Hi10P and 8-bit encodes

[Bob2004]

You said that or you worded your argument in such a convoluted manner that I thought you did.

You are seriously underestimating the power of modern GPUs - how old is the NV 9800 for example? - that is plenty of power for real time dithering (exactly like the cat picture on the wiki article). Again, video type does not effect the GPU's ability to apply shaders before output (even with DXVA as DXVA is just a chip on the GPU dedicated to h264).

Because that way - quality fetishists can use fancy filtering to achieve their boners and the rest can just watch movies in a standard video format?

Kinda contradicting though, you talking about the horrible performance of shaders on hardware while saying that you do not notice the massive performance differences between EVR|VMR and MadVR and Haali Renderer (Installed with Haali Splitter).

Aerah, you seem to be forgetting that 10-bit provides benefits that cannot be gained by using any amount of shaders, filters, or whatever, when playing the video. The main benefits of 10-bit are in a) preventing banding from being introduced during the reencoding process, and b) increasing precision, thus improving compression, and in turn leading to much more efficient use of bitrate - and thus higher quality, or lower filesizes.

Obviously neither of these things can be achieved by the video player (as you well know - it's been explained many times already in this thread), so I'm not really sure what point you're trying to make here.

Also, the argument about whether MadVR uses more cpu or gpu seems completely irrelevant. It is a slightly higher-quality renderer that also uses more resources. EVR is perfectly good, so you don't really lose out on much by not using MadVR. It's just an option that is available if you have the resources available. Since it is completely unneccesary if you want to play 10-bit, and the quality improvement is so small as to be almost unnoticeable, it seems pretty irrelevant to this discussion.

[RedSuisei]

You said that or you worded your argument in such a convoluted manner that I thought you did.

So you seem to have a comprehension problem, I take it.

You are seriously underestimating the power of modern GPUs - how old is the NV 9800 for example? - that is plenty of power for real time dithering (exactly like the cat picture on the wiki article). Again, video type does not effect the GPU's ability to apply shaders before output (even with DXVA as DXVA is just a chip on the GPU dedicated to h264).

If you have the cash to get a modern GPU, then you also have the cash to get a modern CPU capable of playing 10-bit in the first place.

Because that way - quality fetishists can use fancy filtering to achieve their boners and the rest can just watch movies in a standard video format?

Are you saying that with this, the rest can just watch with lower quality?  Because if you do, then the rest can also get a lower resolution 10-bit video to cope with their old hardware. Also, if you are an encoder, you'd know that any filter that actually gives good results are too heavy for real-time playback.
(Btw I don't know of any shaders that actually does deband; all I know to do real-time deband is via ffdshow. If you know of any of this debanding shaders for MPC-HC, please point a link to me as I would like to test it.)

Kinda contradicting though, you talking about the horrible performance of shaders on hardware while saying that you do not notice the massive performance differences between EVR|VMR and MadVR and Haali Renderer (Installed with Haali Splitter).

Whoever said there aren't any performance difference? There is, but it is on the GPU performance. CPU load? Using EVR I get 15-19% CPU usage, using madVR I get 16-19% CPU usage. I certainly wouldn't call that 1% difference a "massive performance difference." Even madVR's thread on doom9 explicitly states that all work are done via GPU. Stop misleading people with your uninformed statements.

[Aerah]

Dunno, about you but I never noticed gradient/banding problems with large quality 8-bit encodes. :/

On the positive side,
- CoreAVC 3.1 promises performance improvements for 9-bit and 10-bit media (sometime in early '12?)
- New MPC:HC builds promises better subtitle performance (sometimes in '12... they are slow)

I am currently trying to re-encode [jackoneill] The Third MKV with Handbrake.
First attempt gave out a 150MB file which turned out very very bad. (using VerySlow from http://bytebin.wordpress.com/2011/02/06/x264-presetstunes-and-handbrake/)
Now trying with Avg. Bitrate instead of constant quality.

@RedSuisei
"Are you saying that with this, the rest can just watch with lower quality?" also applies to the move to 10-bit. Before 1080p h264 FLAC fansubs - after 720p/480p 10-bit h264 fansubs - low-hardware crowd will have to watch in MUCH MUCH lower quality.

[RedSuisei]

@RedSuisei
"Are you saying that with this, the rest can just watch with lower quality?" also applies to the move to 10-bit. Before 1080p h264 FLAC fansubs - after 720p/480p 10-bit h264 fansubs - low-hardware crowd will have to watch in MUCH MUCH lower quality.

Much lower quality? Most of your "high quality" 1080p releases are from upscaled Blu-Rays anyway (and because of this not much new 1080p torrents are going to be accepted anyway except maybe in some special cases), so I don't think you'd lose that much. The only reason I brought this up is because your statements are hypocritical; from the beginning I don't care if people with lower-end hardwares will need to get lower-resolution releases (even if I myself were also in that crowd; I see the lack of hardware capability as my own shortcoming and not the releaser's responsibility to cater to that). However, you from beginning mentioned how 10-bit will force people to watch in lower quality, then you go ahead and said that the rest who lack the sufficient hardware to do real-time filtering can watch in lower quality. What a hypocrite.

Also, you don't notice banding on large quality 8-bit encodes, there are two possibilities: either your eyes aren't sharp enough to notice those (in which case, you shouldn't care about this "lower quality" you speak of), or the 8-bit had such a large bitrate to preserve the dither.

[Aerah]

@RedSuisei
"Are you saying that with this, the rest can just watch with lower quality?" also applies to the move to 10-bit. Before 1080p h264 FLAC fansubs - after 720p/480p 10-bit h264 fansubs - low-hardware crowd will have to watch in MUCH MUCH lower quality.

Much lower quality? Most of your "high quality" 1080p releases are from upscaled Blu-Rays anyway (and because of this not much new 1080p torrents are going to be accepted anyway except maybe in some special cases), so I don't think you'd lose that much. The only reason I brought this up is because your statements are hypocritical; from the beginning I don't care if people with lower-end hardwares will need to get lower-resolution releases (even if I myself were also in that crowd; I see the lack of hardware capability as my own shortcoming and not the releaser's responsibility to cater to that). However, you from beginning mentioned how 10-bit will force people to watch in lower quality, then you go ahead and said that the rest who lack the sufficient hardware to do real-time filtering can watch in lower quality. What a hypocrite.

Also, you don't notice banding on large quality 8-bit encodes, there are two possibilities: either your eyes aren't sharp enough to notice those (in which case, you shouldn't care about this "lower quality" you speak of), or the 8-bit had such a large bitrate to preserve the dither.

The "lower quality" for "the rest who lack the sufficient hardware" is the first-half of '11 8-bit releases by master-level fansub groups. The visual thing 10-bit crowd is complaining about is just un-smooth transitions - especially noticeable in gradients - I don't why this cannot be fixed by filtering.
There is no "lower quality" in this suggestion. There is only higher quality achieved without switching to a software decoding.

The hyporcrite is the one who promotes Hi10p which alienates low-end CPU power but cries foul when a suggestion is given which alienates low-end GPUs.

[OnDeed]

Btw as OnDeed mentioned, madVR doesn't take much more CPU power than any other renderer. At the very least, using an old 2.4GHz C2D shows almost no increase in CPU usage by using madVR or EVR. It takes much more GPU though.

I didn't say that... actually people do say it takes more cpu. And wait a moment, EVR has quite some cpu overhead, compared to overlay, so it's not a huge win to beat it But people seldom need to worry about madvr's cpu usage as said, since it needs a gaming-worthy gpu anyway (thus it probably won't find itself in a pc with slow cpu).

P.S.
Regarding re-encoding downloaded rips... that is imho silly, obnoxious and unreasonable. Just a PITA. If anyone really wants to recommend that to people, he or she better stop and go right for the "get you stuff elsewhere" line. Getting other encodes at other places is a practical choice, re-encoding isn't.

[parusit]

Regarding re-encoding downloaded rips... that is imho silly, obnoxious and unreasonable. Just a PITA. If anyone really wants to recommend that to people, he or she better stop and go right for the "get you stuff elsewhere" line. Getting other encodes at other places is a practical choice, re-encoding isn't.

Totally agree.


p.s. Maybe too much of self-centered thinking here(any side). To the point it looks like Discrimination.

[Aadieu]

Time out!

Will someone tell me already why it is that persons involved in this discussion are getting radically different results from Hi10p playback, with little to OVER9000 level banding, as seen in screenshots of the same scene from the same file?

Also a good time to figure out what part of that is attributable to decoders, what part - to settings, what part to GPU/processing capabilities, and what part to filters?

Cause while it's pretty easy to blame my screenshot on my "doing it wrong", is it also not easy to suspect the better results of "doing it TOO right" (aka introducing results of various technologies that vastly improve visuals and misrepresenting these results as those of the 8bit >> 10bit switch, when in fact they might have shit-all to do with colour bitrates or encodings)?

...ARE WE EVEN SURE THAT ANY OF THE DIFFERENCE "SHOWN" ANYWHERE AT ALL CAN BE ATTRIBUTED TO ANYTHING BUT GPU AND MADVR? Or any other differences due to owners of finetuned hi-end systems posting their top-of-the-line 10bit screens vs. basic users with average systems and no finetuning providing the 8bits for comparison?

Also, could somebody that got "pretty" results (screenshots) on a 10bit release do comparison shots with the same group's similar-filesize 8bit version of the same anime, while running the same player and the same presets?  (Yeah I know that guy from UTW did that, but he was selecting screens of his own encoding to illustrate his own point - not exactly a perfect unbiased situation, a clean experiment would have to be done by someone without a prior conclusion and without the encoder's knowledge; and, in any case, both of his encodes were badly bitrate-starved, yet both looked pretty good and not exactly radically different, either)

[dragon191]

I think this community has a lot of people who can't play high resolution 10bit files properly. So I hope that BakaBT will decide that there will remain an 8bit SD slot. I mean, it took ages before XviD was waved goodbye, so I don't see a reason to not do the same here. :>

[cyberbeing]

Time out!

Will someone tell me already why it is that persons involved in this discussion are getting radically different results from Hi10p playback, with little to OVER9000 level banding, as seen in screenshots of the same scene from the same file?

Also a good time to figure out what part of that is attributable to decoders, what part - to settings, what part to GPU/processing capabilities, and what part to filters?

Cause while it's pretty easy to blame my screenshot on my "doing it wrong", is it also not easy to suspect the better results of "doing it TOO right" (aka introducing results of various technologies that vastly improve visuals and misrepresenting these results as those of the 8bit >> 10bit switch, when in fact they might have shit-all to do with colour bitrates or encodings)?

If TV->PC levels conversion and/or YUV->RGB conversion is done poorly, you will create banding which doesn't exist in the source.

You or something you are using is doing something is doing something that hurts quality. Since this actually shows up in your screenshots, I'd suspect PotPlayer, or how you have PotPlayer set up may be to blame. Could you retest using MPC-HC + FFDShow with a standard CCCP install? I'd also like to see you take a screenshot with madVR, which will avoid anything your GPU Driver may or may not be doing which hurts quality.

The other factor is your monitor. Especially in the past, there were many LCD displays which would show banding when displaying a smooth 8-bit gradient. Do you see banding in any of the following smooth gradients?

RGB Spectrum
RGB Linear Gradient
Gray Linear Gradient
Red Linear Gradient
Green Linear Gradient
Blue Linear Gradient
Gray Horizontal Gradient
Gray Vertical Gradient

...ARE WE EVEN SURE THAT ANY OF THE DIFFERENCE "SHOWN" ANYWHERE AT ALL CAN BE ATTRIBUTED TO ANYTHING BUT GPU AND MADVR? Or any other differences due to owners of finetuned hi-end systems posting their top-of-the-line 10bit screens vs. basic users with average systems and no finetuning providing the 8bits for comparison?

Also, could somebody that got "pretty" results (screenshots) on a 10bit release do comparison shots with the same group's similar-filesize 8bit version of the same anime, while running the same player and the same presets?  (Yeah I know that guy from UTW did that, but he was selecting screens of his own encoding to illustrate his own point - not exactly a perfect unbiased situation, a clean experiment would have to be done by someone without a prior conclusion and without the encoder's knowledge; and, in any case, both of his encodes were badly bitrate-starved, yet both looked pretty good and not exactly radically different, either)

EveTaku Ben-To 01v2 8bit (359.3 MB) w/ EVR-CP
EveTaku Ben-To 01v2 10bit (304.1 MB) w/ EVR-CP

EveTaku Ben-To 01v2 8bit (359.3 MB) w/ madVR
EveTaku Ben-To 01v2 10bit (304.1 MB) w/ madVR

None of these look as bad as your screenshot and I'm not doing anything special. Decoded Video -> Renderer, that's it.

For reasons stated above, madVR will have less banding compared to EVR-CP, but there is still a noticeable difference between 8bit and 10-bit either way.

As far as discrete GPUs which work with madVR, you are probably looking at ~$55-75 for something decent. madVR really doesn't need GPU which is particularly powerful, just something above most entry-level discrete, mobile, and integrated GPUs. As stated by RedSuisei, there is a slightly higher CPU load using madVR (+ 1% to 5% depending on CPU speed), but unless you are running your CPU constantly at 90%+ it shouldn't be noticeable.

The computer I'm testing is rather low-end by today's standards, and it handles 10bit 1080p under ~35Mbps with FFDShow, LAV Video, or CoreAVC + madVR + xy-VSFilter just fine:

AMD X2 4800+ (939) @2.4Ghz
2GB DDR400
NVIDIA 7800GTX 512
WinXP SP3 x86

[Aadieu]

If TV->PC levels conversion and/or YUV->RGB conversion is done poorly, you will create banding which doesn't exist in the source.

You or something you are using is doing something is doing something that hurts quality. Since this actually shows up in your screenshots, I'd suspect PotPlayer, or how you have PotPlayer set up may be to blame. Could you retest using MPC-HC + FFDShow with a standard CCCP install? I'd also like to see you take a screenshot with madVR, which will avoid anything your GPU Driver may or may not be doing which hurts quality.

The other factor is your monitor. Especially in the past, there were many LCD displays which would show banding when displaying a smooth 8-bit gradient. Do you see banding in any of the following smooth gradients?

RGB Spectrum
RGB Linear Gradient
Gray Linear Gradient
Red Linear Gradient
Green Linear Gradient
Blue Linear Gradient
Gray Horizontal Gradient
Gray Vertical Gradient

OK... tell me you're pulling my leg here?!? These things don't look smooth at all. o_O

BUT... I'm 250% sure I've seen plenty of gradients without banding on this computer and this monitor. Hell, no ass in porn ever looked banded... And even the less-banded screenshots of EveTaku Bento look - well - less banded... and no photograph has ever shown banding or gradient problems... AND I've watched plenty of banding-free animes, encoded at anything from ~700 to 6000+kbit/s, resolutions from sub-SD to FullHD, with none of this crap happening.

Plus, hell, even the background of the "topic summary" stripe here has a gradient - and thas all nice and smooth? And windows aero looks plenty banding-free??

...what gives?!?

[Aadieu]

EveTaku Ben-To 01v2 8bit (359.3 MB) w/ EVR-CP
EveTaku Ben-To 01v2 10bit (304.1 MB) w/ EVR-CP

EveTaku Ben-To 01v2 8bit (359.3 MB) w/ madVR
EveTaku Ben-To 01v2 10bit (304.1 MB) w/ madVR

None of these look as bad as your screenshot and I'm not doing anything special. Decoded Video -> Renderer, that's it.

ACTUALLY... madVR looks better (with madVR 8bit best, slightly better than madVR 10bit... but EVR 8bit and 10bit tied for worst, with 10bit perhaps even worse???) ... and both the EVRs, for whatever reason, look just as banded as my own screenshot *on this screen, as connected to this computer*. Ok,what gives?!?!

[DmonHiro]

ACTUALLY... madVR looks better (with madVR 8bit best, slightly better than madVR 10bit... but EVR 8bit and 10bit tied for worst, with 10bit perhaps even worse???) ... and both the EVRs, for whatever reason, look just as banded as my own screenshot *on this screen, as connected to this computer*. Ok,what gives?!?!

Well, that settles it then. It's either your monitor, or your eyesight.

[Aadieu]

Ok, seeing as how the rest of the world don't look banded and there ain't any such eyesight disorder anyway - what the hell is wrong with my TV??? And why does it only show some banding, but not across the board? And why almost none in madVR screencaps?!?!

[Bob2004]

The reason those gradients that cyberbeing look banded is because most of them aren't smooth gradients - ie. the colour changes in bands anyway. That's what they're meant to look like (a couple of them aren't - such as the RGB Spectrum and grey vertical gradient).

And if the problem was your monitor, then obviously we would not have been able to see it in your screenshot, would we? The screenshots would look identical to us if your monitor was the only problem.

Your software is most likely what's causing your issues (although your monitor and/or eyesight might also be problematic). My advice is to just completely uninstall PotPlayer and whatever else you're using, install CCCP, and see how it looks at default settings in the included player, without changing anything. There should be a noticeable improvement.

Anyway, this thread has gone massively off-topic now. The discussion has pretty much run its course, so maybe someone should lock it?

[OnDeed]

I think this community has a lot of people who can't play high resolution 10bit files properly. So I hope that BakaBT will decide that there will remain an 8bit SD slot. I mean, it took ages before XviD was waved goodbye, so I don't see a reason to not do the same here. :>

And the differences between ASP encoders (xvid) and x264 is very large, both in compression and in the visual quality that is achievable. Which isn't the case here, 10-bit might help a lot when you have gradients in a cg-rendered anime, but for hand-drawn stuff, there really won't be banding unless the encoder screws up. In that case, 10-bit will likely only give you the small general compression benefit (0-5% bitrate reduction?).

In short, xvid was a morally dead format compared to h.264, since it simply never was very good. But that can't be said about 8-bit h.264 today at all (when compared to 10-bit h.264). Plain high profile h.264 stays to be a perfectly good format (assuming that x264 gets used, naturally.)

[DmonHiro]

While OnDeed is correct, there is something he/she is forgetting. Almost every single xvid release is compatible with almost any player. It's actually quite hard to make an incompatible xvid .avi. That is not the case with h264. Most players cannot play anything that is encoded in a too high profile, for example. And since most encoders don't care, a vast majority of HD 8bit encodes were never compatible with hardware players to begin with. Expect media players, like WDLive, I guess. So yeah...

[Puiu]

Here we have over 20 pages of nothing but illogical bickering.
It is clear that the vast majority want 10bit to be treated equally with 8bit. If the file is smaller and the quality > or = then i see no reason why we can't close this thread right now. If you have an 10 year old PC then that's your problem, don't go causing problems for the rest. 

as i said in my last post:

So who wins the poll? The one with the most votes or only the one with >50% of the total votes?

Can't we just make a compromise? Some want 6 months and some want them now. How about we treat 10bit equally with 8bit releases starting this spring? Until then upload the FullHD 8bit ones too along side 10bit (that's if the 10bit ones are better of course).
The codecs and players are already good enough to play the files, but we can let them work a whole winter to sort out the remaining bugs, performance issues and also the fansubbing groups to become more proficient at using 10bit.

In my opinion, by then all new HD (720/1080p) releases will be in 10bit.

PS: for me the madvr 10bit looks better then the rest of the screens for ben-to (i also use madvr + potplayer +lav). Does anyone have screens of the BD 8/10bit?

[dragon191]

Here we have over 20 pages of nothing but illogical bickering.
It is clear that the vast majority want 10bit to be treated equally with 8bit. If the file is smaller and the quality > or = then i see no reason why we can't close this thread right now. If you have an 10 year old PC then that's your problem, don't go causing problems for the rest. 

as i said in my last post:

So who wins the poll? The one with the most votes or only the one with >50% of the total votes?

Can't we just make a compromise? Some want 6 months and some want them now. How about we treat 10bit equally with 8bit releases starting this spring? Until then upload the FullHD 8bit ones too along side 10bit (that's if the 10bit ones are better of course).
The codecs and players are already good enough to play the files, but we can let them work a whole winter to sort out the remaining bugs, performance issues and also the fansubbing groups to become more proficient at using 10bit.

In my opinion, by then all new HD (720/1080p) releases will be in 10bit.

PS: for me the madvr 10bit looks better then the rest of the screens for ben-to (i also use madvr + potplayer +lav). Does anyone have screens of the BD 8/10bit?

Thing is, wanting to keep an 8bit slot is not causing a problem for any 10bit fans, since they should still have their awesome 1080p/720p/whatever in 10bit. So it's actually those fanboys causing problems for the ones that can't view it.
Seriously, there is no harm in just having an 8bit slot.

[DmonHiro]

Seriously, there is no harm in just having an 8bit slot.

Obviously there is, if the staff of BakaBT asked the question in the first place. If there was no harm, nobody would be asking the question.