Date: Fri, 6 Feb 1998 19:55:49 -0800 (PST)
From: "William R. Kerr" <bilker@SpiritOne.com>
To: redhat-list@redhat.com
Subject: Re: RedHat for "Merced"
Unconfirmed rumors below. Note also that they form a chain, so any one
might turn the whole lot into wasted bandwidth (not that I'm overly
concerned on that score).
First, some facts:
1. Merced's native architecture is 64-bit, with an extremely interesting
instruction and register set. Given some really good compilers, in
native mode, it will really scream.
2. Merced also supports a mode that "emulates" (not precisely accurate)
a Pentium. Using this feature, all sorts of legacy x86 systems
will run with existing binaries. Without this capability of hosting
the enormous body of existing OS's and apps., it is likely that no
one would be interested in Merced at all.
Now the unconfirmed. I think all of this is true, but I can't prove it.
1. The CMOS process used to fabricate the first generation of Merceds
is older (i.e., larger and slower) than that used in PII's. Therefore,
when legacy x86 systems/apps are run, they will run *slower* than on
a PII (and perhaps slower than on 233Mhz Pentiums).
2. Microsoft is not planning on porting NT to the native Merced
architecture. (Say what?!!! Yep, this is fairly definite. Microsoft's
plate is currently full with NT5, Win98, ACPI, etc. Hey, it's not as
though they have thousands of talented programmers on 6 continents.)
deductive commentary: NT will run on Merced, but *slower* than on a PII.
Given that Merced is more expensive and generates more (a *lot* more)
heat, the obvious question is "why bother?"
3. Intel, recognizing that availability of software sells processors,
is taking steps to avoid the above deduction. They are investigating
several tracks at once:
a) Intel is attempting to port NT to native Merced. (This is *really*
unconfirmed, probably the most unconfirmed item here. But, it
is a definite probability. Well, likely, really. If MS won't
port NT, someone should. Right?)
b) Intel, wishing to place its eggs in more than the Redmond basket,
is porting one or more commercial Unix systems to the native
Merced architecture. HP-UX, of course (due to partnership with HP
in Merced design), also (possibly) SCO (yucch). There are some
partnership relationships involved here (according to rumor).
commentary: remember the Pentium math bug? One of the things that is
rumored to have fallen out of that is that Intel became aware that
a *lot* of people are running non-MS systems on x86. Apparently,
this was news to upper management, despite years of flogging Xenix.
And it got them thinking that Intel might not need be forever
joined at the hip to MS.
c) Finally, (and this is whatcha came for, folks) it is rumored that
there is a small, semi-official effort in Intel to port Linux to the
native Merced. You can list the reasons why this rumor could
be true: Intel needs to do *some* sort of system-level testing
of the native instruction set; no one knows the eventual commercial
importance of Linux; millions of people buy Intel chips today to
run Linux; part-time work by a handful of Intel engineers costs
virtually nothing, especially when coupled with validation testing;
Intel wants to sell processors, not subsidize Gates, etc.
Grain of salt folks. I don't know that this stuff is true, but I've heard
it through multiple, non-proprietary channels, and it makes sense.
A plausible (and really cool) outcome: next year Linux screams on native
64-bit Merceds while NT is relegated to plodding on slow x86 emulation.
As if NT needs processor assistance to be slow. Man, talk about a pie
in the face.
Gimme a bogomip number. :)
--
PLEASE read the Red Hat FAQ, Tips, Errata and the MAILING LIST ARCHIVES!
http://www.redhat.com/RedHat-FAQ /RedHat-Errata /RedHat-Tips /mailing-lists
To unsubscribe: mail redhat-list-request@redhat.com with
"unsubscribe" as the Subject.