microarchitecture
Picture two phones that both run exactly the same apps. From the outside, and to every app, they behave identically — same instructions, same results. Yet open them up and one has a tiny efficient engine and the other a big aggressive one that does many things at once. Microarchitecture is the detailed inside design of a particular processor: the specific, hidden way a chip is built to honour its architecture (its ISA).
Microarchitecture is where all the clever performance tricks live: how deep the pipeline is, whether the processor executes instructions out of order, how many it can start each cycle, the sizes and layout of its caches, how it predicts branches. None of this is visible to a correct program — software just sees the ISA contract and gets the right answers. But these hidden choices decide whether a program runs in one second or ten. A new microarchitecture is what chip-makers ship year after year under the same architecture name.
The honest point: architecture is the contract; microarchitecture is one implementation of it. Intel and AMD chips share most of the x86 architecture yet have completely different microarchitectures, and Intel itself ships a new microarchitecture every generation while keeping the architecture stable. Confusing the two leads to the megahertz myth and to thinking a faster clock alone means a faster computer — it does not, because microarchitecture changes how much each clock cycle accomplishes.
Apple's M-series and a Qualcomm Snapdragon both implement the Arm architecture, so the same Arm program runs on both — but their microarchitectures differ in pipeline width, cache sizes, and out-of-order machinery, which is why they perform differently.
Same ISA, different microarchitectures, different performance.
Microarchitecture is invisible to a correct program by design — it changes how fast and how efficiently, never (if done right) what the answer is.