>>85654
You're right, but also sorta wrong, but it's a nuanced situation.
>x86 is archaic
It is and it isn't. the interface has backwards compatibility back to the earliest releases, but under the hood these no longer map to the same commands they once did. An opcode as simple as "add two registers" will map to multiple micro-opcodes which may or may not try to optimize the instruction but will definitely map to whatever internal interface the hardware uses. It's arguably a limitation but not nearly as much as it would be if the old architectures were maintained. ARM and other RISC architectures don't really suffer from the same problem because they minimize instructions by design, and their regular opcodes are closer to Intel's internal platform than their external interface. Intel does not publish information pertaining to their external interface, they just occasionally add new opcodes that map to different micro-ops and then compilers will eventually use them for optimizations.
>PowerPC never took off
It actually found niches in servers and military hardware for a time. I think you're romanticizing the appeal of it. It's not an amazing spec, and it lost to Intel for a few reasons. Sure, with more funding and support maybe it could have competed, but...
>open-sourcing hardware led to fewer options
The actual architecture is far from open source, there's just a spec to write against now, which was always sorta the case. Intel still hides plenty, including secret opcodes. Hardware is not "open source" the same way software often is, and even if it were, it's difficult to validate and enforce. All of them are encumbered by licensing anyways.
You're completely underestimating the reason why everyone went to x86 in the first place: supporting multiple architectures was absolute cancer, and the further back you go this only gets worse. Things are better now than ever. Most compilers, GCC especially, can spit out binaries for basically anything. Even then, you need to ensure there's a C library and some other stuff. And you have issues, like the new M1 Macs which are ostensibly ARM but often require some custom tweaks to get stuff working. And this only solves the issue if you're willing to compile it or someone will package it (and all its dependencies) for you.
>The only real alternative is ARM... because smartphones and micro-systems
It caught on there years ago due to power requirements, but I think this is a case of Intel never having to design for that case before and completely fumbling. They've had some major holdups the past decade or two. Their attempt at x64 bombed. Their attempt at low-power bombed. Their attempts at shrinking their dies has bombed multiple times in a row now. It's not their strength. Intel got fucked by their own incompetence the same way Detroit did. More power works great until you have enough and then portability and power start to matter more, and if you can't design for that you will die.
Ultimately I don't think the decline in options is because of the spec being open. It's because IBM, Microsoft, and Intel conspired to make it the dominant force so they could all maintain a monopoly and Windows support for non-x86 has never been good. Microsoft really killed alternatives on the home desktop more than anything as a means of lock-in, and it was so effective it made alternative parts go from rare to unavailable.
>>85655
Per above, Intel's spec is basically a huge lie so it's pretty wrong to say it's "open source". Furthermore, it's encumbered by patents and licensing, so not actually open even if the true spec was. RISC-V is still a step in the right direction for openness. The addons are responsible for a lot of important stuff like encoding and wireless that are needed for a modern CPU, but getting the core adders and opcode processing to be an open design is huge. The addons are closed on modern chips, too, and are also encumbered by patents and licensing. People making ARM SoCs are also paying for these addons to be on their board, and they are black boxes. RISC-V lets us go from "the entire motherboard is a black box" to "the bits from the integrator are open". Once that becomes standard, we can see open source module alternatives appear. We can also effectively force competitors to contribute to the same design. Intel and AMD have both made some performance gains in the past few generations, but being able to see each other's work and improve on it would theoretically import those gains.
You're right, the inability for someone to cook their own modern silicon is a real issue, but it still lowers the barrier of entry to competition and allows home users who are interested in CPU design to get involved and offer feedback.