Hi all,

Just curious, how quickly does Ogg Vorbis encode on your computers on average? I've found the following on my machine (AMD Athlon64 2800+, 1.8Ghz, DDR400, Socket 754)) -q8 with oggenc (Xiph.Org libVorbis I 20030909 (1.0.1))..


Values are the "rate" reported by OggEnc after encoding finishes.

32bit mode: ~8.5-9.0x.
64bit mode: ~16-17x.

For the 64bit test I was running Debian AMD64, and for the 32bit test, Debian Sid i386. I'm interested to see how other CPUs perform (on average).

FLAC sees a similar increase in speed between 32 and 64bit modes.

Thanks,
Lem

Impressive speedup

My good old AMD K6-2+ 500 MHz usually hardly manages to encode in realtime... ~0.9-1.1x.

did you use specific CPU optimized compiles ?

With oggenc 1.1RC1 (20040629) -q 6 --advanced-encode-option impulse_trigger_profile=3 --advanced-encode-option impulse_noisetune=-5:

Sole app running, in-focus: 7-7.5x
Typical multi-task environment, minimized: 3.5-4.5x

Intel Tualatin Celeron 1.4GHz, PC100, 440BX chipset

3.0x on my P3 500

7-9x on a Athlon XP1700+
0.9x on a K6-2 333 Laptop CPU

I'm not aware of any specific AMD64 optimisations, just what Debian is providing on their AMD64 mirror (Alioth). I assume it's just a recompile in 64bit mode. I'm not sure how to find out if it's optimised for AMD64 either..

Has anyone got an Intel EM64T CPU to try out 32bit vs 64bit OggEnc?

Just a little curious, are you using 32 bits Float or 64 bits Float ?

How do I find out?


I tried downloading the vorbis-tools source package from Alioth, doing ./configure, make and make install - I got the same encoding speed as the precompiled binary. I didn't see anything special in the GCC compile options (just -O2 and some architecture specific switch setting it to x86_64).

-O3 should increase the speed even further