The thing with ARM is it's an ISA, an instruction set. The latest iteration, ARMv8, hasn't really changed much since its introduction way back when in early 2010. What drives the innovation you seek is how the processor using it is implemented. Like for example, AMD owns the x86-64 ISA, but between Core 2 and Skylake, Intel has dominated in performance. And it's not like Intel can really add anything willy-nilly to the ISA to somehow generate that performance.
Sony and Microsoft didn't do anything completely different. They just asked AMD to make a chip for them and all AMD did was marry Zen 2 cores and an RDNA GPU together into a single package. Something that AMD already does by making APUs. The only reason why nobody else would want to do it is it costs money to ask for something AMD doesn't already make. The SoC's in the consoles are not part of AMD's product line, and AMD isn't just going to make them just because.
Of course they do, and companies that don't have the resources to develop their own CPU are stuck licensing their cores. But for people who do have the resources, they can develop their own CPU.
However, using something open like RISC-V doesn't imply any implementation of it has to be open. Everyone can still make proprietary CPUs with open ISAs.
On your first point, RISC is based on the Basic Instruction set since the 8086 chip. The whole point in RISC was to keep the execution core to its basic and run all new and complex instructions (ran on CISC equivalents) with a basic command set. In fact ARM only add extra instructions that change the hardware operation at a Kernel based level rather than core. So rather than deal with the instruction sets, they deal with changing between Big and Little sets of cores.... and what a waste that has proved, and one that the Ecosystem loves because it means that the Fab plants can charge the consumer more for the product. So, in other words, an Octacore chip provides you with 2 sets of cores that cannot run concurrent, but switch between then. This increases cost and is a poor implementation of resources, but the public literally buy into it at their own expense. Being an ASIC's designer myself, I find ARM a wasteful and gimmick company that provides little value to the eco system. In fact, it often raises it bar (in terms of performance) by using CISC IP.
Secondly, Intel and AMD can make the changes to the architecture as a customer demands, but now they dont have to. Digital design and Sequential logic for CISC operations is easy enough, and companies like APPLE can use their own engineers to modify the Chips without AMD knowing, so long as they don't infringe on Intels or AMD IP in the process when it has not be licensed.
And Yes Intel tried to dominate the ARCHITECTURE increments from 586 + by paying programmers and gaming houses to ignore the competitions iterations of Hardware Changes. So when 3Dnow came along, Intel came along with SSE1 and used unfair practice to force MOST, not all, but most software developers to follow this architecture rather than its competition. So ARM does not have to dominate the progress of the new iterations of hardware progression on the RISC ecosystem. AMD, beat Intel at its own game on so many occasions with its SSE1A, albeit at a small compromise on performance.
As with AMD's bulldozer, it showed to the Public what AMD and Intel really do, and that is just re-oganising pipelines and physical structures of operation handling in the chip. This is essentially all ARM does too. AMD finally figured out by trial and error, with Ryzen 1, that memory material and speed of Logic Gate operation due to back ground tech was killing it, rather than its Architecture design.
Essentially, all the tech is the same, it is made up of Logic gates with Next state logic design (state diagram melee or Moore) with essentially the same IBM registry setup from the 60's with Shift registers and JK flipflops. Each iteration is just trying to reorganise these simple circuits and make the interconnects and VIAs as short as possible, the CISC element is proving (as RYZEN is now showing) that the CISC elements are wasteful and often dont produce as many pipeline hits as it used to. Hence GPUs are becoming a better option for faster computation... Hence RISC is catching up with some of the designers like Samsung, that only Loosely follow ARMS ASIC design in favour of their own. The GPU element more and more being used to speed up complex operations on the CPU,
China will eventually Move in its own direction once it figures out the CISC from the Ryzens first iterations. Then it will stop selling rare earth materials and destroy the western dominance on the electronic market. Trump could see it, and tried to stop it, but Clueless Biden will allow China to have the time it needs.
And, yes, AMD does own the 64 bit iteration of the CISC systems, but only because INTEL created its own iterations to only work with a 64 bit command set. AMD made a hybrid that would work with all 8-64 bit structures, and no amount of bribery was going to stop AMD's dominance. Now Intel Could move over to 128 bit (not that it needs to as 64 bit covers the memory problems it was meant to solve), but could do 2 bits simultaneously... Oh yes! it already does that with AVX, its just that they are waiting for the chance to capitalise that never came (and the fact that Parity checks became massively more complex and Spectre kicked in).
But to finish up. ARM only has dominance because no comany wants to become the AMD of RISC. Not until China chooses to.
Facebook
Twitter
Instagram