coppice (coppice.delete@this.dis.org) on November 16, 2014 11:24 pm wrote:
> David Kanter (dkanter.delete@this.realworldtech.com) on November 16, 2014 9:39 pm wrote:
> > Linus Torvalds (torvalds.delete@this.linux-foundation.org) on November 16, 2014 5:17 pm wrote:
> > > After all, the main selling point of Itanium was high IPC. Hmm? It certainly wasn't frequency.
> >
> > I'd argue the main selling point of Itanium was the ability to run NSK, VMS, and HP-UX...
>
> That's not a selling point. That's a consequence of the selling points. Those systems could have been ported
> to practically any ISA. They were ported to Itanium because it was 64 bit and backed by Intel. People assumed
> Intel would do whatever it took to make a success of the platform, and that they had done the due diligence
> work needed to ensure that success was feasible. If Intel had developed something like the current x86_64 at
> the time they developed Itanium, people would have ported their OSes to that. Then saying a selling point of
> x86_64 was that it had the ability to run NSK, VMS and HP-UX would have been just as accurate/inaccurate.


No, they were ported to Itanium because Itanium was originally HP's follow on to PA-RISC. It was already a VLIW architecture, probably quite similar to what IA-64 ended up as, before Intel was even involved. The partnership with Intel was intended to benefit both parties - HP not having to continue investing in fabs and gaining Intel's market power, Intel gaining an architecture they believed was needed for the future to replace x86.

HP-UX was always going to be migrated to Itanium, the VLIW project it came from WAS PA-RISC 3.0. As HP acquired other operating systems like VMS that was also the obvious course of action. That's why I was always annoyed by a certain someone who kept talking about Itanium's "growth" in the marketplace. It was simply doing what it had been originally designed to do, replacing PA-RISC. The growth was not organic, and as I always predicted, the growth ceased and went into decline once HP finally quit selling the last PA-RISC servers.

Had Intel not chased down the rat hole after the 10 GHz P4 dream, allowing AMD the ability to create a 64 bit architecture for x86 and gain Microsoft's blessing of it, the stores would probably be filled with IA64 PCs now. If Intel introduced a 64 bit x86 architecture at all, they'd have made sure it was limited by not adding registers, and probably avoided a flat address space to make it difficult to use, guaranteeing IA64's superiority. Since IA64 was patent protected, it provided them a legal monopoly they could not achieve with x86 without running into problems from the FTC.