[vorbis] Followup to Rok Papez's tests

damaged justice dj at frogfarm.org
Wed Jul 19 09:48:58 PDT 2000



I first downloaded Vorbis about a month ago, around the time Rok Papez posted
comparisons of Ogg and Lame [*]. Here's my much rougher test on a single file.

CPU: AMD K6-233
MP3: LAME 3.70, mp3blaster 2.0b6
Ogg: Vorbize 0.6 (Xiphophorus libVorbis I 20000718), ogg123-0.1

Data set: Ministry, "The Fall" (51,955,724 byte WAV)

* * *

Ogg Vorbis command line:

time vorbize -v -w /tmp/Ministry_TheFall.ogg /tmp/Ministry_TheFall.wav

Output:

Vorbize by Kenneth Arnold
-----------------------------
Encoding status:
Reading from /tmp/Ministry_TheFall.wav
Writing to /tmp/Ministry_TheFall.ogg
Input file information:
Samples per second: 44100
Bits per sample:    16
Channels:           2

50740 out of 50737 kB (100.0%) encoded*..........
Done.

Statistics:

Elapsed time:  795.13 sec.
File length:   294.53 sec.
Rate:          0.3704

real  13m15.273s
user  13m3.600s
sys   0m3.430s

* * *

LAME command line:

time lame -b 160 -h -k /tmp/Ministry_TheFall.wav /tmp/Ministry_TheFall.mp3

Output:

LAME version 3.83 (www.sulaco.org/mp3)
GPSYCHO: GPL psycho-acoustic and noise shaping model version 0.77.
Encoding /tmp/Ministry_TheFall.wav to /tmp/Ministry_TheFall.mp3
Encoding as 44.1 kHz 160 kbps stereo MPEG1 LayerIII ( 8.8x)  qval=2
    Frame          |  CPU/estimated  |  time/estimated | play/CPU |   ETA
 11275/ 11275(100%)| 0:05:21/ 0:05:21| 0:05:21/ 0:05:21|    0.9177| 0:00:00

real    5m21.174s
user    5m19.120s
sys     0m1.860s

Repeated @ 192kbps:
real    4m39.733s
user    4m37.680s
sys     0m1.800s

Repeated @ 192kbps with Variable Bit Rate (default quality setting, 4):
real    9m49.699s
user    9m37.580s
sys     0m2.220s

* * *

Results:
             Format        Size                     Encoding Time

             Original WAV: 51,955,724 bytes
       LAME .mp3 @160kbps:  5,891,135               5:21.174
       LAME .mp3 @192kbps:  7,069,362               4:39.733
 LAME .mp3@ VBR (avg 193):  7,091,746               9:49.699
          Ogg Vorbis .ogg:  7,056,175              13:15.273

* * *

Decoding:

'top' reports average CPU usage:

mp3 160: 3.7-4.9%
mp3 192: 4.1-5.1%
mp3 VBR: 4.9-5.8%
    ogg: 47-48%

* * *

Summary: 

While at first glance Vorbis appears to lag behind LAME in encoding speed, the
margin of advantage is considerably less at higher bitrates or when using
Variable Bit Rate, both of which are increasingly supported by software and
actually being used by users. Vorbis requires only slightly more CPU for
encoding, but the current decoder is very demanding; I didn't formally test
the plugin for OGG playback under XMMS, but did note that it also is very
taxing on the system.

One of the goals of Vorbis has been to achieve equal or better compression
than MP3, with better quality. In terms of file size, Vorbis does quite well,
running neck and neck in this area. Audio quality is a somewhat subjective
phenomena, but even my tin ears could detect notable improvement in Vorbis
compared to the last time I tried it (around early June); trebles stand out
much more clearly, and it sounds much less 'flat'. Unlike last time, I no
longer hear any significant difference between the formats. 

Overall, Vorbis appears to be succeeding in its attempt to produce a superior
alternative. The current implementation appears stable, and its only weakness
from a user standpoint is the high CPU cost of playback. Feature-wise, it's
nearly ready for prime time; real conversion will only really take off when it
is truly mindlessly convenient (widespread support of Vorbis in scripts, etc),
but the ready-for-use playback plugins are a big help in this area. I would
consider it ready for my own personal use at least as far as all of my future
encoding, but I don't think I'd invest the time to convert my already somewhat
large MP3 archive (unless I get that new CPU :).

Kudos to all involved. I'm greatly anticipating new developments.

-dj
 screaming in digital

* http://www.xiph.org/archives/vorbis/0122.html


-- 
http://dj.frogfarm.org/               ...for the best in unapproved information
"There is an email in which somebody says exactly that."
"Is that the one that said he was going to kill all the lawyers and judges?"
"No, it is a different one."     - MPAA v 2600, June 6 Protective Order Hearing

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.




More information about the Vorbis mailing list