#low-precision

Question 1: Single choice

Going from 8-bit to 16-bit integer arithmetic takes roughly four times as much silicon for the multipliers, not twice as much. Why?

Twice the bits double the circuit, and the wider registers that feed it double it once more1
Wider numbers need a four times higher clock rate to finish a multiplication in one cycle2
An $n$ -bit multiplier forms $n \times n$ one-bit products, so its area grows like $n^2$ 3
16-bit numbers need floating-point exponent logic that plain 8-bit integers can do entirely without4