wich codec is faster?

[sarmin]

i convert much and i want something fast so i don't have to wait much. wich codec is the fastest between xvid and divx? and what encoder should is use do get the fast encode. i convert all my video files in 2 passes.

[Sharktooth]

it depends on what settings you will use.

[CWR03]

If your only concern is speed, why not do 1-pass encodes?

[Revgen]

i convert much and i want something fast so i don't have to wait much. wich codec is the fastest between xvid and divx? and what encoder should is use do get the fast encode. i convert all my video files in 2 passes.

If you're comparing DivX vs. Xvid than I'd say Xvid is definitely faster with all systems.

However X264 single-pass is just as fast as Xvid 2-pass while giving better quality. At least for me anyway.

[Teegedeck]

...and also for x264 goes: it completely depends on the settings.

[sarmin]

k if i use the smae settings and 2 passes for both then wich one is faster.

[Sharktooth]

they have different settings... and speed depends also on the system setup (for example Xvid prefers AMD CPUs)...
that means, try it by yourself and see what's faster for you.

[xyloy]

is the fastest between xvid and divx
On my system(1,67 GHz AMD Sempron), with "maximum" settings(:all "advanced" and useful options enabled, full motion estimation, 2 pass...), XviD v1.2 took 5 hours, and DivX v6.2.2 took 32 hours(even x264 with 'crazy' settings, such as 24 frames for M.E. Range, RDO Leve 2, and 16 ref. frames, only took 29 hours!) for a one hour and a half-lenght movie.

That should give you an idea on wich is fastest(whatever the system setup is).

[sarmin]

so what converter should i use when i'm converting to xvid. i don't want to use vdub cuz it very slow on my computer. here is my computer specs:
3000+ amd 64
1024mb ram ddr 400mhz
x700 128mb

[GodofaGap]

VirtualDub has no influence on codec speed.

[weaver4]

The latest DivX codec is faster than XviD with similar quality. I have tested a couple of movies using XviD and DivX with avi.net and autogk. I used single pass, on a single core processor.

[Sharktooth]

It all depends on the settings and on the system setup.
Please stop asking "what... what... what...", you already had your answers. Try it BY YOURSELF.

[Sagittaire]

On my system(1,67 GHz AMD Sempron), with "maximum" settings(:all "advanced" and useful options enabled, full motion estimation, 2 pass...), XviD v1.2 took 5 hours, and DivX v6.2.2 took 32 hours(even x264 with 'crazy' settings, such as 24 frames for M.E. Range, RDO Leve 2, and 16 ref. frames, only took 29 hours!) for a one hour and a half-lenght movie.

That should give you an idea on wich is fastest(whatever the system setup is).

I can use Libavcodec and make 0.01 fps for ASP but conclusion is not "Libavcodec is slow". If you want make speed comparison you must make comparison with exactly the same quality and size for output.

The fastest mode for DivX is by far faster that the fastest mode for XviD. The fastest mode for x264 is faster than the slowest mode for XviD and done by far better quality.

[Teegedeck]

The fastest mode for x264 is faster than the slowest mode for XviD and done by far better quality.If you are to correct over-generalisations made by others, please don't make up new over-generalisations yourself.

To be more precise, it's a bit like this:


XviD is better, though slower, than x264 down to compression ratios of about 1/4 of of MPEG-2's (overgeneralisation... :-))*
x264 is faster and better than XviD at compression ratios below about 1/4 of MPEG-2's

* In order to be a bit more precise: IMHO x264 looks worse if put against XviD with Shartooth's EQMv3 LR at quantizer=4 or settings aiming for higher bitrates than that. Perhaps x264 looks even worse than XviD with EQMv3 ULR rev2 at quantizer 3 or 4 but that's a matter of taste, blocks (XviD) against lower level of detail (x264). Most of the time, XviD will be slower than x264.

At filesizes lower than what XviD with EQMv3 ULR rev2 at quantizer=4 can produce, x264 is always better and certainly always faster.

The point is, I would never go for lower quality than EQMv3 LR at quantizer=4 so for me XviD is always better than x264. Apparently you, Sagittaire, wouldn't go for that sort of filesizes/bitrates, so I guess for you x264 is always better than XviD. It's all a matter of usage and taste.

[Didée]

It's all a matter of usage and taste.
... and of the method-of-evaluation. If some artificial numbers say that X is "better and by far" than Y, then ... you know. Moreover, if one has spent much of time with gathering artificial numbers that say that something is "better and by far" than something other, there is danger that personal visual experience gets biased in a certain direction. (Orwell ... "how many fingers do you see?" ;) )

[Sagittaire]

To be more precise, it's a bit like this:

XviD is better, though slower, than x264 down to compression ratios of about 1/4 of of MPEG-2's (overgeneralisation... :-))*
x264 is faster and better than XviD at compression ratios below about 1/4 of MPEG-2's

No it's always better simply because for equivalent size at your XviD with EQMv3 ULR q4 I can use very higher resolution for x264 and very better detail level.

Here sample for SWIII 720p at 1080 Kbps, french and english 5.1 aac+ 140 Kbps, 1.4 Go for 140 min, source HDTV, as you know certainely it's a very uncompressible source.
http://multimediacom.free.fr/Video/SWIII_Sample.mp4

If you choose 720*400 for EQMv3 ULR q4 for DVD Backup I choose simply 720*576, 1024*576 or 1280*720 for x264 and quality will be always very better and by far.
http://multimediacom.free.fr/Video/Matrix.mp4
http://multimediacom.free.fr/Video/Sample_720p.mp4

[Teegedeck]

Sagittaire, you cannot prove wrong what I said if you use a different approach of testing than I suggested.

At the same filesize and same resolution there is a point when x264 will start to look the same or even worse than XviD.

Here's an easy way for you to find out where x264 stops being better than XviD:

Encode MPEG-2 (unfiltered!) to XviD with EQMv3 HR (for the settings, see the 'XviD presets' thread or use these profiles (http://forum.doom9.org/attachment.php?attachmentid=5894&d=1150539946) for MEGui) at quantizer=4. Encode with x264 in two pass (at the same resolution) to reach the same filesize.
Encode MPEG-2 (unfiltered!) to XviD with EQMv3 HR at quantizer=3 Encode with x264 in two pass (at the same resolution) to reach the same filesize.
Encode MPEG-2 (unfiltered!) to XviD with SixOfNine at quantizer=4 Encode with x264 in two pass (at the same resolution) to reach the same filesize.
Encode MPEG-2 (unfiltered!) to XviD with SixOfNine at quantizer=3 Encode with x264 in two pass (at the same resolution) to reach the same filesize.

I'm positive that at some point you will stop seeing (with the things in your eyesockets ;)) an advantage of x264; perhaps already with the first test.

And I'm pretty sure that you will start seeing an advantage for XviD somewhere along (if you use a good display).

No it's always better simply because for equivalent size at your XviD with EQMv3 ULR q4 I can use very higher resolution for x264 and very better detail level.Errr... No.

If you choose 720*400 for EQMv3 ULR q4 for DVD Backup I choose simply 720*576, 1024*576 or 1280*720 for x264 and quality will be always very better and by far.
Sagittaire; please do read carefully what I write. I w r o t e:

Perhaps x264 looks even worse than XviD with EQMv3 ULR rev2 at quantizer 3 or 4 but that's a matter of taste, blocks (XviD) against lower level of detail (x264).[...]
At filesizes lower than what XviD with EQMv3 ULR rev2 at quantizer=4 can produce, x264 is always better and certainly always faster.So I don't see a contradiction if in your samples x264 looks better. I haven't had a look at them, yet, though.

[Sagittaire]

Here's an easy way for you to find out where x264 stops being better than XviD:
Encode MPEG-2 (unfiltered!) to XviD with EQMv3 HR (for the settings, see the 'XviD presets' thread or use these profiles for MEGui) at quantizer=4. Encode with x264 in two pass (at the same resolution) to reach the same filesize.
Encode MPEG-2 (unfiltered!) to XviD with EQMv3 HR at quantizer=3 Encode with x264 in two pass (at the same resolution) to reach the same filesize.
Encode MPEG-2 (unfiltered!) to XviD with SixOfNine at quantizer=4 Encode with x264 in two pass (at the same resolution) to reach the same filesize.
Encode MPEG-2 (unfiltered!) to XviD with SixOfNine at quantizer=3 Encode with x264 in two pass (at the same resolution) to reach the same filesize.

No for my eyes with same resolution x264 done always better or egal performance than XviD. Use best possible setting (RDO, trellis, ME6, multipass) with magical CQM don't change anything. There are simply transparency threshold for high quality for all video codec and in this case choose AVC or ASP don't change anything and I will choose to use higher resolution for x264. Moreover for me H263 quantisation (with PP4 and soft threshold/strength for deblocking process) is the best way for my eyes ...

[GodofaGap]

On my system(1,67 GHz AMD Sempron), with "maximum" settings(:all "advanced" and useful options enabled, full motion estimation, 2 pass...), XviD v1.2 took 5 hours, and DivX v6.2.2 took 32 hours(even x264 with 'crazy' settings, such as 24 frames for M.E. Range, RDO Leve 2, and 16 ref. frames, only took 29 hours!) for a one hour and a half-lenght movie.

The fastest mode for DivX is by far faster that the fastest mode for XviD. The fastest mode for x264 is faster than the slowest mode for XviD and done by far better quality.

At the same filesize and same resolution there is a point when x264 will start to look the same or even worse than XviD.

Ok I'm bored.

There should be a rule against anecdotal speed and quality statements. This discussion doesn't really help anyone.

[Teegedeck]

There are simply transparency threshold for high quality for all video codec and in this case choose AVC or ASP don't change anything and I will choose to use higher resolution for x264. If your threshold for perceived transparency is as low as, say XviD, EQMv3 HR @quant=4, be happy about it! :)

As they say on hydrogenaudio.org regarding lossy audio compression: There's no sense in encoding to a higher bitrate/filesize if you can't tell the difference. I'm pretty unfortunate; my perceived transparency threshold is between SixOfNine @ quant=3 and quant=4 for most movies. For some only lowering VHQ settings and ME precision does deliver perceived transparency, with the ~10% filesize bloat that comes with it.

[Sagittaire]

If your threshold for perceived transparency is as low as, say XviD, EQMv3 HR @quant=4, be happy about it! :)


Well certainely for 1080p encoding but not for CQIF resolution. This threshold is function of the resolution.

There should be a rule against anecdotal speed and quality statements. This discussion doesn't really help anyone.

The only way for make that seriousely is objective comparison way (no way for subjective speed comparison).
http://forum.doom9.org/showthread.php?t=105763&highlight=speed

[GodofaGap]

Well ok, but then every one who disagrees will with the results will go PSNR blah blah blah. But at least your test is transparent. :)

I just don't like people talking about 'maximum settings' (what are they and do they even make sense?) and people devising tests for other people to do to prove their own view. If you believe what you say prove it yourself please or shut about it.

I really don't see how a discussion like this is useful to anyone, it's a lot of talk but little of anything said here is proven.

[Sagittaire]

Well ok, but then every one who disagrees will with the results will go PSNR blah blah blah. But at least your test is transparent. :)

I just don't like people talking about 'maximum settings' (what are they and do they even make sense?) and people devising tests for other people to do to prove their own view. If you believe what you say prove it yourself please or shut about it.

I really don't see how a discussion like this is useful to anyone, it's a lot of talk but little of anything said here is proven.

Well as you can see with my test (720*400 uncompressed source with monocore sempron 2.1 ghz) DivX encode to [10, 90] fps with various setting. The only way for make speed comparison is objective way ...

[CruNcher]

@sarmin
if your really searching the fastest ASP codec and Detail Preservation is no reason for you to chose then i sugesst you Jomigos HDX4 Codec http://www.hdx4.com/ as it is the fastest but also the one that drops the most details. Im not quiet sure if the new DivX might be able to go as low but i doubt it so it should be still the fastes for the moment. And im sure we gonna see soon their Ultrafast H.264 codec too ;)

[Teegedeck]

And don't forget about Ahead's MPEG 4 ASP codec in Nero. If I recall it correctly it's very fast and (meanwhile) the quality is nothing to sneeze at.

[weaver4]

I think the original question was which is faster, xvid or divx.

Well if you think that avi.net and autogk set up the default parameters correctly, and they do equal quality for both xvid and divx then my own testing show that latest DivX codec is slightly faster using single-pass.

[gameplaya15143]

divx4/divx505 are much faster on my system than xvid1.2. Never used divx6 so I can't comment there.

I can capture at 640x480 with divx505 on my P4 2.4ghz (just barely anyways), so that should say something there.

Faster yet would be the ATI codecs (directshow, no 2pass).. ghastly quality (disable b-frames ;) ).. but lightening fast :eek:

[Blue_MiSfit]

but why oh why would you not care about quality? there simply HAS to be a balance between speed and quality!!

For me, it's all about quality. I dont mind getting 7fps average for my second pass in x264.

The way I see it, the only reason to care about speed over quality is if you have to cap. In this case, huffyuv.

Thats all I have to say :)

~MiSfit

[xyloy]

I just don't like people talking about 'maximum settings' (what are they and do they even make sense?)
Sorry I was too lazy to post them, but they make sense(options wich give better quality like QPEL are used, and the Motion Estimation level is the highest as possible, so to speak) :
(my test was "DivX vs. XviD - with same settings on both sides when possible - at 300 Kbps, and with a 1 hour 30 minutes CG anime movie, 656*272 pixels, 25 FPS")
XviD v1.2 Alpha:
http://membres.lycos.fr/xcomcmdr/XviD1.png
http://membres.lycos.fr/xcomcmdr/XviDCrazySettings.png
http://membres.lycos.fr/xcomcmdr/XviDCrazySettings2.png
http://membres.lycos.fr/xcomcmdr/XviD4(AdvancedOptions).png
http://membres.lycos.fr/xcomcmdr/XviD4(TrellisQuantization).png

=> The encoding took only five hours and 22 minutes.

DivX 6.2.2 Pro(Trial):
http://membres.lycos.fr/xcomcmdr/DivX6.2.2CommunityProCodecCrazySettings1.png
http://membres.lycos.fr/xcomcmdr/DivX6.2.2CommunityProCodecCrazySettings2.png
http://membres.lycos.fr/xcomcmdr/DivX6.2.2CommunityProCodecCrazySettings3.png
(oops, sorry for the french translated settings' names, but I think it's understandable anyway..)

=> The encoding took... 32 hours and 16 minutes.

So I think the difference is so huge that whatever the system setup is, with the same settings the conclusion will be the same: XviD is a lot faster than DivX (and the quality is a little better, but that's a another topic).

Of course I use x264 wich was a billion galaxies away better, and a little faster than DivX(29 hours 17 minutes), albeit the "M.E. Range" and "Max. Ref. frames" settings I used didn't make sense(I just wanted to play with x264 and so I used "crazy" settings):
--pass 1 --bitrate 300 --stats ".stats" --ref 16 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --b-rdo --bime --weightb --direct auto --subme 7 --trellis 2 --analyse all --8x8dct --me umh --merange 24 --cqm "jvt" --progress --no-dct-decimate --no-psnr
(same setings for the 2nd pass)

[Sagittaire]

1) For MPEG4 ASP.

- Well IMO QPel for ASP is good for low quantizer (high bitrate) but not high quantizer. GMC (1WP or 3WP) is slow and don't improve really the quality

- Use Optimized H263 for DivX and not psy mode

- I think that Chroma optimizer is pre-process filter for XviD like denoising for DivX.

- It's not a obligation to use "insame" quality for DivX. If you want really the best quality you can can use -pq 9999 but it's useless too.

2) For MPEG4 AVC

Your setting are simply ridiculous slow ...

- use slowest setting for first pass is useless
- use 16 ref is useless
- use jvt matrix here (high quant) is useless, flat16 (or similar) matrix is really better with inloop and by far
- use merange 24 is useless

Use these setting ...

--pass 1 --bitrate 300 --stats ".stats" --ref 5 --mixed-refs --bframes 3 --b-pyramid --weightb --direct auto --subme 3 --trellis 0 --analyse all --8x8dct --me dia --progress --no-psnr

--pass 3 --bitrate 300 --stats ".stats" --ref 5 --mixed-refs --bframes 3 --b-pyramid --weightb --direct auto --subme 5 --trellis 0 --analyse all --8x8dct --me hex --progress --no-psnr

--pass 2 --bitrate 300 --stats ".stats" --ref 5 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --b-rdo --bime --weightb --direct auto --subme 7 --trellis 1 --analyse all --8x8dct --me hex --progress --no-dct-decimate --no-psnr

... and quality will be better with faster speed for 3 pass ... lol

[Manao]

xyloy : firstly, you should have used "H263 optimized" instead of simply "H263". "H263 Optimized" == "H263" + "Trellis". Trellis increases PSNR by 0.3 dB - which is more than the increase between VHQ 0 and VHQ 4 in XviD.

Secondly, you claim you're using the same settings. That's wrong. First you forgot trellis, but even so, how can you claim the settings to be the same ? Do you know how DivX is coded ? Of course not. So I'll assume that by same settings you mean either "highest possible settings", in which case you're blaming a codec for actually having a wider range of speed/quality tradeoff available to the user, or "same quality", in which case you need to go forward with PSNR figures, or something else a bit objective.

But then, if you assert the quality to be the same, do the following test : encode the movie with VHQ 2 and VHQ 4 in XviD. Can you see the difference ? I bet you can't. Though you definitely can see a speed difference.

A subjective comparison can't distinguish small quality differences. Once you added trellis, XviD and DivX should have the same quality. There will still be the huge speed difference. But I bet you that you can lower DivX quality settings without noticing any differences, even compared to your XviD clip. The speed, however, will increase tremendously.

So, once again, you have to take great care when you want to compare codecs. And your methodology is not perfect.

Finally, don't get me wrong here. I'm not saying DivX is actually better than XviD. It's just that I'm a bit fed up about people claiming it's slower than XviD because as soon as they see a quality slider, they can do nothing but push it to the maximum. XviD could have had far slower quality settings than VHQ 4, which still would have marginally improved the quality. They decided not to implement them, while DivX did. That's why DivX is ""slower"".

[sarmin]

what settings should i use when converting a movie to divx? should i use the foming or shaping option? should i use any redution? or psy mode?

[xyloy]

Your setting are simply ridiculous slow ...

- use slowest setting for first pass is useless
- use 16 ref is useless
- use jvt matrix here (high quant) is useless, flat16 (or similar) matrix is really better with inloop and by far
- use merange 24 is useless
I know it is uselessly slow, but as I said they are just "crazy" and just for testing
thanks for the matrix tip

- Well IMO QPel for ASP is good for low quantizer (high bitrate) but not high quantizer. GMC (1WP or 3WP) is slow and don't improve really the quality

- Use Optimized H263 for DivX and not psy mode

- I think that Chroma optimizer is pre-process filter for XviD like denoising for DivX.
thanks for the tips, I didn't know what to choose.
Chroma Optimizer tries to manage the "red stairs" effect.

First you forgot trellis, but even so, how can you claim the settings to be the same ? Do you know how DivX is coded ? Of course not. So I'll assume that by same settings you mean either "highest possible settings", in which case you're blaming a codec for actually having a wider range of speed/quality tradeoff available to the user, or "same quality", in which case you need to go forward with PSNR figures, or something else a bit objective.
Firstly, I said "if possible". The goal is to see wich can be better, so if a codec offers more useful options, let's use it. Otherwise, it would be like compare H.264 Baseline Profile with VP7 and claim VP7 is the winner over H.264 High Profile.....
Secondly no I don't know how it is coded, but I know it is based on the same MPEG-4 ASP specifications, that DivX uses 1 warpoint for GMC and 1 B-frame.

As for PSNR, I prefer to see wich is better with my eyes, not with a math comparison my eyes may not agree with.

[Manao]

that DivX uses 1 warpoint for GMC and 1 B-frame.No, it uses as much bframes as it wants ( that's what "adaptive multiple consecutive" means ).
As for PSNR, I prefer to see wich is better with my eyes, not with a math comparison my eyes may not agree withAs do I. Since, if you do the test VHQ 2 vs VHQ 4 you'll see that you don't see any difference. So the eyes sometimes aren't precise enough. PSNR, however, would tell you immediately the difference. And the huge advantage of PSNR is that it isn't biased. Everybody is biased ( though often not consiously ) toward XviD, because XviD is the "good guy". It's imho impossible to be objective when rating two codecs visually ( except by doing blind tests, which are a pain to organize ).

[xyloy]

No, it uses as much bframes as it wants ( that's what "adaptive multiple consecutive" means ).
Oops, you're right. :)

As for QPEL, it's my mistake, I should calm down when I do codec comparisons :D
I will re-do DivX vs. XviD comparison ASAP(as soon as I have my new HDD). ;)

[DivX_Luver]

Look to answer the original question DivX 6.2.5 is faster than Xvid on MOST systems. The reasons for this may vary but it is true. When encoding with DivX alwayse use the balanced setting when comparing to XviD at its default settings. And do a 2 pass encode with the same bitrate and quality settings. Also make sure that you use the DivX Home theatre profile in both DivX and XviD. This will ensure a FAIR test not perfect but fair. Then use your eyes to see which you think is better. You should also compare them with all post processing optimizations off and with the same decoder (I recomend FFDSHOW). If you have a Dual-core processor remember to enable advanced thread optimization in DivX and to set XviD's thread setting to 2 no more or it will create inaccurate results!!! And of corse as everyone else it depends on your system setup and the settings that you are using to encode your video!
HAPPY ENCODING : )
DivX_Luver

[Sagittaire]

Look to answer the original question DivX 6.2.5 is faster than Xvid on MOST systems

No, you can't say that. If you want say that the only way is to make speed = Fct (quality) at constant size. And the only way for quality is metric test here ...

[sarmin]

what do you think gives the best quality and speeds between these to conversions. cuz then i will be sure what codec to use between xvid and divx. i'm going to tell you my favorite settings then you can tell me wich codec is best depending on the settings. but if i have made a mistake in one setting then correct me.

Divx options:
encodeing mode= balanced
bidirectional codeing=adaptive multiple conseutive

i'm not useing qpel and gmc cuz my dvd can't play the files then.

noise redution=OFF
Quantozation=H.263 optimized
pyshovisual enhancements=off

xvid:

adaptive quantization=CHECked
quatiozton type=h.263
b-vops
max b-vops=2
quantizer ratio=1.50
quantizer offset=1.00

i havent changed something in infra-frames tuning,overflow treatment and curve compression so should i change something there?

motion search precsision=6 ultra high
vhq mode=2 limited search
use vhq for bframes to0=checked
use chroma motion =checked
trellis=checked

and all the min. quantizers in quantizer restrictions changed to 2 instead of 1.

so whats your opinion?

[foxyshadis]

You're comparing divx's balanced mode with an xvid mode that is something like "slowest". (But not quite "insane".) With such a disparity you'll get exactly that - much higher speed out of divx and somewhat higher quality out of xvid. You'd be better off deciding how fast you want it to run - a 1 hr encode, an 8 hr encode (and avisynth will dominate this encode time if you use more that light filtering), and set your settings to meet it.

Consider using custom matrices as described in the presets thread as well, they can offer much more noticeable quality improvements, if you use the right one for your needs, than simply slowing the encoder down.

[sarmin]

that's strange i converted a movie(135mb) and it took xvid 19min to convert. and divx 30 min. i used 2 pass encoding for both.

i used vdub-mpeag2

[sarmin]

where is the presets thread?

[Teegedeck]

Press Strg+f, type 'presets' and voila:

XviD presets - preliminary toughts and suggestions (http://forum.doom9.org/showthread.php?t=107897)
:)