CoreCodec/H.264 Codec "CoreAVC" [Archive] - Page 77

Dark Shikari

15th May 2008, 07:59

Anyone want to have a shot at comparing to see if DivX H.264 claims of speed to their new decoder are correct on their page? As they are boasting they beat CoreAVC. My initial tests in a small comparision says otherwise.
See: http://labs.divx.com/ProjectRemouladeIt'll be interesting to see the actual speed values here in real tests. One thing I'm noticing is that their assembly code is drastically worse than CoreAVC's, at least at first glance (see my analysis comments in their thread).

CiNcH

16th May 2008, 12:21

I have just realized that jitter is not just high with DVBViewer:

http://members.inode.at/762450/coreavc/jitter.jpg

Same graph with MainConcept and VMR Deinterlacing (or hardware deinterlacing) produces 0 jitter.

TheShadowRunner

16th May 2008, 16:22

hey all, I have an issue with the "Output level" colorspace set to "Autodetect" and Directvobsub.
I use VMR9.
When I disable Directvobsub, all is fine, CoreAVC detects VMR9 and sets PC levels. Autodect works.
If I insert DirecvobSub in the graph, it breaks Autodetect. Although the renderer is VMR9 alright, CoreAVC sets TV levels.

I'm guessing CoreAVC works as follows; if its "video out pin" is directly connected to VMR renderders, Autodect sets PC levels. In any other case it uses TV levels.
But inserting Directvobsub breaks this behavior even if the renderer is VMR, it will set TV levels.

I hope this can be fixed in order to be able to use "Autodetect" at all times. (when using Vobsub or not)

Also, there is an issue with CoreAVC 1.7 when changing resolutions.
I use Reclock and VMR9.
With CoreAVC 1.6.5, no problem; when Reclock changes resolution, the video resumes just fine.
With 1.7, after Reclock resolution change, the display is flashing/full of artifacts and the only way to correct it is to fake a change in the CoreAVC property page and hit "apply". Very odd.
Again, no problem with previous CoreAVC versions.
See you,

TSR

BetaBoy

17th May 2008, 22:24

I need to address something.... and this is a direct response to DivX's Project Remoulade speed claims at CoreAVC with the beta 1 release of their H.264 decoder as well as the current challenges to DivX's claims in their Project Remoulade thread, so I think it best to be stated here.

Over the past few years Doom9 himself has done many codec comparisons and the best thing as he has learned from his mistakes in going from one codec comparison to the next is to include as many details as possible (yes it was mostly encoding related but it is still relevant). As DivX stated in their Labs post they have decided it was best to directly compare CoreAVC's speed to that of their new decoder. Now that is nothing short of 'great' for us and the community at large as I have already stated, but my point of this post is that such claims were done in more of a 'PR' context, and only explains about half of the details of why they claim it is faster and that it simply falls short of providing true details (aside from the 8 core support) and has only left more confusion here at D9 and in general.

ATM we will not respond to such speed claims (other then what I stated that it being slower against all my local QA content.) till later Beta stages of their decoder release leading to the final release. But in the end as it matures with each release (and it will) if its better, great! It will be an awesome challenge for us in our current CoreAVC 2.0 development (and even upcoming 1.8.x and 1.9.x releases).

So DivX post more info next time, not for us... but for the community to verify and decide for themselves.

CiNcH

17th May 2008, 23:15

I for my part will not take the decoder using 1% less CPU time, but the decoder that delivers a well deinterlaced and judder free image...

bob0r

17th May 2008, 23:33

@BetaBoy

They basically say their decoder can do more than 4 cores (threads), maybe you should take a hint from that.

BetaBoy

18th May 2008, 00:17

@BetaBoy

They basically say their decoder can do more than 4 cores (threads), maybe you should take a hint from that.

I'm not referring to that but hint taken and we already need to know what to do for 2.0.

BetaBoy

18th May 2008, 00:19

I for my part will not take the decoder using 1% less CPU time, but the decoder that delivers a well deinterlaced and judder free image...

CiNcH I noted your issue but there were still outstanding issues with DVBViewer we had addressed with developers last time I checked with Haali.

Shinigami-Sama

18th May 2008, 00:33

CiNcH I noted your issue but there were still outstanding issues with DVBViewer we had addressed with developers last time I checked with Haali.

maybe its time for CoreDBV?
lord knows theres problems with it everywhere in this forum, and dozens of work arounds...

BetaBoy

18th May 2008, 00:41

maybe its time for CoreDBV?
lord knows theres problems with it everywhere in this forum, and dozens of work arounds...

lol... we squashed that idea about 3 years ago ;-) on work arounds we added a few for DVBViewer to the last 2 releases. On ping the guys here to get a update on where we are atm.

TheShadowRunner

18th May 2008, 01:24

being ignored sucks.

bob0r

18th May 2008, 11:55

being ignored sucks.

Thats a generic core trademark :)
Anyways, if you put some reproducable files online in a zip file or something, we can reproduce and confirm the problem.

However the AUTO option should be taken very lightly i think, its best to set your own levels.

Most bluray/hddvd/satellite(some are pc level) are TV level.
On my DELL LCD screen output is PC level.
So i simply set input:TV ouput:PC

But i still agree bugs are bugs and being ignored is horribly annoying!

BetaBoy

18th May 2008, 13:01

Note.... _nothing_ is ignored we add any potential bugs posted here and the ones that get emailed to us.

CiNcH

18th May 2008, 13:04

CiNcH I noted your issue but there were still outstanding issues with DVBViewer we had addressed with developers last time I checked with Haali.

Well, latest issue does not seem to be related to DVBViewer at all but more to VMR deinterlacing (hardware / DirectShow deinterlacing or whatever you want to call it). I have the same jitter with GraphStudio/GraphEdit when using Haali Media Splitter.

But it is quite cool you are addressing DVBViewer issues. Hope you inform DVBViewer developers about issues that you have found...

Ice =A=

18th May 2008, 17:00

bob0r: They basically say their decoder can do more than 4 cores (threads), maybe you should take a hint from that. I really can see no sense in supporting more than 4 cores.
Will there ever be an 8 core cpu that is too slow to decode even the most demanding videos with "just" 4 of its cores?!? I hardly doubt that...

LoRd_MuldeR

18th May 2008, 17:39

I really can see no sense in supporting more than 4 cores.
Will there ever be an 8 core cpu that is too slow to decode even the most demanding videos with "just" 4 of its cores?!? I hardly doubt that...

A few years ago, when everybody was watching SD video, nobody would have expected that you need more than one core for video playback.
Then HD video was born and new complex video formats (e.g. H.264) became popular. Suddenly 2 or even 4 cores are needed for smooth video playback!
You can expect the same to happen again: At the moment we use 720p or 1080p, but what if 2k or 4k resolutions become popular?
Also keep in mind that future CPU's won't get improved performance from higher clock speed in first case, but from higher number of cores...

IgorC

18th May 2008, 17:52

Then HD video was born and new complex video formats (e.g. H.264) became popular. Suddenly 2 or even 4 cores are needed for smooth video playback!
Can you show 1080p H.264 (up to 50 mbit) content that would stutter on decent dual core? I have a very low budget dual core and every 1080p video plays fine on it.

You can expect the same to happen again: At the moment we use 720p or 1080p, but what if 2k or 4k resolutions become popular?
Also keep in mind that future CPU's won't get improved performance from higher clock speed in first case, but from higher number of cores...
2k 4k are still very futuristic. I don't see it coming at least for 2 years.

LoRd_MuldeR

18th May 2008, 18:02

Can you show 1080p H.264 (up to 50 mbit) content that would stutter on decent dual core? I have a very low budget dual core and every 1080p video plays fine on it.

CABAC plus enough Bitrate should be able to push every CPU to it's limits :p

2k 4k are still very futuristic. I don't see it coming at least for 2 years.

Well, 1080p was "futuristic" too only a few years ago...

lexor

18th May 2008, 18:13

CABAC plus enough Bitrate should be able to push every CPU to it's limits :p
bleh, P4 decoded HD DVD perfectly fine in the first gen players. So for practical purposes, CABAC and enough bitrate is not a problem.

Well, 1080p was "futuristic" too only a few years ago...
Yes, 10y+ at least for 2k and up, Igor was definitely way too optimistic. Still to require higher number of cores would suggest that each core isn't going to noticeably increase in performance, which I think is highly unlikely.

Manao

18th May 2008, 19:37

1080p was futuristics a few years back, but HD DVD / bluray were already planned. Nothing equivalent to those two are planned for 4K yet. (2K == 1080p, btw).

LoRd_MuldeR

18th May 2008, 19:48

1080p was futuristics a few years back, but HD DVD / bluray were already planned. Nothing equivalent to those two are planned for 4K yet. (2K == 1080p, btw).

Not exactly :p

AR 4K 2K
1.66 4096 × 2464 2048 × 1232
1.78 4096 × 2300 2048 × 1150
1.85 4096 × 2212 2048 × 1106

And AFAIK they already use it in cinmea production, because it's they only way to keep nearly the same amount of information as "classical" film cameras do...

DigitalDeviant

18th May 2008, 20:14

1080p was futuristics a few years back, but HD DVD / bluray were already planned. Nothing equivalent to those two are planned for 4K yet. (2K == 1080p, btw).

Nothing? (http://www.nhk.or.jp/digital/en/super_hi/index.html)

Manao

18th May 2008, 20:27

LoRd_MuldeR : 2K and 1080p are close enough. As for 4K, it's barely used in the cinema yet, so it won't come anytime soon for us.
DigitalDeviant : I'm more talking about format accessible to average Joe. SHD isn't for tomorrow (NHK themselves doesn't expect it before 2018~2020)

HDDVD/Bluray came roughly 12 years after DVD. Unless Bluray utterly fails commercially, I don't think the next standard will come before 6 to 10 years. Airwave numeric TV will stop at 1080i because of bandwidth issue, so will satellite. I'm less sure about cable/IPTV, but I don't think there'll be a market for >1080p TV only for cable/iptv. So I really don't expect anything over 1080 for us anytime soon.

ADude

18th May 2008, 20:57

I don't expect anything over 1080 ever.

Note that MP3 players have significantly poorer sound quality than the preceding generation - CD players.

Many surveys show that the average HD TV owner thinks that stretched, upconverted SD progamming is "HD". The group of people who have not yet upgraded to HD, have (as a whole) an even lower quality standard.

And, studies have shown that with the normal sitting distance from the home TV, the human eye cannot discern resolution greater than 720p with the most common HD TV screen sizes (32-50 inches).

So, 2k is not ever going to happen.

Manao

18th May 2008, 21:08

Sorry for OT...Note that MP3 players have significantly poorer sound quality than the preceding generation - CD players.Strange analogy. The players got more portable, able to store more music, which was what people wanted. Quality degradation came as an inevitable but acceptable tradeoff.Many surveys show that the average HD TV owner thinks that stretched, upconverted SD progamming is "HD". The group of people who have not yet upgraded to HD, have (as a whole) an even lower quality standard.Sadly, that is mostly true. Though i hope they'll see the difference when actual HD programs become mainstream and their eyes get used to it.

LoRd_MuldeR

18th May 2008, 21:20

So, 2k is not ever going to happen.
Wait until the market is saturated with Blue Ray players and "Full HD" screens in a few years. They will come up with a new "technology" for sure :rolleyes:

I guess it will be 4K, since 2K would be too close to 1080p. Whether it will still be delivered on optical media is a different question ...

Dreassica

18th May 2008, 21:29

U would need 80 inch screens to even see the difference anymore, as to see 1080p u already need 50"to see a real difference. I doubt many people will ever want a 80 incher in their living room.

LoRd_MuldeR

18th May 2008, 21:47

U would need 80 inch screens to even see the difference anymore, as to see 1080p u already need 50"to see a real difference. I doubt many people will ever want a 80 incher in their living room.
Maybe a canvas and a beamer then? Takes almost no room, unless you unroll the canvas. And it's bigger than any screen...

(BTW: This is really getting off-topic ^^)

Revgen

19th May 2008, 07:59

I have a 24 inch screen on my computer, and I absoutely see the difference between 720p and 1080i. I can even see a difference in 1080i content even when my desktop resolution is at 1200x800.

Sorry, but there is a difference. At least to my eyes anyway.

TEB

19th May 2008, 11:17

Its all about viewing distance. Our eyes need to be able to differentiate the different pixels..
There are alot of discussions on this on avsforum and other equivilent forums regarding how close one need to sit vs. screen size.
Other factors like correct 24p handelig, high contrast, deinterlacing is at large the main factos to what customer see as the "wow" effect..

just my 2cents..

smok3

19th May 2008, 12:34

if you look at red, then it appears that 4k is allready here;
http://www.red.com/

Ice =A=

19th May 2008, 12:45

And, studies have shown that with the normal sitting distance from the home TV, the human eye cannot discern resolution greater than 720p with the most common HD TV screen sizes (32-50 inches).
So, 2k is not ever going to happen. Exactly what I was thinking.
So there are only two options left: Decreasing the distance to the TV, which obviously makes not much sense. Or increasing the screen size, which is very limited in "normal" living rooms.

smok3

19th May 2008, 15:21

i can certainly imagine screens being sold/cut by meters to fit on certain wall in the not-so-near future, i guess that would be the time to include DPI as marketing :)

so what you are saying is that for example 16:9 82 cm diagonal screen is nonsense for the full-HD res?

the_corona

19th May 2008, 15:41

ChronoCross

19th May 2008, 18:46

Kinda getting off-topic, no? :-)

Anyways, in their update Divx say that they think the problem stems from a diff. color-space being used and that the "problem" would be resolved swiftly. I'm wondering, no, hoping that CoreAVC will respond quickly as well and we consumers are in for a treat :-) Any statement from CoreAVC regarding this matter?

cept if you read the divx thread even when coreavc outputs the same colorspace Coreavc is still beating Divx. Most people are seeing that it's slower than even ffdshow

MatMaul

20th May 2008, 22:53

hey !
any chance to have NV12 as output format ?
I have chroma upsampling problems with YV12 and the VMR9 renderer, and NV12 solved them.

CiNcH

8th June 2008, 15:39

0 jitter with CoreAVC Trial + DVBViewer + ORF 1 HD (720p50). Seems they can work together. As I pointed out, problem is with interlaced content, where output is jittery.

CiNcH

8th June 2008, 18:49

Seraphic-

8th June 2008, 19:53

1 posting Phishing... i'll bite... Use :search: in this thread you'll find about a dozen pages back I mentioned our encoders will be out later this year...

What I would like to tell you all is that because of the D9 Elite asking for it, that we will soon be releasing a CoreAVC 64 Professional Edition. Don't ask how soon but it is on the short term roadmap here to get it QA'd and out.

BetaBoy, will the upcoming CoreAVC Encoders support High 4:2:2 Profile 480p/720p/1080p encoding?
And what price are we talking about for the encoder software?

IgorC

9th June 2008, 02:40

Sorry for offtopic

Note that MP3 players have significantly poorer sound quality than the preceding generation - CD players.

The situation is totally oposite. Good mp3 encoder is transparent on midle/high bitrate.

LAME at -V 5 (135 kbits) had 4.6/5 points in public test.
-V 2 (192 kbits) officialy transparent even on good Hi Fi system.

People wouldn't accept mp3 if it had low quality.

Shinigami-Sama

9th June 2008, 07:53

People wouldn't accept mp3 if it had low quality.

you don;t know many people then
I've seen many people that are amazed when they hear an mp3 that even sounds like fmradio...

BetaBoy

9th June 2008, 10:05

Kinda getting off-topic, no? :-)

Anyways, in their update Divx say that they think the problem stems from a diff. color-space being used and that the "problem" would be resolved swiftly. I'm wondering, no, hoping that CoreAVC will respond quickly as well and we consumers are in for a treat :-) Any statement from CoreAVC regarding this matter?

Well... we have a few more updates to release over the next few months before we publish our 2.0 Alpha.

On the OT discussions.... I am enjoying them, really ;-)

deekey777

9th June 2008, 11:33

Watching Austria - Croatia in 720p50 now. CoreAVC seems to introduce some kind of aliasing effect at the white lines of the soccer/football pitch, which is pretty weird for progressive material. Deinterlacing is explicitly set to 'None'. MainConcept is fine..

[EDIT]
Due to the color space, forced it to YUY2 now.

Actually all games are produced in 1080i50 by UEFA, HD Suisse and ORF HD (and nl1 HD) do downscale them to 720p50.

News in Geman: http://www.heise.de/newsticker/Fussball-EM-Gewaehltes-HDTV-Ausstrahlungsformat-bringt-nur-Nachteile--/meldung/108493

CiNcH

9th June 2008, 11:39

Actually all games are produced in 1080i50 by UEFA, HD Suisse and ORF HD (and nl1 HD) do downscale them to 720p50.

I know that much. But deinterlacers and scalers used by ORF are not that bad. It was a colorspace problem...

Hardware decoding currently fails for 720p50 under XP (for both nVIDIA and ATi in bitstream mode). Pretty jerky playback. No problem under Vista with EVR. MainConcept and CoreAVC and software decoding is fine too.

BetaBoy

11th June 2008, 07:49

Ok guys.... As many of you know I have mentioned this before but we are about to finally go into beta phase of CoreAVC 64, and are looking for a few elite D9 testers that have 64 bit systems that can test and provide feed back on.

Now this does not feature our upcoming 2.0 branch changes, but it does feature many of the things not in 1.7 like our levels slider (brightness, contrast, saturation).

PM me you system specs.... thx

JohnnyFu

12th June 2008, 23:18

What is the recommended software setup for CoreAVC 64, e.g. MPC HomeCinema - x64 ? wich renderer should be used, wich not ?

BetaBoy

19th June 2008, 05:36

Along with CoreAVC 64 we are bundling 64 bit versions of TimeCodec as well as the Haali Media Splitter. Everything else is on a per user system basis and for them to report back on.

Shinigami-Sama

19th June 2008, 06:59

betaboy
will the 64bit be only for vista or xp64 as well?

BetaBoy

19th June 2008, 14:00

ATM XP/Vista 64 yes..... Were you thinking Linux? If so, OEM wise it will now also be available as a licensable library for it (Linux) as well. One thing to note.... Haali will include both the 32 bit and the 64 bit versions in the same installer package so there is no need for two diff installers.