Xvid speed slowerthan Divx5 ?

[Wuntvor]

Hi!

Ive been reading a lot of messages stating that Xvid is faster than divx5 , but had the feeling this wasnt true (for me, that is) so I decided to test this.
The setup:
I used Virtual Dub 1.4.9 , and loaded an mpeg1 file (resolution 320x240 30 fps and 21 minutes long named Thumbwars) and set no sound and fast recompress.
Then I tested different settings in DivX5 and Xvid, but i used a set quantizer=4 for all the tests, to be as fair as possible.
In Divx I used "Slowest" , in Xvid i used ME 5 and 6 just to see the difference ..
anyway these are the times I got:

divx5 5:35
divx5 with b-frames 7:55
divx5bqp with bframes and quarterpel 13:18

xvid ME5 h263-qaunt 5:45
xvid ME6 h263-quant 6:04
xvid ME6 h263-quant with lumamask 6:40

xvid ME5 mpeg-quant 5:57
xvid ME6 mpeg-quant 6:35
xvid ME6 mpeg-quant with lumamask 6:52


Now 2 thins are interesting for me, one is that Divx5 i faster than xvid with h263 and ME 5 , instead of the other way around.
The other very interesting thing us that all the Xvid files were smaller than all the divx files ;-)

Now it doesnt really matter , I will continue to use Xvid , instead of Divx5 , even if its slowewr, I just wonder why everybody else gets faster then divx results? Am I missing something?

Since this is my first message i would like to express my grattitude to Nic and trbarry, for speeding up the DirectShow filter ;-)
And to Isibar , Nic, Koepi , Gruel, h and all the rest , Good Job!!!

Oh, my machine is an Athlon Tbird 1200 with 384 MB, running Win2k is that has any importence.

regards
/wuntvor

Edit:
Forgot to say that i used the 17 Buiold from Nics site, so it should be the "latest fastest" version av Xvid...

[rui]

Well, divx5 pro can only compete with xvid if it uses b-frames (IMHO it's even a tinny little bit superior, but thats only my opinion, not the general opinion).

So, judging by your test, i can say that, at equal or almost equal quality, xvid is faster.

[Wuntvor]

Hehe!
I do not question the quality of Xvid, but using the same quantizer , Divx is faster in my test, others on this board has posted messages claiming the opposite, i just want to know if I miss something.

regards
/wuntvor

[Hanty]

Am I the only one that feel the continual debate on how many nanoseconds faster each codec is than the other is getting old? Using closest possible equal settings (which is impossible to judge since they are radically different)with the same filters, on fresh installed OS machines, with no other applications running memory resident + repeated test on different processors with different rams then you might discover that Xvid and DivX5 swings back and forth 15 min of each other.

Now if the same figure was 2+ hours then I might consider letting the issue into a debate which is about quality and not speed. But since we won't get those figures I'd say the speed issues are trivialities.

[canadian_fbi]

Originally posted by Hanty
Am I the only one that feel the continual debate on how many nanoseconds faster each codec is than the other is getting old?

...

Now if the same figure was 2+ hours then I might consider letting the issue into a debate which is about quality and not speed. But since we won't get those figures I'd say the speed issues are trivialities.
that's the same reason i'm happy with my old 8x cd-r drive. because going from 10 to 5 minutes for a burn isn't that important for me. similarly, if an encode takes three hours instead of three hours fifteen minutes or something, it's not a big deal. but i have an athlon xp 1700+. on my old athlon 500, maybe that'd be a 45 minute difference, maybe an hour in some cases, and that can be significant.

my personal opinion is that speed is a very distant second to quality and that you're right - i don't care too much if xvid shaves ten seconds off an encode, but i do care if it looks better. this is somewhat, but not very much, due to the luxury of a fast processor. but with the price of athlon xps nowadays, it's a lot easier to get significantly faster encodes than it is to get significantly higher quality ones :)

[athos]

on my machine (p4 1.3ghz) Koepis latest (17-2) was notably faster than Nic's latest when encoding. I run Windows XP and have Visual Studio .NET installed (although Nic's installer installs the necessary VS7 files anyway).

[Wuntvor]

I just want to say that of course quality is more important than speed, its just that when everybody else has better encoding times you wonder why, and if you make a misstake? Now its seems that one of the builds is faster than the other (Koepis, I used Nics), so that might be the answer.

Regards
/Wuntvor

[Nic]

How strange,

VC7 is quite weird, it runs quicker with some settings & slower with others, I went back to VC6, but then everyone complained that my vc7 compile was faster.....

Ill have to do some real tests & find the best overall settings (either using VC6 or VC7)....My build on my Duron was definitely faster....how weird, but thanks for the info :)

Cheers,
-Nic

[Koepi]

You installed my build and compared? *giggle* ;)

SCNR,
Koepi

[Nic]

What do you think im doing, while im writing this :D

-Nic

ps
Ive just forced a VC6 build to be faster than my VC7 or even Koepi's.....hmmm :confused:

[athos]

Originally posted by Koepi
You installed my build and compared? *giggle* ;)

SCNR,
Koepi

Hehe, well I usually run your build, but then I tried Nic's and it didnt feel as fast, so I reinstalled yours to compare. Note though, that this comparison was made on march 17th, and there were several speed updates in the CVS this day, so maybe your two builds were not from the same cvs version?

So anyway, do you (koepi) use any special settings for speed when optimizing, or should your builds be eqivalent to uManiacs (i assume he uses the "default" settings)?

[Nic]

Well to clear this nonsense up :) :

My build did my test clip (1pass) in: 2mins:45sec
Koepi's build did it in: 2mins:30secs (!!!WTF!)
My present build up on my site (just uploaded) does it in: 2mins:25secs (now thats better!)

Sorry it was my fault.....I set up the optimisations wrong. (I always custom mine)

Cheers,
-Nic

[Wuntvor]

Nic, you're the Man!
Hope nobody still feels this thread was stupid.. ;-)

regards
/Wuntvor

[Koepi]

This thread is still nonsense.

Athos, I too use custom optimisation settings.

Congrats Nic for squeezing out a little more BUT: so much to "the fastest ever" :P don't pronmise such things when you don't keep them ;)

Did you base your new build on the new CVS?
If you'd look into the changelog you could see why there probably is a small speedup (should be more releveant on a p4).

Regards,
Koepi

[Nic]

The weirdest thing was, that it was the fastest ever ??? But then I miss set something, apologies for misleading anyone....but at least now it is true :)

So the thread did have a use :)

Cheers,
-Nic