Media Encoding

Traditionally, Media Encoding is one of the strong suits of the P4 line. We use three of the more popular media encoding tools out there for our testing - LAME, DivX and Xvid. LAME is the defacto standard for high quality mp3 encoding while DivX and Xvid are two popular video encoding codecs.

Encoding time is in seconds so lower is better. Each processor was put to the task of encoding the soundtrack to Attack of the Clones roughly a 760 megabyte file into a 133 megabyte mp3. LAME 3.90.3 was run from the command prompt with the --alt-preset extreme switch which generates VBR mp3 files of roughly 250 kb/s. The P4 does not give an inch to the Athlon 64 here; both the 3.0 and the 3.2 did extremely well. There is a large improvement between the 3400+ and the 3200+ but not enough to bring it within range of the P4.

Both the DivX and Xvid encodings were done with the Gordian Knot package with the settings recommended from the Doom9.org encoding guides. The Gordian Knot rippack 0.28.7 & Gordian Knot Codec Pack 1.6 were the encoding tools used. Note that we did not encode sound as we thought the LAME mp3 encoding test was sufficient and compressing sound here is a bit redundant. The film clips used were off BMW's The Hire DVD, for the DivX test, The Chosen was used and for the Xvid test we chose Powder Keg. The Chosen is roughly a 11 minute clip while Powder Keg was closer to 14 minutes. For both clips, the output size of the file was 150 megs and the video was encoded in two passes.

As with the LAME test, the P4 pulls ahead of both AMD chips in DivX encoding. Times are significantly higher on both the 3400+ and the 3200+ although the encoding time between the 3200+ and 3400+ are indeed much improved. The standard quality setting was used in the DivX test.

The tables are turned in the open source Xvid encoding test. Encoding times were significantly higher for a clip that was just slightly longer. The 3400+ is over a minute faster than the P4 3.2 and although the time gap is roughly the same as the DivX test, the encoding process was about 24 minutes so the percentage gap between the 3400+ and the 3.2 are not as large as they were in the previous test