Yeah I should clarify my point a bit better. I tried to answer these kinds of questions in this story: https://erik-engheim.medium.com/the-case-for-risc-v-on-desktops-and-severs-60b0106c636b
But let me try to do a summary here.
1) RISC-V in deskop variant will still only need less than 200 instructions. So far less than the 1500 instructions of x86. Of course transistors for branch prediction, out of order execution and caches wlil dwarf transistors required for the decoders in a high performance desktop chip.
However small cores still matter if you implement specialized hardware for things like AI, image processing, encryption etc. That will increasingly be the way companies like Apple will deliver high performance to users.
It is not your email applications which require screaming performance, it is your video editing and image processing software. Here specialized hardware will play a major role. You can look at something like Esperanto Technologies SOC-1. They make a chips with a large variety of different RISC-V cores. Some RISC-V cores are general purpose cores with lots of transistors. Other cores are simple specialized cores for AI tasks. However they still use the RISC-V instructions.
For implementers and devlopers it should surely be beneficial to see the same ISAs used all over the place. x86 and ARM cannot offer anything like that. Specialized co-processors and accelerators will have to use entirely different ISAs.
2) No desktop processing is not that parallel, but most desktop processing is already fast enough. The things which are not fast enough do either benefit from specialized hardware accelerators/co-processors or multiple cores. Server tasks and compiling has very obvious benefits from going multi-core.
If you look at the M1, AI, Image processing and video encoding is NOT done on the GPU. It is done on other specialized hadware such as the Neural Engine and other co-processors.
The point here is that those accelerators can be made as RISC-V processors. That is the advantage of an architecuture which can be tailored for different tasks. x86 and ARM ISA are not suitable ISA for making accelerators and co-processors. They are only usable as general purpose processors.
RISC-V has the advantage of being able to cover all of them. You can do embedded CPUs, general purpose CPUs and specialized accelerators.
That makes it a valuable architecure as you can build tooling which can be utilized for a large variety of hardware. In the age of hetergenous computing, RISC-V will be king. It was made for this world.
