Does composite video / audio in fact requires more CPU usage?

[Solid One]

When testing and researching about performance of N64 emulators on Raspberry PI 2, I stumbled across this wiki article: https://github.com/RetroPie/RetroPie-Setup/wiki/Optimization-for-Nintendo-64#notes-on-audio

Here, it says:

Note's on Audio
Audio
Use HDMI as composite requires more CPU usage.

This weekend, I was testing lr-glupen's performance on my Rpi2. I was on a house with only CRT televisors, and I used this composite cable for gameplay: https://www.modmypi.com/raspberry-pi/accessories/audio-cables/av-composite-cable-3.5mm-to-3-x-rca-3m

Those same tests were made on my home, playing some N64 games on my 39'' TV, using HDMI for gameplay.

I noticed that games on HDMI run smoother than on composite cable. I got a frame skip dropdown on games such as GoldenEye 007 and Duke Nukem Zero Hour, when playing on CRT TVs using composite video. This frame skip drop didn't happened on HDMI.

The same applies to PSP games running on PPSSPP, zdoom running Brutal Doom on 640x480, etc. Performance was lower on composite than on HDMI here too.

Some questions I'm wondering here:

1. Even using composite video running retroarch and emulation station on lower resolutions (720x480, I reckon), am I sure to affirm that composite video in fact is heavier than HDMI running on higher resolutions by default (1280x720, I guess)? I thought the contrary should happen because of resolution differences...
2. If Composite video indeed requires more CPU usage, how much is the difference of CPU usage between HDMI and Composite?
3. Is there a way to get the same HDMI performance on composite cable? I know I can use a HDMI -> RCA converter, but it'd be interesting to find another way, if possible.