[Info-vax] OT: IBM Offering $9-10 Per Share for Sun

[johnwallace4 at yahoo.co.uk]

On Apr 6, 1:22 pm, Lee Witten <nos... at nospam.com> wrote:
> johnwalla... at yahoo.co.uk wrote innews:f34e7983-8ffb-416d-8bcd-c62918794768 at f32g2000vbf.googlegroups.com:
>
> > Fortunately these days in this picture we're not stuck with classic
> > x86 architecture. We're talking about its next generation successor,
> > the AMD64 architecture, which just happens to have succeeded in being
> > remarkably compatible (transparently so, in the vast majority of
> > cases) with classic x86 when it needs to be, whilst also behaving like
> > a real 64bit architecture when *that* is what's needed.
>
> Agreed.
>
> > As for x86 systems hardware: it's already been pointed out that the
> > innards of any reasonably modern low-end Alpha (from any AlphaStation
> > onward) were remarkably PC-like in terms of chips and functionalities,
> > and the same could presumably be said to some extent for Itanium,
> > especially if Quickpath for Itanium ever actually comes out (even
> > though it's no longer going to be socket-compatible, it presumably
> > needs to be architecturally compatible, unless silly subsetting games
> > and the like come into the picture).
>
> There are enough differences to make support a non-trivial effort.
>

Well,  it all depends on what you mean by "trivial" and by "support".
Are there many differences between an Alpha/VMS driver for card/
chipset XYZ and an Itanium/VMS driver for card/chipset XYZ? So why
would an AMD64/VMS driver introduce significant new differences
(examples welcome)? The system core chipset may (obviously) be
different between the two families; whether Quickpath would make such
a porting exercise any easier remains to be seen.

Presumably if there was a sensible business case for doing it (porting
and qualifying on a limited subset of Proliant/AMD64 servers), it
would get done, right?

> And if the hardware is identical, there won't be much if any advantage
> for using itanium/alpha, other than if you need to run vms.
>
> One could have seen the itanium floating point as a strong advantage,
> but it's almost nil versus the nahalem (sp) generation of x86 CPUs.

Indeed. Plus for some classes of HPTC, this week's fashion seems to be
to use graphics cards as low cost high performance FP compute engines,
so long as the user doesn't mind some changes to the apps.