Mark Roulo (nothanks.delete@this.xxx.com) on June 2, 2022 8:15 pm wrote:
> --- (---.delete@this.redheron.com) on June 2, 2022 1:06 pm wrote:
> > Intel could doubtless do somewhat better if they gave up some compatibility, and
> > a lot better if they abandoned all compatibility. But they won't do that. So...
> > And this means the whole package. It's not just ISA, it's memory model, it's IO model, it's cache protocol
> > and locking primitives, it's socketed DRAMs, etc etc. Of COURSE all that stuff costs; if it didn't Apple
> > (mostly free to use whatever they want) would be copying it instead doing something different.
>
> It isn't clear to me that Intel is likely to do better (enough to matter) if Intel could start all over.
>
> The problem is that you have to ship implementation, not just architecture, and
> by starting over you have no installed base and thus very little revenue.
>
> The RISC guys (SPARC, MIPS, Alpha, POWER) all got to start over versus x86 and for the integer
> loads that mattered to Intel they all eventually lost. A large chunk of the losing was because
> Intel could spend so much more on actually IMPLEMENTING a given architecture.
>
> Apple's ARM implementation has a similar advantage today over many of their ARM competitors: Apple sold almost
> 250 million iPhones in 2021 and also sold a bunch of iPads and some M1 Macs. And all of these products are high
> margin. Which makes it easier for Apple to justify spending dollars on their next ARM implementation.
>
> Clearly a given architecture can be so botched given where the
> market goes (e.g. 68K, VAX) that even this won't save them.
>
> But it isn't obvious that Intel's x86 *IS* that botched.
>
> The history of MIPS, SPARC, POWER and Alpha competing against x86 suggest not.
>
> And, of course, Intel has tried a few times to get away from x86: i432, 80860,
> 80960 and Itanium. Those didn't turn out very well as competitors, either.

960 was actually very successful, but not as a direct competitor to x386.
Then they got StrongArm as result of deal with DEC and caught reversed-NIH syndrome that ended
up in both stagnation 960 and not getting much value of StrongArm/XScale.

>
> So, yes, it is *technically* possible to do a better job than x86. And Apple (arguably) is
> doing it. But it isn't obvious that Intel can -- because of business/revenue limitations.
>