MPEG-2 Was Good Enough Because the Disc Was Big

Four generations, four arguments · Medium

The DVD-Video specification could afford MPEG-2's inefficiency because it had 4.7 gigabytes to spend. The constraint moved when the disc did.

A Format Built Around a Given Size

MPEG-2 arrived in 1995 as the compression engine for broadcast digital television and, soon after, for DVD-Video. The Moving Picture Experts Group had designed it to replace analogue transmission and VHS tape with something that held up at broadcast resolution, and for those purposes it was exactly sufficient. It was not, by any modern measure, efficient. It left a great deal of information in the bitstream that a smarter encoder would have discarded or summarised, and it made no attempt to model the way human vision weights different parts of an image differently. What it had going for it was that the hardware to decode it was cheap to build, the standard was unambiguous, and the disc it lived on was large enough that the inefficiency didn't matter.

A single-layer DVD holds 4.7 gigabytes. A two-hour film encoded at an average video bitrate well below the DVD-Video specification's 9.8 Mbit/s maximum fits with room for audio tracks, subtitles, menus and chapter markers. The disc was sized to the format, or the format was sized to the disc; either way, they matched. MPEG-2's redundancy was absorbed by the capacity, and nobody encoding a studio release had reason to wish for a more parsimonious codec. The inefficiency was simply a cost that the medium had already paid for.

This is why the DVD era produced no competitive pressure on the codec itself. Studios licensed MPEG-2 encoders, pressed discs and shipped product. The standard held for a decade as the dominant home video format without any meaningful challenge from more efficient alternatives, because the bottleneck was never the codec — it was manufacturing, authoring and distribution logistics. MPEG-2's patent holders, collected through MPEG LA's patent pool, took royalties on every disc player sold, and the whole system was stable.

When the Disc Shrank

The constraint changed the moment the target storage shrank. A CD-R holds roughly 700 megabytes — about one-seventh of a single-layer DVD. Fitting a feature film onto one required not just higher compression ratios but a qualitatively different approach to what information was worth keeping. MPEG-2 could reach those ratios, but the quality at CD-R bitrates was poor enough to be commercially unacceptable, and the decode complexity didn't decrease as bitrate fell: you paid the full processing cost and got a visibly degraded picture.

MPEG-4 Part 2 — the standard that underlay the legitimate DivX 4 releases, which followed the hacked DivX ;-) 3.11 Alpha build — addressed this directly. It modelled inter-frame motion more precisely, used variable-length coding more aggressively, and could allocate bits where they were perceptually needed rather than spreading them evenly. At the bitrates that fit on a CD-R, the difference over MPEG-2 was visible. The codec mattered more when the disc was smaller.

The same logic applied in every direction. When streaming became the delivery mechanism, storage gave way to bandwidth as the binding constraint, and the question was no longer what fills a disc but what a given connection can carry at acceptable quality. H.264, standardised in 2003, achieved roughly twice the compression efficiency of MPEG-2 at equivalent quality — a ratio that meant the difference between a stalled buffer and a watchable stream on a mid-2000s broadband connection. HEVC pushed further still, though its fragmented patent licensing slowed the transition. Each generation justified itself not by being new but by facing a constraint MPEG-2 had never been asked to solve.

MPEG-2 is still in active use. Broadcast television infrastructure has retained it longer than consumer formats, partly because re-equipping transmission systems is expensive and partly because broadcast bandwidth, while not infinite, is allocated rather than competed for. The DVB standards body maintains specifications that still name MPEG-2 as a supported profile. For a signal that fills a fixed pipe at a guaranteed bitrate, the argument for upgrading is weaker than it looks. MPEG-2 was never bad engineering. It was engineering calibrated to a constraint that, for a substantial part of the industry, has not yet decisively changed.