All things being equally unoptimized, is it more efficient to move many large parallax layers* with position: absolute; and left/top ... or is it better to use CSS3 transforms and translate (or translate3d) ... or is HTML5 canvas best? *Sprites have their own test, and parallax is only used here because it is a real world example.
| normal | w/ translateZ(0) | scaled 110% | w/ translateZ(0) | |
|---|---|---|---|---|
| left/top | -- | -- | -- | -- |
| translate | -- | -- | -- | -- |
| translate3d | -- | -- | -- | -- |
| background-position | -- | -- | -- | -- |
| canvas | -- | -- | -- | -- |