Seems we are talking past each other.
1. We agree. Why do you think that I see RISC as a new thing? This puzzles me.
2. And...? I have been pretty clear that legacy technology naturally has advantages in mindshare, marketshare and other things. The logical flaw is when people come to believe that this advantage exist due to technological superiority. It doesn't. It is simply first-mover advantage.
3. The first RISC CPUs where dramatically simpler the CISC processor at the time and outperformed most of them. There is no rule that says complex systems must be higher performance the less complex systems. In fact the opposite is very often the case.
4. We know from the 90s that simpler RISC designed where quite good out outperforming x86. Intel won by throwing a lot of money at the problem, creating some reather complex chips and more aggressively moving to smaller nodes in the chip manufacturing. There is just no compelling evidence that x86 complexity is an advantage for performance.
5. Okay so they were wrong. But I suppose this shows that it is simply a complicated question. Adding more decoders is obviously harder for x86, but evidently not impossible. They may have had some smart heads who could figure it out. Micro-op cache may be similar to the use of micro-ops in RISC and CISC processors. Both RISC and CISC CPUs can use it, but micro-ops gives is more necessary for a CISC CPU otherwise you cannot have functioning pipelines. A RISC CPU mainly needs it to be superscalar and do out of order execution. I am not an expert, but I would speculate that a micro-op cache has more benefits to a CISC CPU than to a RISC CPU. Thanks for the Tachyum Prodigy mention. That sounds like something very interesting to have a look at or maybe even write about. Transmeta was certainly an interesting endevour.
6. Don't think I agree with this analysis. Windows already has something similar to Rosetta 2: https://beebom.com/apples-rosetta-2-vs-windows-x86-emulation-explained/
I would instead argue that they key reason we have not seen this transition is that that the PC market is impossible to coordinate to do a transition to ARM, because it is so fragmented by many vendors.
The PC market has a catch-22. If somebody puts in a huge investment to make a fast desktop class ARM chip like Apple, they don't have any guarantee that they can recoup that investment. To many important factors are out of their hands. Whomever pours billions into a desktop ARM chip relies on Microsoft, Dell, HP an many other are onboard.
Because no such gurantee exists, all the PC market gets is half hearted ARM chips. That means low performance, which means they are unlikely to excite consumers.
Dragging RISC-V into this discussion doesn't make a lot of sense. You call RISC-V a toy because it has not obvious clear path into the desktop PC market? But who says RISC-V even need to enter this market? It is simply one of many potential markets for RISC-V. RISC-V is a very versatile architecture and may exists as co-processors and accelerators in PCs rather than as the main CPU.
One huge market where RISC-V has an obvious opportunity is in the phone market. First for cheap low end phones. But nothing should prevent it from going high-end. Android phones run software written in bytecode not native code. Thus the obstacles to RISC-V is mininal.
