Hello All,
I realize that I am using pre-release code. I realize that the
bitstream format will change in a couple of days. I realize that at
this point comparisons between MP3 and Vorbis still don't mean much.
BUT I couldn't resist. I ripped a random track off of one of my CDs
and did some comparisons between Vorbis and MP3. I thought you guys
might be interested in the results.
Info:
MP3 encoder was notlame using VBR encoding. Minimum bitrate
was 112KB/s. Maximum bitrate was 256 KB/s.
OGG encoder was sample_encoder from CVS source. XMMS reports
bitrate as 140KB/s.
Wave File Size: 52,454,348 byes
Wave Play Time: 4:57 minutes
MP3 Encode Time: 16:10 minutes
MP3 File Size: 5,848,223 bytes
Playback CPU Usage: 3%-4%
OGG Encode Time: 9:31 minutes
OGG File Size: 5,226,262 bytes
Playback CPU Usage: 44%-50%
Stats:
OGG encode 41% faster than MP3 encode
OGG file 11% smaller than MP3 file size
OGG plays with 92% more CPU time than MP3
MP3 achieves 9:1 compression ratio
OGG achieves 10:1 compression ratio
Summary: OGG looks and sounds good. File size is smaller than
similiar sizes resulting from the method with which I usually encode
MP3s. CPU time for playing is much, much too high. There may be
inefficiencies in the XMMS plugin or in the libvorbis library. I do
not know which to blame.
Is the efficiency playing Vorbis/.ogg files expected to improve? Is
that just an artifact of being pre-release, unoptimized code?
later,
joseph
P.S. I compiled everything using the default Makefiles (produced after
running ./configure). Compiler is gcc 2.95.2, system is Debian
GNU/Linux unstable (woody).
--
the "LaterDude"
ICQ: 52640402
martinja@ice-works.com
http://www.ice-works.com/personal/LaterDude/index.html
All opinions expressed are my own and not necessarily those of
my employer unless otherwise noted.
--- >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@xiph.org'
containing only the word 'unsubscribe' in the body. No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.
>CPU time for playing is much, much too high. There may be >inefficiencies in the XMMS plugin or in the libvorbis library. I do >not know which to blame. > >Is the efficiency playing Vorbis/.ogg files expected to improve? Is >that just an artifact of being pre-release, unoptimized code? >The XMMS plugin isn't particularly inefficient. Libvorbis is clean, working code - but it hasn't been optimised at all yet, really. Functionality before speed! On the other hand, 50% cpu seems overly high. This machine takes around 4-5% cpu to decode an mp3 (similar to what you were seeing). Right now, vorbis comes in at around 25% cpu time - still not great, but better than what you're seeing, and quite tolerable. I was seeing around 50% cpu usage with a debug build, full optimisation cut this almost in half. The decoding complexity shouldn't increase in the future, as I understand things, so this should be an upper limit on what decoding will take. There are a number of obvious places which could be optimised - but it really isn't time for that right now. At this point, clarity and completeness are far more important. Eventually, an optimised decoder should be of comparable speed to an mp3 decoder, or at least not much slower. Michael --- >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@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
> Hello All, > I realize that I am using pre-release code. I realize that the > bitstream format will change in a couple of days. I realize that at > this point comparisons between MP3 and Vorbis still don't mean much.OK. At this point, you're entitled to be curious.> BUT I couldn't resist. I ripped a random track off of one of my CDs > and did some comparisons between Vorbis and MP3. I thought you guys > might be interested in the results.[..]> Summary: OGG looks and sounds good. File size is smaller than > similiar sizes resulting from the method with which I usually encode > MP3s. CPU time for playing is much, much too high. There may be > inefficiencies in the XMMS plugin or in the libvorbis library. I do > not know which to blame.Partly libvorbis and partly the 'wrong tool' effect. Top and other system utils undersample and as such a process can consistently slip in under the radar, so to speak. I'd be a bit surprised if MPEG playback is really only 3% or so.... run something else mildly processor intensive at the same time and I think you'll see the mp3 playback usage mysteriously jump. A better way to measure is to do 'fastest possible decode' to /dev/null or such and compare run times. That said, libvorbis playback is currently poorly optimized and way too slow.> Is the efficiency playing Vorbis/.ogg files expected to improve? Is > that just an artifact of being pre-release, unoptimized code?Yes, entirely. Monty --- >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@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.