xvid encode speed

[dlzinc]

Hello,

I'm wondering if there's some sort of problem with my machine/setup or is it xvid or .. ?

Anyways, I've been testing xvid rc4 koepi's build with the following settings changed from the defaults:

unrestricted profile, qpel, gmc, hvs-best-picture custom matrix
2 pass (defaults)
target bitrate 900kbps

motion search 6, vhq 4, chroma, cartoon
all min quants to 2
trellis

gmc seems to make my filesizes drop below the target bitrate more often (which is figure is a good thing sometimes?), while having it off and vhq1 practically halfs the encode time...

The real problem is:
I have two systems:
a----
AMD Athlon 1.4 GHz (not XP) 133MHz FSB
512MB SDR

b----
Intel Pentium 4 2.8 GHz w/o HT 533 quad pumped FSB
512MB DDR 266

both systems run WinXP Pro SP 1 with all the latest updates, clean install, using gknot to make the avs, and virtualdubmod 1.5.10.1 to actually do the encode. BOTH SYSTEMS take about the same amount of time to finish the encode (they finished 10 minutes apart, both taking about 30min for 1st pass, and about 2 hours for the second). That is just ... wrong. Is it me or ?
Is there anything I can do about it? (comments/suggestions?)

[JohnMK]

Are you using "Turbo" mode in XviD options? This dramatically increases Pass 1 speed.

[dlzinc]

hmm.. no i'm not using turbo mode
i tried it for a while while on second pass, it didn't appear to make much of a difference (timewise), i'll try it on first pass... but in a relative sense, first pass time doesn't matter too much anyways since it only takes ~30min whereas 2nd takes 2 hrs

but still... my athlon 1.4 took the same time to encode something using the exact same settings as my p4 2.8 ..... that just seems so wrong

[sysKin]

Originally posted by JohnMK
Are you using "Turbo" mode in XviD options? This dramatically increases Pass 1 speed. Eh, no? It doesn't change 1st pass' speed, at all.

[App0lly0n]

hi

wheres turbo. I updated the xvid version, and now i cant find it :(.
thanx

[celtic_druid]

Under advanced options.

[App0lly0n]

i still can find it

Advanced i have

tabs

GLOBAL , QUANTIZATION, TWO PASS, ALT CURVE, CREDITS, DEBUG

not under any

the other xvid setting werent as many. i installed 1.6 rip then XviD-1.0.3-20122004.exe

[celtic_druid]

Sounds to me like you are using an old devapi3 build.

[Koepi]

Yupp, sounds reasonable celtic. Deinstall all earlier builds (via the software control panel) and then reinstall the newest xvid build.

Regards
Koepi

[App0lly0n]

prob fixed

thanx guys -- it was koepis developer build

[SLA]

well is it normal that first pass is much faster than second pass? i get ~40-50fps@firstpass with my a64 and around 10-20@second pass? it doesnt matter if i use latest nic or koepi stable build, second pass is always slow.

[Koepi]

:search:

Simple question, simple answer. Yes, it's normal.

Cheers
Koepi

[Pen-Pen]

but nobody answered the OP's question : why are his two systems equally slow, when they shouldn't... sounds like a mistery to me too... (or like a good reason to buy AMD ^^)

[IgorC]

on my celeron 2 ghz, 1.0.3 and 1.1 (both celtic druid build) are fastest.

[Benihana]

I'm not sure why that's happening. At the very least, you'd expect a huge difference between render times, considering that your system A is a generation behind the P4, 133 FSB and SDR memory. Your system B should perform much better, well by just looking at the system setups anyways. I'm not sure either. Try doing a large task, one that would take at least 10 hours to complete the second pass, maybe the difference will be found there.

Then again, maybe your "default" 2 pass settings may be different on each machine. Look at those and see if you see any difference in them.

[MSlv]

I encoded Triple X last night to 1cd(700MB) XviD with AutoGK 1.85b in order to make a comparison between XviD and RV10. My question: can you tell me if my encode was fast or slow?

From autogk log:
[1/8/2005 10:59:54 PM] Source aspect ratio: 16:9
[1/8/2005 10:59:54 PM] Source resolution: 720x480
[1/8/2005 10:59:54 PM] Found NTSC source.
[1/8/2005 10:59:54 PM] Source seems to be pure FILM.
[1/8/2005 10:59:54 PM] Found 173490 frames
[1/8/2005 10:59:54 PM] Encoding audio.
[1/8/2005 11:11:18 PM] Forcing encoding with sharp matrix.
[1/8/2005 11:11:18 PM] Using bitrate spikes control.
[1/8/2005 11:11:18 PM] Picking up credits information.
[1/8/2005 11:11:18 PM] Credits start frame: 164220
[1/8/2005 11:11:18 PM] Credits quality percentage: 50
[1/8/2005 11:11:18 PM] Audio size: 115,287,056 bytes (109.95 Mb)
[1/8/2005 11:11:18 PM] Overhead: 6,939,648 bytes (6.62 Mb)
[1/8/2005 11:11:18 PM] Video size: 611,776,496 bytes (583.44 Mb)
[1/8/2005 11:11:18 PM] Running compressibility test.
...
[1/8/2005 11:16:50 PM] Duration was: 5 minutes 31 seconds
[1/8/2005 11:16:50 PM] Speed was: 26.16 fps.
[1/8/2005 11:16:50 PM] Compressibility percentage is: 32.20
[1/8/2005 11:16:50 PM] Using softer resizer.
[1/8/2005 11:16:50 PM] Chosen resolution is: 512x224 ( AR: 2.29 )
[1/8/2005 11:16:50 PM] Predicted comptest value is: 52.16
[1/8/2005 11:16:50 PM] Running first pass.
...
[1/9/2005 12:05:04 AM] Duration was: 48 minutes 13 seconds
[1/9/2005 12:05:04 AM] Speed was: 59.96 fps.
[1/9/2005 12:05:04 AM] Expected quality of first pass size: 46.69
[1/9/2005 12:05:04 AM] Running second pass.
[1/9/2005 1:08:49 AM] Duration was: 1 hour, 3 minutes 44 seconds
[1/9/2005 1:08:49 AM] Speed was: 45.36 fps.
[1/9/2005 1:08:49 AM] Job finished. Total time: 2 hours 11 minutes 10 seconds
/end_of_log

[jon.schaffer]

Originally posted by MSlv
can you tell me if my encode was fast or slow?

The value you should have bold is the encoding speed (in frames per second) [as we do not know the length of the movie]

59.96 fps (1st pass) and 45.36 fps (2nd) quite high speeds! The mean is around 52 fps... more than real-time (generally 25 fps). You have a nice hardware. I personally used to encode at 6 fps without any filter...

[MSlv]

the movie is 2h long
My specs:
AMD 64 3000+ @ ~2000 MHz
512 DDRAM
Gigabyte nForce MB

Oh, and is P4 better than AMD?

[Doom9]

I think hardware questions (encoding speed is one) belong in the hardware forum. Let's not clutter a codec forum with issues that do not concern the codec.

Encoding speed depend on the complexity of the source, codec used, settings used, filters used, and on many parts of your system: mainboard, ram, cpu, harddisk, and the settings for those (FSB speed, RAM speed, whether you have source on one hd and destination on another), etc.

You should also read some hardware sites on the subject. They'll tell you that when using CCE, a similarly clocked P4 using comparable hardware is faster, when using XviD it's the reverse. DivX also has a P4 affinity. But now you can see that the question if intel or AMD is faster also depends on what tools used, so it makes little sense to ask a global question like "which cpu is better".

Giving an fps rate, movie length, codec name (not even build), no hardware info and no info on what scripts were used (filters in an avisynth script make a world of difference) and asking if that's fast or not is simply and utterly useless and a waste of everybody's time and my bandwidth.

[MSlv]

sorry. Promise it ain't gonna happen again.

Oh, and the codec shootout rocked! Thanks for making the effort and sorry again for being stupid.
:o
oh, and sorry for your wasted bandwith and time. I know it must be expensive...


ps: am I expecting a ban any time soon?