I'm afraid it's not exactly the truth.
You may take a look at the test project
https://snap.berkeley.edu/snap/snap.html#present:Username=dardoro&ProjectName=MIcro%20lag%20test&editMode&noRun
Exactly the same amount of work - 60 iterations of "move(8)" for 400 sprites is performed in three ways.
Simple loop initiated by broadcast took 1-1.5sm as expected.
The same loop started by "launch" for each sprite took 5-7 times longer.
Iterations "powered" by broadcast tick are 10 times longer.
Amount of graphic work is exactly the same only control structures differ. Also difference in sprites movement is astonishing.