Intel drops 8086 instruction support from Montecito

Intel drops 8086 instruction support from Montecito

Post by JF Meze » Sat, 21 Jan 2006 08:45:26


> http://www.yqcomputer.com/ +scraps+once-crucial+Itanium+feature/2100-1006_3-6028817.html?tag=nefd.lede

Essentially, for Montecito, Intel is dropping the hardware emulation of
the industry standard instruction sets.

Instead, anyone wishing to run programs for x86 chips on Montecito must
use Intel emulation software called IA-32 Execution Layer, or IA-32 EL,
that was first released in 2004.

The change, which Intel had refused to discuss until now, reflects the
company's diminished Itanium ambitions, which cast the chip as being
only for higher-end servers.

(so I am not the only to have read that Intel/HP as giving IA64 a small
niche at the high end).


Here is an interesting one:
##
Support for IA-32 EL is necessary for operating systems on Montecito,
according to the manual. "All OSes running on Montecito have a
requirement to have IA-32 EL installed," the manual said.
##

The manual can be accessed at:
http://www.yqcomputer.com/

(haven't checked the URL, if it doesn't work, go to the news.com article
where the link is available).


So, according to the above statements, VMS relies on that IA32-EL being
present ?-) Perhaps VMS was already ported to the 8086 and has been
running on that IA64 thing in 8086 emulation mode :-) :-) :-) :-)
 
 
 

Intel drops 8086 instruction support from Montecito

Post by GreyClou » Sat, 21 Jan 2006 09:13:29


After scrounging around my Mac developer header files, I ran into a lot
of Itanium based code inside the new headers.
Makes ya wonder if Intel is going to offer Jobs a good deal on Itaniums.


--
Where are we going?
And why am I in this handbasket?

 
 
 

Intel drops 8086 instruction support from Montecito

Post by young_ » Sat, 21 Jan 2006 09:48:17

In article < XXXX@XXXXX.COM >, JF Mezei < XXXX@XXXXX.COM > writes:

And this is a very good thing. Did you read the article?

"IA-32 EL provides much better performance and flexibility for 32-bit
applications on Itanium," spokeswoman Erica Fields said of the choice. "With
Montecito, we took back the silicon area that was being used up by the x86
hardware support."


Paul had fun with that whole view. Read his comments:


http://www.yqcomputer.com/

Name: Paul DeMone ( XXXX@XXXXX.COM ) 1/19/06

what anyone could see in a Montecito die photo a year
ago. In fairness to CNET even the supposed chip experts
at MDR didn't notice either. :-P

http://www.yqcomputer.com/ +scraps+once-crucial+Itanium+feature/2100-1006_3-6028817.html?tag=nefd.lede

"With the next Itanium chip, Intel has abandoned a feature it once banked on
but that never proved successful, CNET News.com has learned.

Circuitry to let Itanium run software for x86 chips, such as Pentium and Xeon
chips, is not present in the forthcoming "Montecito" processor, according to
the 176-page reference manual for the chip published this week (click for
PDF)."


BTW, CNET idiotic statement of the week:

"The change, which Intel had refused to discuss until now, reflects the
company's diminished Itanium ambitions as a chip just for higher-end servers."

Or it could be because IA32-EL is ~3x faster and dropping
x86 support from Montecito silicon freed up die area for a
1MB L2I that will speed up IA32-EL based x86 performance
even more as well as native IA64 performance.

----

Get it, the software emulation is about 3x faster, make sense
now?

Hope that helps!

Rob
 
 
 

Intel drops 8086 instruction support from Montecito

Post by Alan Grei » Sat, 21 Jan 2006 09:59:32


Nope. Not a chance. However Apple almost went down the Itanium route a
few years ago so you may be seeing the remains of that. Or it has just
trickled into header files because Itaniums exist and run Unix. Do you
honestlty think Apple would move to X86 now if they thought there was
even the remotest chance of moving to Itanium in the forseeable future?
--
Alan Greig
 
 
 

Intel drops 8086 instruction support from Montecito

Post by Tom Linde » Sat, 21 Jan 2006 10:10:24


That hurts my ears, the plural surely has to be Itania.
 
 
 

Intel drops 8086 instruction support from Montecito

Post by JF Meze » Sat, 21 Jan 2006 10:10:53


I didn't say it was bad.

But this is sigificant since it is one of the vestiges of the era when
Intel really though IA64 would replace the 8086 as high volume industry
standard chip.



At the time they started that IA64 thing, DEC already had its VEST, (did
they have FX32 already back in 1994 ?), and Apple already had had its
68k emulator on PowerPC. It find it somewhat strange that they would
have decided to go for an 8086 hardware solution, especially in light of
time to market (even if IA64 had been on time).

I am personally surprised that hardware feature would have made it past
Merced. Back in 1999, DEC people were already making presentations on
how slow IA64's built-in 8086 emulation would be.
 
 
 

Intel drops 8086 instruction support from Montecito

Post by David J Da » Sat, 21 Jan 2006 10:56:51


"Itanics"

...or, in the case of UN*X-like o.s.-es, "ITANIX".

:-)

--
David J Dachtera
dba DJE Systems
http://www.yqcomputer.com/

Unofficial OpenVMS Hobbyist Support Page:
http://www.yqcomputer.com/

Unofficial Affordable OpenVMS Home Page:
http://www.yqcomputer.com/

Unofficial OpenVMS-IA32 Home Page:
http://www.yqcomputer.com/

Coming soon:
Unofficial OpenVMS Marketing Home Page
 
 
 

Intel drops 8086 instruction support from Montecito

Post by GreyClou » Sat, 21 Jan 2006 11:54:51


It's possible. He may be looking into something for his high end
towers, but only if he gets a sweet deal.

--
Where are we going?
And why am I in this handbasket?
 
 
 

Intel drops 8086 instruction support from Montecito

Post by Alan Grei » Sat, 21 Jan 2006 20:04:29


Why? There is no significant performance advantage to justify this?
Itanium is not now for the low end or the desktop. That has been stated
over and over again by Intel and has also been stated explicity by Rich
Marcello (who actually claims it was his idea). It's a fantasy to
believe that Apple will use Itanium any time soon - or ever for that
matter.

Intel could give him the entire stock for free and I'm sure he wouldn't
use them.
 
 
 

Intel drops 8086 instruction support from Montecito

Post by koehle » Sat, 21 Jan 2006 22:42:05

In article < XXXX@XXXXX.COM >, "Alan Greig" < XXXX@XXXXX.COM > writes:


Yes. Since OS X sits on a BSD kernel on a Mach micro-kernel, once
Apple gets endian dependencies out of its own code they can port
rapidly to almost anything.

Maybe Jobs will look around and buy up Alpha licenses so he can get
a really fast processor to run OS X on. Heck, he could offer both
big-endian and little-endian OS X on the same hardware.
 
 
 

Intel drops 8086 instruction support from Montecito

Post by bill » Sat, 21 Jan 2006 23:13:23

In article <JIO3svd1lNd+@eisner.encompasserve.org>,
XXXX@XXXXX.COM (Bob Koehler) writes:


Not being in the least bit interested in Macs I never looked but I think
this is wrong. OSX, from what I have read, is a Mach micro-kernel and
the FreeBSD userland. No BSD kernel. Running two kernels really sounds
silly anyway. But then, it's a Mac, who knows. After all, the latest
suggestion read to me like someone wanting to run VMS on top of Wine, on
top of OSX. I make that to be:
VMS -> Wine -> Mach -> BSD
Maybe we should throw SIMH in there somewhere. We couls always try to
set the record for the most layers in an emulated system. :-)


Excsue me, but he already had that. He is apparently abandoning the
really fast Power architecture in favor of x86. Why would he now want
to move to a dead architecture? He ahas already taken a big enough step
backwards. (I am not saying that Alpha as an architecture would be another
step backwards, only that the Alpha is dead. We have had the wake. We
threw dirt on it. It is time to let it rest in peace.)


And the advantage of that is???

bill

--
Bill Gunshannon | de-moc-ra-cy (di mok' ra see) n. Three wolves
XXXX@XXXXX.COM | and a sheep voting on what's for dinner.
University of Scranton |
Scranton, Pennsylvania | #include <std.disclaimer.h>
 
 
 

Intel drops 8086 instruction support from Montecito

Post by JF Meze » Sun, 22 Jan 2006 00:28:53


You need to look at the reasons Apple dumped PowerPc to go to the 8086:

Apple was having difficulty in sourcing fully scaled chips from laptop
to datacentre. Apple wasn't big enough t justify the costs of developing
its own custom PowerPC versions in all those sizes.

By moving to the 8086, Apple gets fully scaled 8086s from plamtop to
datacentre because the 8086 market is big enough to justify all those
form factors and development costs.

The 8086's market is being widened to even higher end computers.


For Alpha and that IA64 thing, neither can give Apple the scalability
Apple needs. It is a no brainer.

Yes, they *could* have. But Digital decided not to scale Alpha down to
compete against wintels.

And by the way, some recent articles indicated that Apple seriously
considered moving from 68000 to SPARC instead of PowerPC.

Had Apple moved to Alpha, I suspect Digital might still be alive today
because it wouldn't have tried to become just another 8086 box manufacturer.
 
 
 

Intel drops 8086 instruction support from Montecito

Post by JF Meze » Sun, 22 Jan 2006 00:36:55


Don't forget that Apple wasn't using off-the shelf IBM Power chips. It
was getting custom chips made with a few modifications. And Apple wasn't
big enough to tickle IBM's interest. Its laptop chips were already one
generation behind and done by what used to be Motorola's chip division.
(PowerPC was a IBM/Apple/Moto project).

So moving from an old PowerPC chip in its laptops to current 8086s for
laptops represents a boost in performance.

Apple wanted a wide range of Apple POWERPC chips from laptops to big
servers. So a small volume divided into multiple variants of the chip.

the game console makers want one specific variant of the chip and
promise high volumes. Guess which one IBM prefered ?
 
 
 

Intel drops 8086 instruction support from Montecito

Post by Tom Linde » Sun, 22 Jan 2006 01:00:48

On Fri, 20 Jan 2006 10:36:55 -0500, JF Mezei




They also get a second source, AMD.
 
 
 

Intel drops 8086 instruction support from Montecito

Post by bill » Sun, 22 Jan 2006 01:24:25

In article < XXXX@XXXXX.COM >,
JF Mezei < XXXX@XXXXX.COM > writes:


ANd you think that Intel is going to be willing to make custom chips
just to keep Apple happy? Not likely. And that means that Apple could
very likely have used standard Power chips as well. They went with
Intel in order to ride the wave. And that wave is not Itanium it is
x86.


But more work than it would have taken to move from one Power to another.


I have littl;e doubt that Mach could run on anything that IBM turned out.
Ther is nothing in Power specific to *** . It's just the best CPU on
the market today. Moving to Intel is a step backwards driven by marketing
and not by technical excellence. But then, if MS has taught us anything
it is that with the right marketing mediocrity will always win out over
technical superiority.

bill

--
Bill Gunshannon | de-moc-ra-cy (di mok' ra see) n. Three wolves
XXXX@XXXXX.COM | and a sheep voting on what's for dinner.
University of Scranton |
Scranton, Pennsylvania | #include <std.disclaimer.h>