> Our software is order of magnitudes slower than it could be, RISC-V or x86 doesn't change anything.
That's not really true, the most critical bits of the stack or often implemented pretty fast.
This is what DEC learned: "The conclusion was clear: for any given amount of investment, the RISC designs would outperform a VAX by 2-to-1."
- DEC is dead, long live DEC
I'm not suggesting that it is 2-to-1 now but its still relevant.
There's definitely better and worse ISAs e.g. some are unjustifiably complex.
Yet the key advantage of RISC-V is not at all technical, and lies in its licensing scheme.
There is no need to ask for permission to implement RISC-V and thus be able to leverage the best ecosystem out there, which RISC-V is rapidly building.
There is also no need to ask for permission to extend it with custom extensions, to suit your specialized needs.
Using the RISC-V trademarks is only a little more complicated, as it requires proving standard compliance.
Yet this aligns well with most implementations; Compliance is desirable as it enables tapping into existing software and tooling, which is much cheaper than building and maintaining bespoke solutions.
As for licensing I agree that it can be seen as a step in the right direction, but at the same time it seems like we absolutely refuse to learn.
All standards inevitably deprecate, I'd prefer every hardware manufacturers to come up with their own one-off ISA and an easy way to run/interpret existing applications.
Software ecosystems are more about locking you to a specific vendor/standard, they prevent the emergence of better systems.