50M : How much data do you use?

[martkt10]

Originally Posted by Kushan

Not when it's 4k!

Where do you find 4k porn ?

[Kushan]

Originally Posted by martkt10

Where do you find 4k porn ?

Around.

[Ignitionnet]

Porn always leads the way. Was delivering quality video online way ahead of the curve.

I was about to link a list of 4k porn sites however said page is, err, somewhat risque*.

*it has hardcore porn on it.

[qasdfdsaq]

Originally Posted by Ignitionnet

They used H.264. Codecs have come along a little bit since - see H.265 and VP9.

Yes, I'm well aware of H.265, and it's well into the realms of diminishing returns, it makes very little improvement in bandwidth usage, we're talking single digit to low double digit percentages. It doesn't introduce any new coding concepts and doesn't do very much that advanced implementations of h264 can't already do - in any case even the maximum hypothetical "marketing headline" improvements of 30% won't get close to reducing a 500Mbps h264 stream to something that can be strung over 100Mbps ethernet without degrading to Youtube quality - and face it, nobody's going to buy an 8K setup and settle with Youtube quality.

---------- Post added at 03:48 ---------- Previous post was at 03:46 ----------

Originally Posted by Ignitionnet

Porn always leads the way. Was delivering quality video online way ahead of the curve.

I was about to link a list of 4k porn sites however said page is, err, somewhat risque*.

*it has hardcore porn on it.

What's more impressive, is that while it's still hard to find even 1080p60 content in mainstream media, most of the 4K porn listed above is full 60p (to be fair, finding 1080p60 porn seems to be harder than 4K60).

That said, quality descriptors aside, quality is actually pretty poor, even at 25-30Mbps.

[Ignitionnet]

Originally Posted by qasdfdsaq

Yes, I'm well aware of H.265, and it's well into the realms of diminishing returns,

There appears to be nothing you aren't an expert on.

I can only go by the documents I've read on the respective standards which indicate considerable gains over H.264, so apologies if I am giving incorrect comments, they're based on incorrect information.

[Kushan]

Originally Posted by qasdfdsaq

Yes, I'm well aware of H.265, and it's well into the realms of diminishing returns, it makes very little improvement in bandwidth usage, we're talking single digit to low double digit percentages. It doesn't introduce any new coding concepts and doesn't do very much that advanced implementations of h264 can't already do - in any case even the maximum hypothetical "marketing headline" improvements of 30% won't get close to reducing a 500Mbps h264 stream to something that can be strung over 100Mbps ethernet without degrading to Youtube quality - and face it, nobody's going to buy an 8K setup and settle with Youtube quality..

Where are you getting your information about that? Everything I've ever seen states quite the opposite.

http://www.extremetech.com/computing...o-expectations

This gives a good basic idea of what sorts of benefits H.265 can offer compared to H.264. While it’s not hitting 50% bandwidth savings in most cases, it’s close — quantizer 24 is 57% the size, q=30 is 59%, and q=40 is just 47%. Granted, at a quantizer of 40, the final output is wretched — but it’s wretched at less than half the bandwidth.

[qasdfdsaq]

Originally Posted by Kushan

Where are you getting your information about that?

Mainly from the knowledge and experience I gained while I was coding optimizations for the x264 encoder.

Everything I've ever seen states quite the opposite.

http://www.extremetech.com/computing...o-expectations

Let's start with the fact that the first chart (bitrate) doesn't even go high enough to cover decent quality 1080p bitrates, it's also using a restricted h264 profile (unrestricted on the other hand opens up a whole host of quality and efficiency enhancing options which are disabled on the fixed profiles targeted at standalone media players with limited hardware capabilities). Granted, it's also using a restricted profile for h265, but it's also trying to quantify 'quality' using luminance-only PSNR estimates - which is widely known in the enthusiast coding community as ... well... estimates. It's attempting to approximate human perception of 'quality' based on computer determinations of bit-level differences.

Skipping onto the "file sizes" chart you'll note once again it doesn't get close to figures actually required for decent quality HD, a quantizer of 40 is worse than Youtube's 160p setting and a quantizer of 24 is still below the bare minimum for acceptable HD (a quantizer of 16-20 is considered the 'good' quality range, with closer to 16 being the equivalent of Bluray). Once again using a limited h264 profile which again contradicts my point earlier, that an unrestricted profile is required to enable all the quality options. Quantizer settings are also entirely arbitrary and cannot be used as a measure of quality, even within the same encoder let alone different codecs; changing the quantization matrix can make a Q40 encode look like a Q24 or vice versa, hence why "constant quantizer" settings are are never used as a "constant quality" regulator. Short version: frame quantizers as a 'quality indicator' are useless, they are simply the output of an arbitrary 'detail mask' that will be completely different for different codecs.

If you look at Ignition's link earlier summarizing the "new" capabilities in h265, you'll see the only real new feature comes from the increased macroblock size of 64x64 vs. 16x16 in h264, which on it's own barely gives a 11% increase in coding efficiency. Most of the remaining capabilities including advanced prediction, transforms, and CABAC, are all available in h264, and in-loop filtering is only useful for masking atrociously low bitrates, and should really be turned off at high quality settings anyway. Things like GOP sizes, B-frames, inter-I references etc. are used more intelligently by default in h265 but only by bringing the defaults up to levels that advanced encoders have been doing in h264 for nearly a decade.

H264 was already such an exceedingly efficient codec that it got close enough to the theoretically possible maximums that it simply isn't possible to improve much further. It's just again, as I mentioned, a lot of content is encoded for compatibility with the lowest-common-denominator thus disabling a vast number of features ultimately resulting in poor quality. Also, all encoding accelerators (e.g. GPU accelerated encoding, Intel Quick-Sync, etc.) are optimized for speed over quality.

---------- Post added at 16:54 ---------- Previous post was at 16:43 ----------

P.S. If you want a look at some "real life" comparisons rather than theoretical predictions, here's a good place to start:

http://forum.doom9.org/showthread.php?t=170986

[Kushan]

Some quite detailed stuff there. It's not something I know enough about to comment on properly, all I can say is that you're the first person I've seen to claim that it's only about 10% better, while every single other source says it's more like 50% depending on the content.

Also regarding the link, is that not comparing x264 and x265 rather than the codecs themselves?

[qasdfdsaq]

Well then we're getting onto the technicalities of what you call a 'codec', x264 and x265 are encoders, not full-codecs, true, but most formats use specialised and separate encoders vs. decoders. But most would consider x264 a codec and h264 the format that it encodes. If you are asking why we're comparing codec software instead of the format itself, it's because video quality is hugely influenced by the software and settings used to encode it, as much as it is the format itself - after all, Youtube and BluRay use the same h264 'format' but different codecs.

Regardless, those are the best known and most widely used open-source implementations of the format, and (x264 at least) used to be considered the gold standard for quality.

Finally, if you carefully read many of the links you say mention "50%" you'll notice a lot referencing that as a theoretically up to and/or only applicable to lower qualities. Which is true - but as you can also see from the same graphs, the higher the bitrate and the higher the resolution, the smaller the difference becomes, by the time you get to 8K which we were talking about it's in the low double-digits as I mentioned earlier.

[dragon]

Billing Period Peak usage Off-Peak usage Total usage
05 Aug - 04 Sep 464.25GB 89.75GB 554GB
05 Sep - 04 Oct 299.08GB 41.23GB 340.31GB
05 Oct - 04 Nov 468.11GB 86GB 554.11GB

According to plusnet although I'm not totally sure that's right, unless I really did do exactly the same usage for 2 months