DivX 6.8 is out

[crnagora]

:helpful:

As the topic says:

DivX 6.8 is out

Personally I call this version the "Special (But too early) Christmas Version" :D

The changelog:

------------------------------------

Welcome to DivX Codec 6.8!

Following up on the improvements we added for HD support
in the 6.7 release, we've added additional optimizations
to our decoder to take full advantage of multi-core
computers. These latest advances allow the decoder to
perform up to 50% faster on multi-core machines making
HD content playback even better!

Additionally we have added the option to utilize custom
quantization matrices into our encoder. This will allow
you die-hard perfectionists to fine tune the encoder for
specific content types such as anime or CGI.

Please give us any feedback you have about this release
via the DivX Labs website:

labs.divx.com

And remember, if you want to show off your DivX video
creations to the masses, try publishing it to Stage6,
our video community website.

www.stage6.com

Stage6 offers free hosting for high-quality DivX videos
and lets you watch them at full screen in your web
browser using the DivX Web Player or download them to
enjoy on a DivX Certified device.

- The DivX Team

------------------------------------

Cheers and thanks to the DivX Team for still bringing MPEG4 ASP updates (As XviD seems die slowly on the ASP Part anyway).

[IgorC]

Cheers and thanks to the DivX Team for still bringing MPEG4 ASP updates (As XviD seems die slowly on the ASP Part anyway).

Please, don't start this kind of sh... divx vs xvid talks.
What actually does Divx bring?
Custom matrices? Xvid had already for YEARS and years.
Fast decoder? I invite you to compare divx decoder with ffdshow for speed and subjective/objective qualities.
Divx isn't better at all than Xvid encoder. Make your own test or search another ones.

[Inventive Software]

Divx isn't better at all than Xvid encoder.
Actually, in some ways it is. It has brought proper multi-threaded encoding, for one, which Xvid hasn't had (and I don't count 1.2, since that's not an official release. 1.2 has it's inefficiencies which are documented in the Xvid forum). CQMs are a welcome addition which will mean it makes it into DivX standalone players soon. And there's nothing wrong with another fast decoder! ;)

[Sharktooth]

finally CQMs!!!

[lancelot]

I do agree with Sharktooth - the CQMs will do the change - big change in the quality issues :) This was sth that DivX missed a lot and one of the reasons that Xvid was used to encode the movies :)

[dloneranger]

Does anyone else have a problem with this version?

I can not encode and playback at the same time smoothly
While I encode a vid in virtualdub (in idle), any video that is played that uses divx has intermittent pauses (even if the player is running at highest priority)

Previous versions do not exhibit the same behavior, and switching back the previous version solved it

[e-Pawel]

Cheers and thanks to the DivX Team for still bringing MPEG4 ASP updates (As XviD seems die slowly on the ASP Part anyway).

'Cause there's just no need to update Xvid :p Even when the DivX team brings new updates, they are still lightyears behind xvid.

And they still didn't manage to get their player to work properly *dohhh* When I exceed a certain size of the main window when resizing, the divx player crops a part of it (on the righthand side). Look at the screenshot:

http://img217.imageshack.us/img217/6417/divxsb1.th.jpg (http://img217.imageshack.us/my.php?image=divxsb1.jpg)

I do agree with Sharktooth - the CQMs will do the change - big change in the quality issues :)

But how to use the custom matrices? I mean how do I know which matrix will be more or less the most suitable for the given video content?

[crnagora]

I didn't start anything nor did I want a DivX vs. XviD showoff battle between the fans.

I just pointed out that we haven't seen a XviD Final for quite some time now (Like a final of the pretty stable XviD 1.2 Branch).

So stop this childish "Hell XviD is better" or "That feature was already featured in Codec X for years".

[Sharktooth]

@e-Pawel: there's no best as per forum rules. so your question is somewhat silly.

[Konrad Klar]

OK.
So, how do I know which custom matrice is appropriate for the given video content?

[Inventive Software]

Trial and error. If it's a well known CQM, try searching for it and see what people have said.

[DigitAl56K]

Fast decoder? I invite you to compare divx decoder with ffdshow for speed and subjective/objective qualities.

So do I :)

[weaver4]

Please, don't start this kind of sh... divx vs xvid talks.
What actually does Divx bring?
Custom matrices? Xvid had already for YEARS and years.
Fast decoder? I invite you to compare divx decoder with ffdshow for speed and subjective/objective qualities.
Divx isn't better at all than Xvid encoder. Make your own test or search another ones.

IgorC: Both XviD and DivX are great codecs, even in the 2005 Doom9 codec shootout they said their was very little difference between them. In the only double blind MOS test that I know of the DivX encoder came out slightly better.

http://www.compression.ru/video/codec_comparison/subjective_codecs_comparison_en.html

The new DivX decoder is very comparable speed and quality wise with the ffdshow now.

Just because DivX is a company does not mean that all their products are totally crap. While the XviD has been doing minor bug fixes for a couple of years the DivX team has be hard at work to improve their product, 6.8 has significant feature and speed improvements over 6.0.

IMHO DivX should be commended for the continued efforts in the ASP world.

[olnima]

Please stop this "which-one-is-better"-discussion. Let's all be glad to have different options like divx, xvid, 3ivx, etc. and it's up to every user to decide, which one to prefer.

The only thing I would like to tell the devs from divx is, that the speed-difference between "balanced"-mode (in modern times You will not find a CPU which drops a frame when encoding in realtime) and one quality-mode higher (even in modern times You will hardly find a CPU which DOESN'T drop a frame when encoding in realtime) is way too big. I own a E6850 and gone back to xvid, because for balanced-mode my CPU is way too fast and for better-mode it's way too slow, nothing in between.

Maybe a idea for developing future versions,

Olnima

[Hard Core Rikki]

Soo, no DivX love for single core or non-HT owners ?

[olnima]

Soo, no DivX love for single core or non-HT owners ?

My idea is not to eliminate existing modes I'm asking and hoping for a new mode speedwise between "balanced"- and "better"-mode.

Olnima

[weaver4]

olnima; don't know if your comment 'stop this "which-one-is-better"-discussion' was directed to me. But, the topic is "DivX 6.8" and many people including myself would consider moving to DivX if the new version had advantages over XviD. It is hard to evaluate the two codecs without a discussion comparing the two.

For example I would like to know more about CQM, and what advantage that gives DivX over XviD?

[Hard Core Rikki]

Depending on the content type, custom quant matrixes can help either reduce the bitrates needed for equivalent quality, or improve video compression quality for the same bitrate, for the same video. A compromise between the two approaches could ensure a slightly higher quality is obtained despite using a slightly lower bitrate.
A cartoon-tuned matrix should ensure a minimal loss of frequencies for that specific type of content, while the same matrix would not be appropriate for other types of videos, like live movies. Of course, its just 'try and see', as its not usually possible to guess how fine the final output will be, especially true for multipass encodes

Custom matrixes for divx videos might bring decoding issues on certified hardware divx players. Will this implementation require firmware updates for certified devices to be able to run videos created with custom matrixes ?

[Brother John]

and what advantage that give DivX over XviD?
No advantage. DivX has drawn level with Xvid, which has been supporting CQMs for years. CQMs are even interchangable between the two since they use the same file format.

In terms of supporting the MPEG-4 ASP standard both codecs are on par now. Off the top of my head I can’t think of anything that’s missing in either of them. Sure, Xvid still has a bunch of knobs and levers more (unrelated to the standard though). And DivX has its very straight forward hardware profile support. Quality and speed was roughly comparable even back in 2005 when Doom9 did his last codec shootout. I don’t have any new tests available, but I wouldn’t expect any major differences between the two in those areas.

Right now, not even price is an argument (http://www.divx.com/dff/).

[Mr_Schizo]

No advantage. DivX has drawn level with Xvid, which has been supporting CQMs for years. CQMs are even interchangable between the two since they use the same file format.

In terms of supporting the MPEG-4 ASP standard both codecs are on par now. Off the top of my head I can’t think of anything that’s missing in either of them. Sure, Xvid still has a bunch of knobs and levers more (unrelated to the standard though). And DivX has its very straight forward hardware profile support. Quality and speed was roughly comparable even back in 2005 when Doom9 did his last codec shootout. I don’t have any new tests available, but I wouldn’t expect any major differences between the two in those areas.

Right now, not even price is an argument (http://www.divx.com/dff/).
DivX still lacks 3 WP GMC afaik. So its not complete at all, even if GMC has no benefits in practise.

[Inventive Software]

I thought the limit for proper warp-points was 2? What does the spec say on this?

[Manao]

The spec says 3.

DivX has a better interlaced support, though it matters as much as GMC for most people here.

Finally, the MSU comparison was somewhat biased since deblocking was used for DivX but not for XviD ( as they say : "Probably, XviD results could be improved by switching on deblocking algorithm (this algorithm isn't used by default)" )

In the end, DivX and XviD are more or less on par. There's only so much you can do with such a limited (feature wise) standard anyway.

[Brother John]

A standard that was never primarily intended for simple rectangular video encoding anyway. :)

[Mgz]

so would the new PS3/XBOX divx playback support divx 6.8 :D

[Scarpad]

so would the new PS3/XBOX divx playback support divx 6.8 :D


Yes it does, I encoded a couple of files using both the Converter and the Dr. and they both play fine on the PS3

[Sharktooth]

IMHO, divx is slightly superior right now.
xvid is free and all, but i tried both and i have to say:
1- divx decoder is quite fast, supports MT and has a nice sharpening post processing option xvid hasnt. ffdshow has more options but we're comparing divx to xvid CODECs
2- the encoder is fast (still MT) and simple to setup (almost "fire and forget"). now it properly supports CQMs (like xvid) and that is a BIG step in the right direction for enhancing video quality for ppl who dont care for the SAPs compatibility. it also supports proper psy enhancements, while xvid's AQ efficiency is questionable.
3- EKG is far more advanced than zones control in xvid.
4- interlaced encoding supports fields order.
5- SAPs compatibility is assured without any hassle. xvid is compatible too, but the newbie doesnt even know what packed bitstream is... that means a bit of effort must be paid understanding the multiple codec options and what to do...

Picture quality is almost the same. Both codecs are over the top, even if, to my eyes, divx looks a bit softer but with less artifacts... but that's subjective. So, if you think those 5 points are valid, then DivX is what you need (few bucks for something easy to setup and that works) otherwise, you know, there's a free alternative.

PS. There is also the DivX media format... which i didnt mention, but that is not directly part of the codec.
did i miss anything?

[weaver4]

Sharktooth, great comments. Don't know if the DivX media format is an advantage or not. I made the mistake of putting all my videos in DivX format and now I am having fits converting them to videos for my iPod. Seems like a lot of major tools will not extract the audio from the DivX container; and I end up with video without sound. I have tried StaxRip, AutoMKV, RipBox264 and NeroDigital. Any ideas?

BTW: I encoded one video last night in avi.net in both xvid and divx and xvid 1.1.3 was 79fps and divx 6.8 was 112fps that is 42% higher, pretty impressive. I like to use avi.net because of the "packed bitstream" and such that you mentioned above; avi.net makes sure it is SAP compatible.

[buzzqw]

@weaver4

have you tried automen ?

BHH

[prof_unto]

Rename .divx to .avi then open and extract the audio with VirtualDubMod
Hope it helps

[Sharktooth]

i usually rename it to avi and extract audio with avidemux.

[weaver4]

buzzqw: Will automem make a iPod file? I know it will make a mp4 file but did not know if it would make one that would work on the ipod.

prof_unto and Sharktooth; I will give it a try I guess a two step process of ripping the audio and then encoding the video is not too bad, but if I would of stayed with *.avi it would be easy.

[buzzqw]

most probably the lower quality settings will be "more" compatible.. but i cannot guarantee.. just try on a very short sample

BHH

[DJ Bobo]

For people wondering about the decoding speed.
I made a comparison between the DivX 6.8 decoder (Deblocking on MAX, without sharpening) and the 4th beta of ffdshow-tryout (Nic's post-processing on 4, so without deringing), and here the results:

45% avg. CPU load for DivX
42% avg. CPU load for ffdshow

Test made w/ a single-core SSE2 machine on a standard b-framed XviD 1-minute clip (640x480 - 23.976fps - 900kbps).
I didn't test ffdshow w/ mplayer's postprocessing method because it's crap.

So DivX isn't that far behind after all.
I prefer the DivX decoder anyway, since I have some older DivX 4 and 5 encodes that don't display correctly with other decoders, and I don't wanna switch decoders all the time.

[DJ Bobo]

Here another test, this time with a slightly higher resolution (768x432), a much higher bitrate (3000kbps), and GMC & QPEL enabled:

60% avg. CPU load for DivX
46% avg. CPU load for ffdshow


With these extreme conditions, DivX is wa~y behind!

[Dark Shikari]

For decoding speed testing, try TimeCodec.

[GmorG McRoth]

or chegepuga

http://forum.doom9.org/showpost.php?p=763259&postcount=15

[DJ Bobo]

I think that CPU load measurement is more "real" than synthetic benchmarks.
Like when you benchmark a graphics card, you may find that Card A has a higher 3DMark score, yet Card B is actually faster with Game X.

PS: It's not like I'm guessing the averages. I run the test at least 2 times (more if the results are inconsistent). The averages are calculated using the logging function of Throttlewatch.

[Sharktooth]

first, make sure PP in both ffdshow and divx decoderd is disabled and then do a first bench, then enable comparable PP options on both decoders and do a second bench.
on my box i cant see all that difference in speed btw...

[DJ Bobo]

Well, Sharktooth, it doesn't make sense for me to disable post-processing, because I always watch with post-processing (as most people would... no?), so for a test to be as close to the reality as possible, I need it enabled.
And then, it's not like I can choose which post-processing plug-in I can use in DivX anyway, and Nic's post-processing is closest to DivX's quality-wise (as far as I can tell).
And even if the DivX decoder is actually as fast as libav, that would mean that the postprocessing is not fast enough, but then again, may be it's higher quality compared to Nic's and I didn't notice it?

By the way, as you can see above with the first "normal" test, there isn't much of a difference speed-wise. The difference becomes obvious only with cranked-up encoding options (which aren't common, I admit)

And hey, if you ever feel the need to compare the decoders without PP, please, go ahead, it would be interesting, but *I* don't feel like it... at least at the moment :p

[IgorC]

I think that CPU load measurement is more "real" than synthetic benchmarks.
Like when you benchmark a graphics card, you may find that Card A has a higher 3DMark score, yet Card B is actually faster with Game X.

PS: It's not like I'm guessing the averages. I run the test at least 2 times (more if the results are inconsistent). The averages are calculated using the logging function of Throttlewatch.
Timecodec is reallife benchmark. You run video under real conditions while output isn't null. It's not synthetic and you safe the time with this little program http://haali.cs.msu.ru/mkv/timeCodec.exe . Don't need to do any calculation neither using throttlewatch.

[DJ Bobo]

@ IgorC
Could it be that TimeCodec is limited by the monitor's refresh rate? 'cause I get results around 60fps (and my monitor has 60Hz refresh rate...)