First pass with DivX 5.0 is off and running!

[dragoman]

Alrighty.....

First pass started with DivX 5.0 codec in Virtual Dub.

Movie is Independence Day Special Edition,

settings are the standard for the codec (max quan 12, min 2), bitrate 1000kps (assumed DivX 5.0 is similar to DivX 4.0 in that 1k=1024bits)

Will keep you updated...

dragoman

[Dr. Nick]

isn't it 1k=1024bits in divx 3.11 and 1k=1000bits in divx 4.12?

[dragoman]

OOops!!

*blush*

You are right, senor!!! However, I still did it right since I used the bitrate calculator in Gordiaknot for DivX 4.

Thanks for catching that......

dragoman

[Dr. Nick]

Sorry...didn't mean to be the grammer nerd!

just that it could have seriously screwed your test...

i unfortunatly cannot test er out, school and all, so i gotta make sure you do it right!

[Emp3r0r]

Can you update us on what kind of speed you are getting... fps?

Also state system speed and encoding specifics please. Just want to know if it is faster?

[JetSter735180]

I'll started a DivX5 project. Actually, I took Star Trek: First Contact, wich was my first DivX3.11 rip.

First thing I can says that it is fast. With DivX3.11, i got 9 fps, with Divx 4.12, i got 12 to 15 fps, and with DivX5 I get 19 to 21 fps. WOW!

My encoding should be finish ant 3 hours. I hope the quality is better.

[dragoman]

22% done with Independence Day Special Edition

Total movie size is 220865 frames.

I have been averaging about 13-15 fps.

I am running a AMD Athlon Thunderbird 800 with 448 MB SDRAM (PC133)

No other programs are running in background (except my ftp, that never goes off :) )

Seems like it is a little faster than DivX 4.xx, which is good because with the inclusion of B-frames I thought it would be slower, despite the hype.

And reading the documentation, it seems like the second pass is where the speed really goes up, since the codec uses the motion-calculations or something (don't quote me) from the first pass to increase the speed 30% over DivX 4 (according to the website).

I'll keep you posted.....seems like it is working good.

I'll encode a second time with DivX 5.0 Pro when this is done in order to make a speed comparison...hopefully, Pro is faster (website claims to have attained twice-real time with Pro version on a fast machine).

dragoman

[saVe]

@dragoman: for me divx 5.0 pro is slower! don't know if i got that right, but are you encoding with the standard version? according to divx.com b-frames are only available in the pro version, so maybe that's why you get more fps than with divx 4.12?

correct me if i'm wrong!

[Acaila]

I'm using the pro version. About speed:

1-pass quantizer mode: no speed difference over DivX4
2-pass 1st pass: no speed difference over DivX4
2-pass 2nd pass: about 20-30% speed increase over DivX4

Total first impression:
Quality is comparable with XviD
Post-processing works nicely
Much smaller filesize than Xvid at same bitrate

You're all using GKnot to calculate bitrate, but how accurate is that going to be with all the new motion calculations and B-frames? I don't think GKnot can do it correctly.

And where the %$@%#! is size prediction? All these great features and they still left that out! Unbelievable!

[dragoman]

Hi,

Just cracked 55%. Still sitting around 13-15kps fps in Vdub first pass. Scuttlebut from around the board and other boards on the net says that the 2nd pass is where you'll see the improvement.

Haven't tried DivX pro yet, still working on the original DivX 5.0 (funny to say that, a codec that was released hours ago :) )

Anyway, @acaila - have you encoded whole movies, or just short clips. For me it makes a difference in how the codec performs, a full movie-length test is what I am interested in.

Still you are helpful, thank you.

You might be right about Gknot, at least with the Pro version. I believe that the regular DivX 5.0 codec performs similar to 4.xx, at least when it comes to filesize, so perhaps using Gknot to set bitrate would work for DivX 5.0, but not for Pro version.

Anyway I will be letting everyone know soon enough, I am encoding Independence Day Special Edition, which is about 2 1/2 hrs long.

dragoman

[Evaldas]

I did a small test. Encoded 10000 frames from Shawshank Redemption, start frame 160000 with xvid and divx5 pro, resolution 640x352.

XviD (build 28/02/2002)
mode - 1-pass quantizer, quantizer 2
motion search precision - 6 Ultra High
smooth quantizer fluctuation - on
luminance masking - off

DivX5 Pro
mode - 1-pass quality based, quantizer 2 (100%)
scene change treshold - default (50%)
use Quarter Pixel - on
use GMC - on
use bidirectional encoding - on
psuchovisual enhancements - off
pre processing source - off
performance/quality - slowest

XviD final size - 93.8 MB
DivX5 final size - 63.5 MB

DivX5 has a little bit less small details, but looks good.
So I'll definetely switch to DivX5.

[gldblade]

>Much smaller filesize than Xvid at same bitrate

Um... That's not right. Average bitrate is based on filesize so if the filesize is smaller, the bitrate is smaller as well.

>You're all using GKnot to calculate bitrate, but how accurate is that going to be with all the new motion calculations and B-frames? I don't think GKnot can do it correctly.

GKnot doesn't deal with motion calculation and B-frames, especially not with DivX4/5 since it can't even edit the log files. All GKnot does is give out a bitrate to the codec and it's up to the codec to meet that average bitrate.

[Top_Cat]

Just running my first test, one pass to see the quality difference. It seems to me that 5.0 is "slightly" quicker than 4.12 by about .20 fps on my system! so about the same then :)

Running the Pro version with ads :( (guess there will be a non-ad version to download from a few warez type sites soon!)

Has anyone tried the "simple deinterlacing" is it the same as the one that was with 4.12? Just I encode a lot of interlaced stuff and would like to use xmpeg but that does not do deinterlacing yet :(

Thanks, will post my subjective views of quality differences soon

Top Cat

[dragoman]

Um... That's not right. Average bitrate is based on filesize so if the filesize is smaller, the bitrate is smaller as well.

I think what he was trying to say was that for the same quality, the filesize is much smaller.

This is what is claimed by Divx.com website....

dragoman

62% and still going.....

[oddball]

Smaller filesizes are all well and good but if I am encoding for 1 or 2 CD's I want it to use ALL the available space! So if I want a 700MB file (with audio) I want a 700MB file! Not a 650MB file! More bits = more quality no matter what way you look at it.

[dragoman]

Okay, first pass of Indepdence Day Special Edition is done.

Time taken was 5hr, 20min. Keep in mind this is 2hr, 34min movie!!

The reason I am doing this is to see how the codec performs over a long movie-length video clip. It is my personal experience that the codecs are truly made for movie-length clips and therefore you can't really evaluate them until viewed in this manner.

But that's just my opinion...


Average framerate was 12fps.

Encoding parameters:

DivX 5.0 regular codec

Resized to 640x272, soft bicubic filter

Max Quant. 2, min 12

1000kps bitrate.

I am running an Athlon Thunderbird 800 with 448 MB of RAM

Starting 2nd pass now....

dragoman

[oddball]

Soft bicubic? Eeew!

[dragoman]

@oddball

Just my personal preference. With a long movie I don't like the artificing I get around edges with a sharper resize filter.

To each his own, right?

dragoman

[Helmut Hamburger]

dragoman,

i agree about the soft bicubic for divx4

but for divx5 u can move up to nuetral and it still looks excellent