The reason being power efficiency just doesn’t make sense any more. Since Apple silicon, ARM is used in desktop computers as well where power efficiency matters less. Power efficiency seems like an artificial limitation rather than a technical or architectural, which means you could just configure it to clock faster at the expensive of more power draw.
I only recently found out just how low the M1 family was clocked at. If the math aligns right (using rough values from benchmarks), it’s IPC is so high, even 6 years later, Lion Cove and Zen 5 don’t have that much IPC as Firestrom. They are faster due to clock speed. Apple since did increase IPC but a larger part is the clock speed as well.
Are there any actual architectural differences which makes ARM harder to clock faster, or is it really just power efficiency? Simpler decode, Apple having lesser ways for L1 cache, just seems counter-intuitive of why ARM clocks slower. Snapdragon, which has similarities with Apple’s architecture, tries to clock a bit higher yet fails to surpass it. Though the core’s specs themselves lack than Apple’s which tells you why.
