I still don't get the manual placement and how it differs from the norm.
The physical design of a chip is like a giant connect-the-dots puzzle. In the olden days, when I first started doing chip designs that only had several thousand transistors, this was done by hand on giant sheets of plastic paper using colored pencils and erasers. A lead designer would break the problem into blocks, and other designers would figure out the best way to do each block (place each dot and figure the shortest line between dots with only so many overpasses over/under other wires allowed, etc.)
With millions, to the neighborhood of billions, of transistors, this became too big a puzzle for ordinary engineers to do by hand. So some giant computer program, a distant relative of the google map router, woud figure out the puzzle. But the result usually ends up looking like a big dense uniform blob. Look at the center of the GPUs for an example. Sort of blends everything all the same looking.
But, just a someone with intimate knowledge of their local neighborhood can sometimes figure out a better shortcut than the google maps route, the smartest most experienced chip designers can still sometimes figure out how to cleverly break the giant puzzle into smaller blocks, and tweak the insides of each block individually to have shorter wires. Look at the center of the CPUs to see this hand-work. The subdivision into smaller blocks and groups of blocks which can each be optimized separately by different specialists.
The A15 core is likely mostly computer placed and routed. This Apple A6 custom CPU core looks more like the old days when a lot of clever puzzle solving went into making really fast chips.