[adrian_b]

All modern high-performance computers use a combination of von Neumann Architecture and Harvard Architecture.

They use a big unified main memory used to store both programs and data, but the CPU has separate instruction and data cache memories, in order to be able to overlap in time the fetching of the instructions from the instruction cache memory with the loading and storing of their operands from/to the data cache memory.

Only some of the cheapest microcontrollers without caches may not have such a combined architecture, but even many of them can overlap fetching instructions from a read-only flash memory with loading and storing operands from an internal SRAM, while also being able to use an external unified memory for both programs and data.