Re: How Intel Can Save the Itanium!

In article <memo.20070402191850.1476F@xxxxxxxxxxxxxxxxxxx>, jgd@xxxxxxxxx (John Dallman) writes:
In article <GMadnUGe9JJ_t43bnZ2dnUVZ_vamnZ2d@xxxxxxxxxxxxxxxxxxxxxxxx>,
billtodd@xxxxxxxxxxxxx (Bill Todd) wrote:

Did they really expect 64-bit Windows (or possibly Linux, but I doubt
they were counting as much on that early-on) to blow away the
established Unix mid-range-to-high-end server market,

They didn't want to believe that it wasn't possible. No, it isn't
rational.

such that there was no need for support from established players
such as Solaris, AIX (well, they had hoped for these two, but
support was dropped long before Itanic first shipped),

But by then the religion had its doctrine.

VMS (which Compaq and then HP have allowed to languish),

It was going to be ported. Indeed, that is still supposed to be going
on, probably somewhere obscure in India. I'd guess they saw it as
lacking growth potential, and thus not a priority.


VMS was ported quite a while ago now. As of the end of this month HP will no
longer sell Alpha systems so all new VMS systems sold by HP will run on
Itanium (End of Alpha sales was going to be in October 2006 but HP
extended it).


David Webb
Security team leader
CCSS
Middlesex University



and Tru64 (which they didn't even bother to port)?

Compaq had announced the start of Tru64 porting a month or two before HP
took them over. Since HP already had HP-UX running by then, cancelling
the port after the takeover was an obvious means of saving money and
showing the Compaq people who was boss.

And (with respect to the previous point), did they seriously expect
that AMD's embrace of 64-bit x86 technology could successfully
be ignored?

That came along a little later. And it wasn't until people got samples
and a basic OS that people began to realise just how much better AMD64
worked than Itanium. I had a very interesting 'phone call with Intel
around then. I had a product in production on Itanium/Windows, having
spent over a man-year on it. It worked OK, but performance was less than
sparkling, to say the least. The AMD64/Windows version came very close
to just building and working. When I told Intel that the porting effort
for AMD64 had been under 5% of the Itanium effort - probably less, but I
could be confident of 5% without adding up a load of timesheets - they
went all quiet and worried.

I suppose one could suggest that (save for the power and x86
performance issues) most of the above might likely not have occurred
had McKinley shipped in 1997-1998, but they certainly became
important individual sources of pain as Itanic's voyage actually
progressed (and I've probably missed a few more at that).

Here's two that were closely related.

If you wanted your B-series (beta) Merced chips upgraded to C-series
(production), you had to pay. Something close to US$1000 each, and a lot
of those Merced systems were dual-processor. The B-series worked, but
had bugs that you needed a compiler switch to avoid, and that cost
performance. I don't know how much, because Intel weren't very efficient
about spreading the word, so my management never got the chance to
decide it wasn't worth the money. I've gained the impression that MS did
pay, for a thousand or so chips, which is the kind of thing that makes
people look at their assumptions anew, and wonder what they're getting
for all this money.

Following this, once the first Itanium II systems came out, Intel
dropped support on the Merced development systems, with only a few
months notice. Which meant you needed to buy new development systems
from HP: their ZX-1 chipset was the only small system (1-4 processors)
chipset, and they only sold it in complete systems. HP's prices weren't
particularly keen, and the Merced systems still worked - but Intel were
trying to save money at the wrong moment. Those Merced systems were
still needing regular BIOS updates and the like.

At this point, seeing that we had put a great deal of work into the
Itanium - I reported over 50 compiler bugs that got fixed - and the
customers still weren't keen, we got fed up, and prioritised AMD64.
Turned out to be a very sound decision: the customer who desperately
wanted 64-bit Windows said "Itanium is dead" the day they had a
disclosure on EM64T. They hadn't dared go for an AMD-only platform, but
once Intel had to complete with AMD, they were all over it.

Nowadays, when vendors try to interest me in supporting Itanium
platforms, I just say "Old joke, not funny any more."

--
John Dallman, jgd@xxxxxxxxx, HTML mail is treated as probable spam.
.

Relevant Pages

  • Re: How Intel Can Save the Itanium!
    ... HP's core objective for Itanium was to ... throwing money at it. ... I had a very interesting 'phone call with Intel ... of those Merced systems were dual-processor. ...
    (comp.arch)
  • Re: What to buy?
    ... The fact that no money was being made on it didn't enter into it at ... compared to itanium, performance was far higher and they were making ... Intel thought that it would be the ... advantage that it was so complex that it could never be copied, unlike x86. ...
    (sci.electronics.design)
  • Re: IA64 and emulator performance
    ... |>> Are you sure they do lose money on Itanium? ... If that's true, I feel reasonably confident that Intel are, ... God only knows whether HP is subsidising Intel, ...
    (comp.arch)
  • Re: HP pulls out of IA64 chip development
    ... Today Itanium has patent protection. ... Those who point out AMD ... as a stiff competitor to Intel should realize that Intel would rather be ... making their money on Itanium where AMD is not a problem. ...
    (comp.os.vms)
  • Re: Identification 32/64 bit Linux
    ... The Itanium platform was given the designation "IA64" because it was ... the IA32 architecture, eventhough it's an entirely different platform. ... emulation and does not quite perform as well as a native IA32 ... not even being used by anyone, which has why Intel has contemplated - ...
    (comp.os.linux.setup)