x265 - Work already started?

[Didée]

...subjective testing of HEVC compression capability ...
It shows impressive results ... In particular, > 50% bitrate reduction, 67% in class B (HDTV), 49% in class C (WVGA) => mission accomplished!
Until there's something to actually try, I'm suspicious of such numbers. Those are compairing reference encoders. I don't know about the quality of the HEVC reference encoder, but who knows, maybe it simply does a better job per se. The quality/efficiency of the JM reference encoder for AVC is known to be anything but impressive ...

Throwing some ballpark figures, x264 surely can cut bitrates by 50% compared to JM's AVC reference, in a breeze. :)

[kidjan]

Spec's released. Everybody start coding :P

From a certain standpoint, I feel like H.265 is a carrot on a stick that actually doesn't have a lot of value for anyone.

The biggest issue with it, obviously, is the complete and total lack of embedded/hardware support. There is no hardware support in any device currently sold that I'm aware of, which means there's something like a billion phones in circulation that are incompatible with the format. Which means encoding your content to this format is really pointless.

Second, the lack of encoder maturity is also a big turn off. If x264 demonstrated one thing very clearly, it's that the codec used isn't nearly as important as the thoughtfulness of the encoder implementation. So from my standpoint, it's just a bad attempt to squeeze some more royalties out of consumers by obsoleting a format that is frankly not obsolete.

[paradoxical]

The biggest issue with it, obviously, is the complete and total lack of embedded/hardware support. There is no hardware support in any device currently sold that I'm aware of, which means there's something like a billion phones in circulation that are incompatible with the format. Which means encoding your content to this format is really pointless.

Second, the lack of encoder maturity is also a big turn off. If x264 demonstrated one thing very clearly, it's that the codec used isn't nearly as important as the thoughtfulness of the encoder implementation. So from my standpoint, it's just a bad attempt to squeeze some more royalties out of consumers by obsoleting a format that is frankly not obsolete.

And yet, all of this was being said by people when H.264 was ratified. Hardware support took many years to become as ubiquitous as it is. x264 didn't become as mature as it is now until many years after the standard came out. By this logic, we should have just stuck to MPEG-1 video.

[JEEB]

Spec's released. Everybody start coding :P
...and where is it released, other than drafts? As far as I know, HEVC has not gone through the FDIS ballot yet, it passed the DIS ballot on January 8th. The current versions of the 10th draft are available here (http://phenix.int-evry.fr/jct/doc_end_user/current_document.php?id=7243), and that is what going to end up being pushed to the FDIS ballot after being finished up, as is noted by the name of the document: 'High Efficiency Video Coding (HEVC) text specification draft 10 (for FDIS & Consent)'.

If you mean the news about HEVC being officially named over at the ITU-T, yeah... That is exactly what it is. They decided that they will vote for it in the standardization FDIS ballot, and that it will get the ITU-T naming of H.265. Yet people really seem to take it out of context a lot :P .

Anyways, I'd guess that people would start moving a bit more after the final version of Draft 10 gets released, which would pretty much be what will get then released, as you can only do minor editing after the FDIS ballot is started. Also for really proper development one would need something to test his work against, say, a reference implementation.

Enter HM, currently at version 9.2 and under active development. It currently does not match up with the under-work 'HEVC Version 1' specification, and is expected to mostly match to it after version 10.0 gets released. But I suspect it will take some extra time for it to become generally 'fully' compatible with the text specification.

That said, Smarter's decoder, and another decoder/encoder project I know of (not the x265 that one Chinese dude started) have already started development, but esp. the latter project is having problems because HM is not matching the specification.

[Asmodian]

By this logic, we should have just stuck to MPEG-1 video.

Well maybe MPEG-2, at least I don't remember a mature MPEG-1 encoder that was better than early implementations of MPEG-2. ;)

I wouldn't call the release of new spec's obsoleting H.264, after all MPEG-2 isn't really obsolete, at least it is very widely supported and new content is still being released in it.

[paradoxical]

I wouldn't call the release of new spec's obsoleting H.264, after all MPEG-2 isn't really obsolete, at least it is very widely supported and new content is still being released in it.

Yep, still used overwhelmingly for US digital TV. And the quality suffers all the more for it, too.

[Asmodian]

Well maybe MPEG-2 should be obsolete but I wouldn't worry about devices dropping support for H.264 yet.

[hajj_3]

My prediction is that HEVC will give us more something like a 25% compression improvement, and probably mainly for high resolution content... Considering how much more cycles it eats, and that high resolution content takes ages to encode to begin with, its use is imo questionable, even more so considering that we don't know if 8k (and higher) will find widespread acceptance by the consumers or if it will go the way of high definition audio for music (SACD, Audio DVD...). As storage and bandwidth is growing I don't see much need for better compression than we have now with x264, at least not at the costs of the need for new encoding/decoder hardware and time (CPU cycles). Just my 2 cents :) *Back to IDLE mode*

x264 is already beaten with the hevc reference encoder, the reference encoder of AVC was poor so imagine how good an x265 encoder will be in a few years time.

Also most broadcasters don't use x264 they use commercial versions so hevc will be significantly better than the h264 that they currently use. BSkyB have been involved in HEVC so its pretty clear that sky in multiple uk countries will switch to hevc at some point. As for requiring new hardware... some nvidia/amd gpu's might be able to hardware decode using the programmable parts of the gpu. h264 was slow on cpu's for a long time so i think hevc will be a pretty big success.

sacd references are pointless as it is difficult to hear the differences whereas halving the bitrate is a dramatic change.

[LoRd_MuldeR]

x264 is already beaten with the hevc reference encoder, the reference encoder of AVC was poor so imagine how good an x265 encoder will be in a few years time.

In what way is it beaten? AFAIK the HEVC reference encoder runs extremely slow, as is expected for a reference encoder. So even if it could beat x264 quality-wise (at the same bitrate!), we cannot use it practically. So unless we get something that runs at reasonable speed (i.e. not significant slower than x264) it would be kind of a "theoretical" success with no practical relevance. Furtheron PSNR or SSIM numbers are one thing, but I'm yet to see an in-depth visual comparison of the HEVC reference encoder and x264.

Another point to consider: The better the reference encoder performs, the more difficult it will be for "alternative" implementations to actually improve over the reference encoder...

[Atak_Snajpera]

i hope that hevc won't be another 'jpeg2000' , 'jpeg-xr' , 'webp'. All those formats offer higher compression than jpeg but our digital cameras still use older technology.

25 - 33 % bitrate savings may not be enough for many companies to convince them to make additional investments in new hardware. especially when they just upgraded from mpeg2 to h.264.

[LoRd_MuldeR]

With the much higher amount of data for "4K" or even "8K" content, those 33% may be more important. And I guess it will require "hardware upgrades" anyway. Not that delivering "4K" content to end-users makes much of a sense. But after "3D" has become more or less standard and the big hype is fading slowly, it seems "4K" will be the next hype...

[Atak_Snajpera]

I can't wait to see 576i upscaled to 4K ;) Even now I'm very disappointed that so many shows are still being recorded in SD. So what that your TV reports video stream as 1080i if this is just upscaled version. This is very common in Poland. For me 4K is just another hype like 3D.

[edison]

With the much higher amount of data for "4K" or even "8K" content, those 33% may be more important. And I guess it will require "hardware upgrades" anyway. Not that delivering "4K" content to end-users makes much of a sense. But after "3D" has become more or less standard and the big hype is fading slowly, it seems "4K" will be the next hype...

I think there will be an another new encode tech for 8k content .

[benwaggoner]

25 - 33 % bitrate savings may not be enough for many companies to convince them to make additional investments in new hardware. especially when they just upgraded from mpeg2 to h.264.
Yeah, I'd say it takes at least a 50% improvement to drive a big industry-wide shift.

MPEG-2 was easily 50% better than MPEG-1 for interlaced content (which the bulk at the time).

But MPEG-4 part 2 was perhaps only a 30% improvement from MPEG-2, and didn't get broad industry support.

H.264 offered >>50% improvement compared to MPEG-2. But even there we see industries like cable still using mostly MPEG-2 due to all the old STBs out there.

[JEEB]

While I do agree that traditional broadcasting most probably is not going to be switching over to HEVC too quickly*, I must say that it doesn't exactly look like MPEG-4 Visual to me.

As far as I can remember, first of all MPEG-4 Visual never got used by ITU-T (thus not having a recommendation code on that side of things). Adding to that, it came relatively soon after MPEG-2 Video, which was IIRC '95 (MPEG-4 Visual came in '99), and the business was just in the middle of moving towards the first one with DVD video, first-gen digital television standards and so forth. Then, looking at the whole MPEG-4 Visual specification, it just feels like a mess; a lot of semi-cool features stuffed into a format while not really gaining that much from them than from some more "traditional" takes at compression. Not to mention that while H.263 did seemingly have in-loop deblocking, this was not present in MPEG-4 Visual for whatever reason. One of the relatively good WTF-listings regarding MPEG-4 Visual is here (http://guru.multimedia.cx/15-reasons-why-mpeg4-sucks/).

Looking at it after-the-fact, in a sense MPEG-4 AVC was really a kind of "Oh, so you just want the usual kind of thing after all?" format, which did actually work well as we have all seen. And HEVC takes it even further, with features that generally seem to be found to be good ideas by other formats that have come up lately, without going all out on "goofy" things like MPEG-4 Visual did.

What I am trying to say here is that while HEVC probably is not going to blow anyone's mind any time soon, comparing it to MPEG-4 Visual is kind of... not right in my opinion, as it already has gone further than MPEG-4 Visual ever did in the specification space regarding adoption.

* No-one wants to have customers crying over why they have to buy new TV sets or receiver boxes just to watch TV, only ~10 years after the last digital'ization was done -- or even less depending on the country and whether or not there was a switch towards AVC + 1080i/p lately.

[benwaggoner]

What I am trying to say here is that while HEVC probably is not going to blow anyone's mind any time soon, comparing it to MPEG-4 Visual is kind of... not right in my opinion, as it already has gone further than MPEG-4 Visual ever did in the specification space regarding adoption.
FWIW, I agree with everything you say there. Every indication at this point is that HEVC is going to be a big enough improvement for it to see broad industry adoption.

Speaking for myself, I expect that most or all practical 4K home video will be HEVC only. It'd be crazy to do anything else given HEVC's advantages at higher frame sizes, and new decoders are going to be needed anyway.

[xooyoozoo]

There was a recent paper published (https://ece.uwaterloo.ca/~z70wang/publications/vpqm13.pdf) detailing HEVC's subjective improvements over H264. The paper is more focused on objective metrics (psnr, etc) and spatial vs temporal distortions rather than the encoders themselves, but it still gives good data relevant to us. It also doesn't give any numbers the JCT docs haven't, but the fact that it's co-authored by Zhou Wang (SSIM's original creator) makes it inherently more quotable. ;)

Long story short is that 'state-of-the-art' objective metrics vastly underestimate HEVC, and with 1080p material (class B), HEVC can produce the same subjective quality for 1/3 of H264's filesize:

http://i.imgur.com/PhBWZKZ.png

There was also a more cursory subjective test (http://infoscience.epfl.ch/record/180494/files/hanhart_SPIE2012_1.pdf) last year on 4K clips. On some of clips, HEVC managed the same quality at 25% of H264's bitrate!

'Real' encoders would obviously have numerous improvements* over the JM used here, and one cannot assume that porting these improvements over to the HM would yield similarly proportionate gains. However, the massive gains shown for high resolution content sure does leave a large buffer zone for all the confounding variables to work things out. :D

*which I think are almost always overstated in these discussions. Looking only at quality per bit, reference encoders suck because they don't come with a steering wheel, not because something's wrong with the engine.

Edit: Forgot to link this fairly important bit from the first paper. Analyzing still-shots of an HEVC clip will also underestimate its overall performance:

http://i.imgur.com/46BlCil.png

[Warperus]

Is there any HEVC recommendation available to public (without subscriptions/payments)?

[drmpeg]

Is there any HEVC recommendation available to public (without subscriptions/payments)?

Here's the latest edits.

http://phenix.int-evry.fr/jct/doc_end_user/current_document.php?id=7243

Ron

[Warperus]

Thank you, drmpeg.

[JEEB]

By the way: why are so many users here still referring to "HEVC"? It's officialy called "H.265" now, isn't it?
ISO/IEC name is still (MPEG-H) HEVC, H.265 is the ITU-T's name for it, only given to it lately (before that it was marked as ITU-T H.HEVC there).

Just like (MPEG-4) AVC is the ISO/IEC name for what is called H.264 over at the ITU-T.

[Biggiesized]

H.HEVC might have been a working name like H.26L.

[JEEB]

H.HEVC might have been a working name like H.26L.
It was indeed. Since numbers are usually only given relatively close to finalization of the specification (although yes, it still isn't finished).

In any case, my point was that both HEVC and H.265 are valid names for the format :) (I generally will be calling it HEVC because that's how I've been calling it for the past XY months, and the fact that it gained another nickname isn't going to change much of that habit)

[pandy]

Can't wait for x265 then :).

https://code.google.com/p/x265/


@ x264 developers:

Can you please stop working on x264 and put all your workforce into x265 (or whatever you wanna call it) from now on :cool::D;)?

What with interlace? And how fast your CPU is clocked? 12GHz?

[aegisofrime]

https://code.google.com/p/x265/



What with interlace? And how fast your CPU is clocked? 12GHz?

AFAIK that isn't "real" x265 though. The developer isn't part of the x264 team like Dark Shikari, akupenguin et al. I wonder what will DS et al call their implementation of HEVC now?

[easyfab]

AFAIK that isn't "real" x265 though. The developer isn't part of the x264 team like Dark Shikari, akupenguin et al. I wonder what will DS et al call their implementation of HEVC now?

Are you really sure that x264 team will do a HEVC encoder ?

[aegisofrime]

Are you really sure that x264 team will do a HEVC encoder ?

Nope I'm not sure at all. AFAIK DS has not commented on the issue. I'm just saying that the x265 linked is not done by the x264 developers.

[Sagittaire]

anyway the work for x264 is incredible ... perhaps time for other dev to work to x265.

[LigH]

Hype victim.

There are very few people willing to spend a week on a movie just to be a "pioneer" using a bleeding edge technology, not understanding that most consumers and casual users will demand a conversion being faster than playing time.

Give the developers the time they need to really understand the concepts, even before they start implementing half-baked code. A good software project is started in theory, and it is "almost done" (regarding its definition) before the first line of code starts. The less you know where you go, the more often you have to go back because you took a wrong turn.

[Audionut]

Not to mention, afaik, it was over 12months after version 1 of H.264 before the first commit of x264. And x264 was not fast initially.

Have we even got a final draft of HEVC yet?
I guess people these days are less patient.

[LigH]

They never had the joy of manually tweaking DivX ;) 3.11α SBC parameters in Nandub until there were no "shit frames" left...

[pandy]

What do you mean with "What with interlace?"?




:rolleyes:


H.265 is designed for progressive video - mostly those with higher resolution in mind (4k and 8k)


So you're saying a 12 GHz CPU is needed to decode H.265 :rolleyes:?


Today probably yes especially for 4k content (8 k for sure) - count H.265 as 4 - 16 times more demanding than H.264 for CPU speed.


Also: the sooner the adoption of H.265 starts, the more likely it would be that there would be dedicated hardware to decode H.265. So a CPU might not even be necessary if NVIDIA or AMD for example would develop some VPUs capable of decoding H.265 ;).

Seriously, just look at those numbers posted by "xooyoozoo":




They sound amazing IMHO. So why not screw H.264 from now on and go for H.265 instead :p?

But why - is see no point to go for H.265 for HD - it have sense for Ultra HD however until moderate priced displays for 4k will be not present then i don't see any gain for H.265 over H.264.

[aegisofrime]

So H.265 does not support interlacing?

That would be awesome :).

Interlacing is a very evil thing IMHO and should have been abandoned a long time ago.

So, if H.265 drops interlacing support, then this actually is very much appreciated :).


They can drop it as long as we get high frame rates.

Recently there's a disturbing trend of some DVDs coming out with 30p content. In this case, I actually prefer 30i as I can deinterlace it with QTGMC to get nice 60p content. With 30p, the best I can do is to double the framerate with Interframe, and the results are not as good as deinterlacing 30i with QTGMC.

[sneaker_ger]

H.265 still supports flagging content as interlaced - it just doesn't have any interlaced tools like PAFF or MBAFF anymore AFAIK.

[Guest]

Guys, debating interlacing versus progressive is off topic here. Further such posts will be deleted.

[IanB]

Split the debate crap off into it's own thread, it's spoiling an interesting thread.

[kidjan]

And yet, all of this was being said by people when H.264 was ratified. Hardware support took many years to become as ubiquitous as it is. x264 didn't become as mature as it is now until many years after the standard came out. By this logic, we should have just stuck to MPEG-1 video.

Except there wasn't an existing base of billions of phones capable of decoding MPEG-1, and MPEG-1 was obviously a pretty basic standard comparatively speaking. Right now there are literally billions of embedded devices with hardware capability for H.264; it's a completely different situation if you take magnitude and proliferation into account, not to mention the relative maturity of H.264 when compared to something like MPEG-1 (please, no comparison)

[kidjan]

My prediction is that HEVC will give us more something like a 25% compression improvement, and probably mainly for high resolution content... Considering how much more cycles it eats, and that high resolution content takes ages to encode to begin with, its use is imo questionable, even more so considering that we don't know if 8k (and higher) will find widespread acceptance by the consumers or if it will go the way of high definition audio for music (SACD, Audio DVD...). As storage and bandwidth is growing I don't see much need for better compression than we have now with x264, at least not at the costs of the need for new encoding/decoder hardware and time (CPU cycles). Just my 2 cents :) *Back to IDLE mode*

...and that isn't even beginning to broach the whole patent subject. HEVC basically resets the clock on royalties.

[paradoxical]

Except there wasn't an existing base of billions of phones capable of decoding MPEG-1, and MPEG-1 was obviously a pretty basic standard comparatively speaking. Right now there are literally billions of embedded devices with hardware capability for H.264; it's a completely different situation if you take magnitude and proliferation into account, not to mention the relative maturity of H.264 when compared to something like MPEG-1 (please, no comparison)

There were more than a billion VCRs sold when DVD came out. There were more than a billion DVD players sold when BluRay came out. Did you fight against those things coming out as well? I really don't see how it's a different scenario other than you've just made an arbitrary line. H.264 deserves no special place of being considered irreplaceable versus any other video codec or format.

[me7]

When you compare the situation with SACD and DVD-A, remember that the CD has practically surpassed human hearing and all of these formats don't offer any practical advantage.
Right now, people love streaming high resolution videos over the internet. The quality is usually so bad that I'd roughly compare it to audio cassettes. A new video codec that improves the bad quality of youtube videos could have a similar impact as the CD had. People are ready for it, even my non-tech-savvy friends stream 1080p videos from youtube every day.

[xooyoozoo]

I really don't see how it's a different scenario other than you've just made an arbitrary line. H.264 deserves no special place of being considered irreplaceable versus any other video codec or format.

I also think it's a "different" scenario, but in an entirely opposite way.

Back then, my laptop couldn't handle H.264 properly. Do I buy a new laptop? No, I do so a few years later because it's a fully-functional laptop. Nowadays, my phone won't handle H.265. Do I buy a new phone? Well, yea duh, give me 8 months for my contract to end.

That's the what the appliancization of hardware has led to, and with some of the biggest (http://pro.gigaom.com/blog/intels-over-the-top-compression/) mobile (http://reviews.cnet.com/8301-13970_7-57387626-78/qualcomm-shows-horsepower-of-next-gen-h.265-video/) players (http://www.tweaktown.com/news/27733/broadcom_s_new_chip_supports_h_265_ultrahd/index.html) brawling for market share and carrier** support using whatever technological leverage they can, I doubt it'd even be possible to get any decent mobile gadget next year without H.265.

**at least in the US, this will singlehandedly ensure quick and peaceful media progressions for the next cycle or two. They love, NEED, the efficiency. "Oh you don't have H.265? Why don't we put your gadget back in the corner over here."

[LigH]

I wish it was just as simple with Ogg Vorbis. Sometimes I wonder if a competitor pays for not mentioning it in the list of supported formats (I know several devices which support Ogg Vorbis audio, but their manufacturers don't advertize that).

Marketing will have an important impact. The MPEG-LA has the power. DivX had a little less. Xiph hardly any. The popularity does not reflect the results of technical comparisons regarding quality and efficiency...

[kadajawi]

Not sure if that was posted here before, but some researchers (?) claim that h265 decoding can be done on an iPad for example.

http://www.v-net.tv/hevc-capable-devices-hit-1b-in-2012-ahead-of-standard/

Only needs a software update.

I think once decent encoders become available, and if device support really comes that fast (though the question is if the encoding settings need to be reduced in order to make it playable...), h265 will be used. It is useful everywhere. My connection is not even good enough for 480p at YouTube (mostly I use 240...), so if they can pack in 480p in a 240p sized stream, that would be awesome. The 720p/1080p people stream on demand is not very good in quality, if the same bandwidth can give better quality, I think it'd be quite popular.

[pieter3d]

At the JCT-VC meetings there was a demo of 1080p HEVC playback on an iPhone5, pure software decode.

[vivan]

It is useful everywhere. My connection is not even good enough for 480p at YouTube (mostly I use 240...), so if they can pack in 480p in a 240p sized stream, that would be awesome. The 720p/1080p people stream on demand is not very good in quality, if the same bandwidth can give better quality, I think it'd be quite popular.It's just because of youtube crappy encoder, not because of H.264. H.265 wouldn't really change anything there - if they wanted, they could improve quality a lot even with H.264.

[foxyshadis]

Not sure if that was posted here before, but some researchers (?) claim that h265 decoding can be done on an iPad for example.

http://www.v-net.tv/hevc-capable-devices-hit-1b-in-2012-ahead-of-standard/

Only needs a software update.

HEVC will never catch on as a software-only update on current devices. It would just eat battery too quickly to be useful, although the newest phones and pads would be able to use multicore decoding and get some improvement, and one or two might figure out how to leverage the GPU for some steps. It will never reach ubiquity until it's baked into hardware, just like AVC, and just like MP3 and MP4 audio before it. At best, a small portion of people will use it for personal collections and a few companies will integrated it into streaming apps, but it will remain niche until battery consumption issues are solved.

It's too bad, because I'd love it if it rolled out everywhere today, particularly the still image part.

[sneaker_ger]

Fraunhofer HHI has presented a software HEVC decoder supporting Main and Main 10 profile on the CeBIT. Allegedly "a single core" is sufficient for 1080p50, "four cores" for 2160p50 (4k). (Their test sequences were really low bitrate, so take it with a grain of salt...)

http://www.hhi.fraunhofer.de/en/media/press/cebit-2013.html?NL=0
http://www.heise.de/newsticker/meldung/Videostandard-H-265-HEVC-in-Aktion-1817722.html (German, English auto translation (http://translate.google.de/translate?hl=de&sl=de&tl=en&u=http%3A%2F%2Fwww.heise.de%2Fnewsticker%2Fmeldung%2FVideostandard-H-265-HEVC-in-Aktion-1817722.html))

[IgorC]

The situation around HEVC is different from H.264's 10 years ago.

Chipmakers were already prepraring HEVC acceleation.
Hardware video decoder gets only small area of chip and today even mobile GPU are much more powerful than that.
Considering that HEVC is 3-4 times more complex and that each new node (32 nm, 22nm... etc) gets 2x more power efficient, HEVC hardware decoding on 20-22 nm chip will consume the same amount of power as H.264 on 40-45 nm chip.
Even 28 nm chips would have a long bttery life.

As an example, the last Atom CPUs (32 nm) have no issues with harware decoding of Blu-Ray (H.264 High Profile). I have tried to play two Blu-Ray streams at the same time (or any other x264 encodes at high bitrate) on Atom N2600 (CPU+GPU TDP 3.5 W), zero dropped frames.

[Sagittaire]

The situation around HEVC is different from H.264's 10 years ago.
As an example, the last Atom CPUs (32 nm) have no issues with harware decoding of Blu-Ray (H.264 High Profile). I have tried to play two Blu-Ray streams at the same time (or any other x264 encodes at high bitrate) on Atom N2600 (CPU+GPU TDP 3.5 W), zero dropped frames.

Well no problem with ARM too but it's hardware decoding (GPU). Atom (CPU) can't decode two 1080p BD H264 stream with softwate decoding.

[Nil Einne]

Except there wasn't an existing base of billions of phones capable of decoding MPEG-1, and MPEG-1 was obviously a pretty basic standard comparatively speaking. Right now there are literally billions of embedded devices with hardware capability for H.264; it's a completely different situation if you take magnitude and proliferation into account, not to mention the relative maturity of H.264 when compared to something like MPEG-1 (please, no comparison)

But the thing with smartphones and perhaps tablets to a slightly lesser extent is they tend to have a short lifespan. Even less so then computers I would say. Most people probably keep their phones for at most 2-3 years, even if they're cheap ones and/or unsubsidised. Amongst other things, by that time the battery starts to die out and while many phones do have replaceable batteries, many people don't bother. So if it takes 2-3 years for most phones to have H265 support, we get most phones supporting in 4-6 years which is a fairly long time, but not that long.

BTW, for the person who mentioned royalties, I would say the MPEG-LA didn't get to where they are today by being stupid. Presuming there must be one of VP9 or Daala or something else (Intel/Real NGV perhaps?) which will be sufficiently improved over H264, even if not able to compete with H265, they have to be careful not to price themselves out of the market. As others have said, there are plenty of areas, particularly streaming videos and legal downloads where plenty of people would love to have smaller files or higher quality even for 1080p. Bandwidth is improving, but not that fast.