Is Divx 5 better?

[BigKing]

Hi!

Is Divx 5 realy better then 4.xx ? Or is it only a new number?

[yosemite]

i'd say: test it yourself!

then u can compare it ;)

[Uli]

Originally posted by yosemite
i'd say: test it yourself!

Right answer to a lame question ;)

greetz, Uli

[LotionBoy]

well it has to be better, because the number is higher. Just like the Pentium 3 and Pentium 4. :-)

LotionBoy

PS: in all seriousness, however, Divx5 is probably better than Divx4, because even if the quality isn't lots better, the encoding speed is.

[theReal]

I don't think the question is that stupid. Some people have lives apart from divx, not everybody has the time to do week-long encoding tests ;)

With b-frames you can get files that are the same quality but are 30% smaller.
...this sounds like a commercial, but it's true for what I have tested.
On the other hand, you must know certain things first if you don't want to end up with shitty movies: don't use any of the new features except b-frames! Only use video widths divisible by 16!
For reliability and stability reasons (i.e. for live capturing) I'd recommend to install dvix5 only in addition to divx 4. Description how to do that is here: http://forum.doom9.org/showthread.php?s=&threadid=19300&highlight=divx.dll

[3D]

I would say from some of my tests that 5.0 is better than 4.12 just in what way seems to be the question. Image quality is only slightly better with high bitrates (I have not tested low bitrates) and not even noticable in most scenes. The big improvement seems to be file sizes which are considerably smaller with the use of b-frames (4.12 file for 2 hr movie @1172 kbs/s 928MB, for 5.0 w/ b-frames @ 1172 829MB.) The GMC feature will also shave some MB off but only an additional 1-3% and at the risk of shit frames.

People have stated that 5.0 is faster but what I have noticed that with b-frames and/or GMC enabled most (not all) of the speed gain over 4.12 is lost to these extra features.

Overall I would say that divx 5.0 is a vast improvement for 1 CD rips but 2 CD rips will not benefit much from the new codec except for a smaller file size (but not enough to make it a 1 CD rip if you use AC3 sound). Right now if you share your creations with a number of people then I would stick with 4.12, 3.11 or xvid because there are still editing and playback issues that are being ironed out and you would want your shared files to be as universal as possible.

[DivxBr]

All this comments about b-frames in Divx 5.0 making smaller files is making me confused.
The bitrate we encode the movie, is related to the data stream that goes out the encoder, so if with the same bit rate the Divx 5.0 codec make a smaller file, the only thing this means to me is IT ISN´T RESPECTING THE BITRATE SETTING. If the file size is smaller, so it is the bitrate. :mad:
The real question must be: THE QUALITY with the same (real) bitrate / filesize is better with b-frames?
;)

[dragoman]

Hi,

I think you need to review the documentation on two-pass encoding.

Two pass VBR (Variable bit rate) is only considering the average bitrate when you input your desired bitrate. This means the codec will do it's best to have the average bitrate come out to the one you specified.

However, the actual determination is made frame-by-frame with the use of something called quantizers (again, if you are unfamiliar with this term review the documentation).

If a movie is so compressible that a quantizer of 2 (max possible) is used in every frame and the average bitrate is still less than your desired average, then there is nothing you can do. Increasing your desired bitrate will have no effect upon the filesize or quality (same as encoding CBR with 100% quality setting).

When a movie is less compressible, that is when the average bitrate setting becomes relevant to the codec. The codec must choose a quantizer setting between established settings (usually 2-12) to determine how many bits to dedicate to a certain frame. Obviously this results is a loss of quality during some portions of the movie.

So setting an arbitrary bitrate will not automatically get you a desired filesize when using a two-pass vbr method. The only time it will is by using a cbr method.

Incidentally, a good way to determine a good bitrate setting is to perform a cbr encode using quant=2, or 100% quality. The filesize you get will be the max filesize. Then look at the average bitrate of this file, and you can adjust it down to fit your desired filesize, should the first encode be too large.

dragoman

[DivxBr]

I know the problem of reaching the limit quantizer of 2, but I don´t think it could be the problem with a 2 hour movie with 1200 kbps (except with a too low resolution, or a too little action movie ;) ).
When using the CQ method, I agree with the proposition of getting less file size with the same quality setting for a better codec, but not with the CBR of VBR method (exception made for reaching the quantizer limit).

[theReal]

Divx5 with b-frames absolutely respects the bitrate in 2-pass encoding. The filesize calculator in Gknot told me my 61min 56sec DV-video is going to be 933MB with a bitrate of 2100 - the actual filesize is now 932MB. This is what I call perfect, even closer than 4.12 has ever been.

Using a fixed quantizer (=fixed quality setting), filesize will of course be lower using b-frames, as well as when you want to encode a very compressible movie with a low resolution at a very high bitrate (that is: your desired filesize is above 100% of first pass compressibility). This is not a bug, but a feature - when you can reach maximum quality at a smaller filesize, perfect!

I don't think the maximum quality of Divx 4 can be higher than that of DivX 5 no matter if you can achieve bigger files in some cases. I encoded some smaller sample videos with fixed quantizers (quality=100%) in Divx 4.12 as well as in Divx 5 using b-frames. The divx 5 quality was a little better in my opinion, maybe the same, but it was under no circumstances lower. The filesize was lower, though.

[psychoelf]

Actually the Pentium III is a better processor then the Pentium IV.
The only gain is that the Pentium IV runs at higher clock speeds, but the Pentium III does more bang per cycle.

I'd still recommend AMD :D though, cuz paying $100+ more for a chip that benches about the same isnt worth it.




Wouldnt it be nice if we could DivX our luggage?
Same content, smaller size, while only losing maybe one sock.

[rui]

Regarding the question BigKing made, i totally agree with this:

Originally posted by 3D
Overall I would say that divx 5.0 is a vast improvement for 1 CD rips but 2 CD rips will not benefit much from the new codec except for a smaller file size (but not enough to make it a 1 CD rip if you use AC3 sound)

[3D]

I totally agree with Rui and theReal. As I stated above the quality of 4.12 and 5.0 are virually the same, there is some small improvements that I have seen but only in certain scenes and the quality is still only "slightly" better.

Divx 5.0 in my opinion is pretty much 4.12 with the pro options added and that is all. This new codec has not resolved the macroblock issue with solid colors which IMHO should have been a top priority fix for this codec, but it wasn't... The rip the I quoted the fize sizes from earlier had only one scence with these blocks and no matter what bitrate I used the blocks were still there and the quants were maxed out at 2. QPel should help this but we all know that it can cause shit frames...


DivxBr - Just to note: I have seen the quant maxed out a number of times for a 2 hr movie (at least for a 2CD rip). The filesizes that I mentioned earlier were from a rip of "The Score" (no credits)it maxed out at 1172 (or somewhere in there) as all the frames were quant 2.

[movmasty]

divx5 is bet.........a

[chemmajik]

THG shows the pro 5 version running faster on a AMD, so either its the improved SSE or floating point compared to 4.12.