Bit level vs byte level

Am I correct in believing that video game consoles such as the NES used graphics and instructions at the bit level?  Why don't modern operating systems use instructions that are only a few bits rather than bytes?  You'd be doing the same operations at 1/8 the scale.

>>1
The NES's CPU was a 6502-compatible, meaning it addressed everything (registers and memory) as collections of eight-bit bytes.  All opcodes in the 6502 are eight bits.  You have bitwise operators (AND, ORA, EOR) but they require full eight-bit operands.

The only instructions that operate on a bitwise level in 6502 are the set and clear flag instructions (SED, CLD, SEC, CLC, SEI, CLI, CLV)

The NES PPU (graphics) and sound hardware have registers wherein each of the bits within are dedicated to specific functions.  Accessing specific bits require the above bitwise operations.  However, the graphics data itself are held in normal 8-bit RAM and on 8-bit boundaries like all other memory and registers of the CPU.  With the exception of one area of memory in the NES called the attribute table, that is a collection of 2-bit values... but again, you need bitwise operations here.

Effciency in processing CPU instructions, for CPU design, lies in making the CPU instructions the same number of bits, and such principles apply in architectures such as MIPS, PPC, ARM, etc.