I don't have time to correct all the errors in the parent post. So very briefly:
- The 6502 had three 8-bit registers: A, X, and Y. A was the accumulator, and received the result of all arithmetic operations. X and Y could hold temporary data, arithmetic operands, and be used as index registers for memory load/store. There was also an 8-bit stack pointer register, SP, hard-mapped to the range 0x0100 - 0x01FF.
- The 8080 had the 8-bit registers A, B, C, D, E, H, L, and a 16-bit stack pointer. In addition, the registers B & C, D & E, and H & L could be used to hold 16-bit quantities for some instructions.
- The Z80 had all the registers of the 8080, plus a shadow copy of the registers for quick use by interrupt service routines.
- The 6502's zero page (0x0000 - 0x00FF) got special treatment by the CPU, using only a single byte to address a location. As such, zero page usually got treated by software as a pile of "slow registers."
- No instruction on the 6502 executed in fewer than two clock cycles. The fastest 6502 I ever saw was 2 MHz.
- By contrast, 4 Mhz Z80 chips were widespread.
- The Z80 helped popularize dynamic RAMs by containing a very basic DRAM refresh counter. The 6502 had no such thing; DRAM refresh was usually provided by custom logic, usually part of the video controller.
- S-100 machines had huge power supplies because they had huge numbers of slots (eight or more being common), and had to have enough reserve power for all of them.
- There was nothing special about the 6502's memory access patterns, and 6502 would get starved out like any other CPU if another device held the bus. On the C-64 in particular, every eight video lines, the VIC would grab the bus for 40 uSecs to fetch the next row of character cells, holding off the 6502 the whole time. This led to all kinds of problems with timing-sensitive operations, and was directly responsible for transfers to/from the 1541 floppy drive to be glacially slow.