H.264 Versus DivX in 2005: What an Early Compression Test Teaches

In 2005, affordable DVD writers were changing home video storage. A single-layer DVD held substantially more material than a CD, reducing the number of discs needed for an archive. DVD authoring software also made MPEG-2 workflows more accessible. At the same time, newer compression tools promised useful picture quality at much lower bitrates.

This historical adaptation follows a comparison by Filipp ?Finar? Kazakov, published in Computerra, issue 19, on May 24, 2005. It preserves the test setup and reported performance while correcting terminology and separating that era's hardware limits from present-day expectations. It is not a benchmark of current encoders.

Why the author investigated H.264

A late-2004 codec comparison on Doom9 drew attention to an Ateme H.264 implementation distributed in Nero's video software. The author decided to investigate it personally rather than accept the ranking as sufficient evidence.

H.264 itself was not a closed format exclusive to Nero. It is an international video coding standard, also known as MPEG-4 AVC; the ITU recommendation history includes a 2003 edition. The commercial encoder, its packaging, and its licensing were separate matters. This distinction matters whenever a product's brand name is confused with the compression standard it implements.

The software package and the practical barriers

The historical Nero package bundled several applications: Nero Vision Express for capture, basic editing, and disc authoring; ShowTime for playback; PhotoSnap for photo work; MediaHome for media distribution; and Recode for the encoding work of interest. The author had not tested MediaHome and did not offer a verified account of its capabilities.

The complaint was about packaging as much as technology. Downloading roughly 34 MB of bundled software was inconvenient on a slow modem connection when the user wanted only an encoder or decoder. The article also discussed the absence of a separate Nero DirectShow decoder and noted H.264 decoding support in ffdshow. These observations describe the software distribution environment of 2005, not a current download recommendation. The old trial-software link is not needed to understand the experiment.

DivX was selected as a familiar consumer-video reference point. The author acknowledged contemporary debates about Xvid and VP6 but chose a widely recognized workflow for the comparison. That selection did not establish a universal ranking of every codec available at the time.

Source material and encoding settings

The test used a short excerpt from a licensed Terminator 2 DVD containing both relatively static and active scenes. Both encoders produced 640 ? 480 output using two passes. The original material was described as progressive, so deinterlacing was disabled.

The author avoided additional advanced DivX adjustments to approximate a typical user workflow. That choice should be understood as a test condition, not proof that every optional setting would have had no effect.

Bitrate, resolution, and the missing frame-rate term

The experiment held output resolution constant and varied bitrate across five runs, from approximately 1,500 kbps down to 500 kbps. This made the effects of heavier compression easier to compare without simultaneously changing the number of output pixels.

The original discussion used ?bits per pixel,? but its numerical examples require a frame-rate term. For video, a useful normalized quantity is:

Bits per pixel per frame = video bitrate in bits per second ? (width ? height ? frames per second).

At 640 ? 480 and 25 fps, 1,500,000 ? (640 ? 480 ? 25) is approximately 0.195 bits per pixel per frame. At 500 kbps, it is approximately 0.065. Those calculations explain the article's rounded figures of 0.2 and 0.066; the excerpt does not independently specify the clip's frame rate. At another frame rate, the result changes.

The author treated 0.2 as a working starting point for DVD-to-DivX encoding. It is not a universal quality threshold. Scene motion, texture, encoder settings, and the viewing task all matter. The comments about avoiding very low resolutions or upscaling beyond DVD dimensions were likewise choices for this experiment, not general limits on useful video resolution.

Reported processing performance

On a Pentium 4 Northwood rated at 3.06 GHz and overclocked to about 3.33 GHz, the author reported roughly 6 frames per second for both H.264 passes. DivX reached about 11 fps in the first pass and 17 fps in the second. Raising H.264 to its maximum-quality setting reduced the reported rate to about 4 fps.

Those figures made two-pass encoding of a two-hour movie a lengthy task on that computer. They do not establish how long modern hardware or a hardware-assisted encoder would take. Two-pass offline encoding also differs from the continuous, real-time encoding expected of a surveillance camera.

Playback worked on the Pentium system even with strong decoder postprocessing. A second computer, a Celeron rated at 1.7 GHz and overclocked to 2.0 GHz, reportedly approached full CPU utilization without postprocessing. Decoding cost therefore mattered alongside file size.

Lessons for evaluating security video

The supplied excerpt does not include a complete scored quality comparison, so it cannot support an invented percentage advantage or a definitive visual winner. Its lasting value is methodological: keep the source and output conditions consistent, test a range of bitrates, examine motion as well as static detail, and measure both encoding and playback load.

עבור מצלמות אבטחה ו מערכות הקלטה, also confirm that recordings preserve the detail needed for the actual task. A smaller file is useful only if the resulting video remains usable and the complete viewing and export workflow supports it.