Why Is Apple M1 So Fast

Explained here. The article contains a gentle refreshers about ISA, pipelining, CISC vs RISC, CPU architecture, out-of-order execution…​ Before it dives into Apple’s not so secret architecture.

M1 is not a CPU. It is a system of multiple chips put into one large silicon package. It departs from the traditional PC architecture. Its Unified Memory Architecture addresses the traditional limitations of shared CPU-GPU memory.

Intel and, to a lesser extent, AMD are left in an undesirable position, as their business models do not fit the new paradigm. Their OoOE is not match for the M1, and there is no obvious way for them to catch up, in part because of the x86 instructions variable lengths, to paraphrase the article.

Does it mean Apple won, game over? I wouldn’t jump to that conclusion. There are many variables other than the technical aspects — otherwise Intel would have been already burried by AMD on several markets where its product is already faster, cheaper and more efficient. But for them these are long hard times to come after a much needed wake-up call.