I always wondered if its economical in terms of performance, to invest lots of transistors into a CPU for maximum clock rates and all sorts of caches and speculative execution as opposed to using that economy to have much more basic cores.

Take the 6502 it only had a few thousand transistors and ran at most a few MZH while todays cpus from Apple and X86 have billions of transistors and a dozen of so cores, and yet the multipler for the later in terms of transistor count is not a multiplier for performance.

If took the transistor budget of a complex cpu today and created thousands of 6502 for example until they were expended, the simple one would win, so why go the other way ?