Most
microcontrollers on the market today have two basic architectures, each with two variants. One architecture refers to the type of bus design used for the program and data memory:
von
Neumann vs.
Harvard. The other classification describes the type of instruction set and the complexity of the CPU:
CISC vs.
RISC. Microcontrollers from
Motorola,
Intel,
Zilog and
National Semiconductor all have
von Neumann architecture and CISC underpinnings.
Harvard architecture has separate program and data memory busses. The program memory bus is variable in width and
optimized for a particular device. The data bus width is usually a standard 8 or 16 bits. The special function registers (timer settings, port configuration, etc.) are also mapped onto the data bus because they are essentially variables. This architecture allows the simultaneous access of both program memory and data memory.
Harvard architecture has several benefits, mainly a long word instruction that occupies only one location in program memory. A single word instruction increases the execution
speed because the opcode and associated data are all contained in a single word. The single cycle instruction execution is also easily performed in the Harvard architecture
because the program and data memory can be accessed at the same time.
A popular maker of Harvard architecture chips is Microchip, who makes the PIC series; by some accounts, they are the #2 microprocessor maker! This is largely due to use in consumer electronics. Scenix is Microchip's AMD: Faster, cheaper, code-compatible, not as diverse.