Oh, them's fighting words =) I wrote ARM assembly before I wrote MIPS, so for me, MIPS looks like ARM. ARM was influenced by Berkeley RISC. MIPS came out of Stanford. ARM and MIPS CPUs were both introduced in 1985. And MIPS is dead, whereas ARM is doing slightly better, so, statistically, if you show a RISC assembly programmer MIPS code, they'll probably say "that looks like ARM".
Now there are two approaches to the performance argument.
First, I will argue that RISC processors provide better performance than CISC processors.
Second, the counter argument to that is that, actually, no, modern RISC processors are just as Complex as CISC processors, and M1 is faster simply because Apple. My second argument is that Apple choose ARM because of RISC. So even if it were true, now, that one could build a CISC that is just as performant as a RISC, the fact that right now, the most performant chip is a RISC is ultimately because it is a RISC.
Do RISC processors provide better performance than CISC? The #1 supercomputer in the world uses ARM. AWS's Graviton offers better price and performance than AWS Intel. M1 is faster power/performance than any x86. ARM holds all the records. But it's just a coincidence?
I think PP's position is that CISC or RISC doesn't matter. One argument I've heard is that its the 5nm production node that matters, and that CISC or RISC, it's all the same nowadays.
So how is Apple on 5nm? Why are the CISC manufacturers stuck on 7nm (or failing to get even there)? When the Acorn Computer team were looking for a new processor, they were inspired by the UC Berkeley and their RISC architecture. In particular, the were inspired by the fact that students were designing processors. They decided that they, too, could build a new CPU if they used RISC, and that was the Acorn RISC Machine. The ARM. I do not believe that RISC and CISC are "basically the same" when it comes to processor design. The fact that Intel is still stuck on 10nm(?) must in part be due to the thing being CISC. One might argue that it's because they made a particularly complicated CISC, but that would only make my point for me. I don't think that "only the instruction decoder is simpler so it doesn't make much difference" holds any water. I would love to hear from actual CPU designers who have worked on M1 or Graviton, but if they said "RISC is easier" then they would be dismissed as biased.
But let's suppose that no, actually, the geniuses at Apple could equally create a CISC CPU that would be just as performant, I'd still argue that the success is because of RISC. M1 is the most recent of a long line of ARM CPUs. They are ARM CPUs because the first in that long line needed battery life - the Newton. M1 is ARM because ARM was RISC when RISC mattered. You may argue that RISC doesn't provide a benefit over CISC now, but it certainly did then.
And how does Apple have such geniuses? Again, this is largely because the iPhone was perhaps the largest technological step-change in history. They have the market cap they do because of iPhone, and they have iPhone because RISC. So even if the argument is "Well, the M1 is fast because Apple has lots of money" well that is because of RISC.
But I still think that its easier to build a faster CPU with RISC, and I expect the first RISC Mac Pro will prove me right. At which point, RISC will own performance/watt, performance/price, #1 supercomputer, and, at last, fastest desktop.
