec is almost taken for granted whereas it wasn't that long ago there was dread about how the licensing around h.264 and the MPEG group would play out and whether it would put online video in an early grave of licensing fees. Not that Google was acting totally altruistically and had to be goaded a bit but their purchase and opening of VP8 and then VP9 and now the Alliance for Open Media and AV1 really changed the whole foundation of todays internet (and also helped Googles bottom line to be fair)
Also thanks to Matroska, Vorbis, Opus and as always FFMpeg and all the developer support behind them over the years really combining to put open source standards at the core of another section of the internet as a whole.
Re: (Score:2)
It's kind of a roundabout self-payout, but very win-win. Giving away the codec gets adoption into the hands of users faster and cheaper than licensing patents ever would. They save more on infrastructure costs for lower bandwidth and storage than they would have earned on royalties.