Subj : Re: X86S
To   : Spectre
From : tenser
Date : Thu Apr 25 2024 07:46 am

On 24 Apr 2024 at 12:09p, Spectre pondered and said...
 
 Sp> Was going to point out as I started reading, the Itanium arrived first, I
 Sp> don't remember seeing it for anything other than server class hardware,
 Sp> and everything had to be re-written/compiled for it.  Where as AMD64 was
 Sp> more an extension for X32 and was backwardly compatible.

The big problem with Itanium is that it was predicated
on having really, really smart compilers that could do
the instruction scheduling (it was VLIW).  Those never
materialized, so it never lived up to its performance
potential.

AMD led quite the coup with x86_64.  What happened there
was that Digital Equipment Corporation was in its death
throes, and they hemorrhaged some of their best Alpha
designers (Alpha, at the time, was the fastest microprocessor
in the world) to AMD.  At the time, Intel was pushing
Itanium hard, and refused to entertain the idea of a
64-bit x86.  This extraordinarily  talented team, now at
AMD, didn't want to futz about with an also-ran x86 clone,
so they came up with x86_64.  It was fast, supported a
large virtual address space (x86 had had PAE for several
years by then, so they already had a large physical address
space), and retained compatibility with 32-bit x86
applications.

That was kind of what the OEMs all wanted, which meant that
the desktop market and low-end servers all went x86_64, and
Itanium was relegated to the high-end, where it only had
marginal market penetration.  Eventually, x86_64 took over
there, too (at least by volume), in part thanks to the
hyperscalars paving the way for large-scale x86 deployment
in server environments.

--- Mystic BBS v1.12 A48 (Linux/64)
 * Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

.