Comparisons between video codecs circulate constantly and frequently contradict each other. The disagreements are usually genuine, because the result depends on choices made before any encoding happens.
A codec is a standard, not a program
A codec defines how a decoder must interpret a bitstream. It does not define how an encoder should produce one, which leaves enormous room for implementations to differ.
Two encoders targeting the same standard can produce very different quality at the same bitrate, and a mature implementation of an older codec often beats an immature one of a newer codec.
A comparison is therefore always between specific encoders at specific settings, never between the standards themselves, however the headline is phrased.
Encoding effort is the hidden variable
Every encoder offers a scale between speed and compression efficiency. Running one at a fast preset and another at a slow preset produces a meaningless result.
Fair tests either fix the encoding time budget or fix the quality and compare resulting file sizes, and they state which approach was taken.
This matters practically because a live stream must encode faster than real time, while a catalog title can be encoded slowly. The right codec differs between those cases.
Source material changes the ranking
Codecs gain their advantages from specific tools, and those tools help some material far more than others.
- Grainy film scans behave unlike clean digital capture.
- Animation and screen recordings have large flat areas and hard edges.
- Fast sports footage stresses motion prediction rather than detail.
A test run on a handful of clips describes those clips. Extending the conclusion to a whole catalog assumes a similarity that usually does not exist.
Metrics measure different things
Automated quality scores are indispensable at scale, but each was designed with assumptions about what viewers notice, and they disagree in predictable ways.
Some reward preserved detail, others tolerate blurring but penalize blocking. An encoder tuned toward one metric can score well while looking worse to a person.
Serious comparisons report more than one metric and include viewing, because the number that matters is what an audience perceives rather than what a formula returns.
Decoding cost never appears in the chart
Bitrate savings are only half the equation. A newer codec typically demands more decoding work, which older devices cannot supply.
For a service, that translates into maintaining parallel encodes, which erases part of the storage saving and complicates every delivery decision.
A comparison that ignores this describes a technical result rather than a deployable one, and the gap between those two is where most real decisions are made.