Actually Tench.. for a change my timing couldn't
have been better.
FUD you say? Don't think so!
" Do you really believe Microsoft feels so complacent about the 64-bit version of Windows 2000 that they'd
delay it until late 2001?"
Why don't you ask some folks at Intel, they'll tell you
the status of Win64. Late 2001? Nah... mid-2001 for
production quality.
FUD? Check out some of my posts. I tell you when I
am FUDding up things. Not this time.
Enjoy!
BTW, check out the last 2 paragraphs also, the ones
the moderated AMD guy trimmed.
Rob
What to do about Itanium.
Computer Systems News July 17,2000
Here it is two weeks into the second half of
2000, when Intel is publicly committed to delivering
the first Itanium systems, the culmination of six or
eight years of heroic effort to produce Intel’s first 64-
bit processor, a project designed to make its reputa-tion
and secure its future, and the chip still isn’t baked
and its chief operating system, NT64, is still in pretty
scruffy shape. In fact, Intel now admits, when asked,
that it doesn’t expect NT64 to be production-grade
until the first quarter of next year, an appreciation of
the situation that jibes with other, earlier reports
claiming NT64 won’t be ready on time (CSN No 339,
338). We ourselves have just heard on the rumor mill
that Microsoft is currently aiming to get NT64 out
around Thanksgiving, which, given the quick onrush
of the Christmas holidays, might as well be Q1.
That leaves Linux to carry the load – HP-UX,
Solaris and Monterey not withstanding – and therein
lays Intel’s problem.
The Itanium, by all reports, still isn’t yielding the
733MHz and 800MHz speeds it’s supposed to when
it’s introduced. Ditto the 133MHz double-pumped
bus it’s supposed to have. Development systems
marked 133MHz don’t appear to be the real McCoy
but merely 100MHz single-pumped buses. The best
chip Intel has been able to put in those development
systems to date is a 600MHz, but mostly they’ve
been 500s. None of the 1,500 development systems
Intel has sent to IHVs or ISVs so far has a B-step
processor in it. The best have been A-3s, and it’s
pretty late in the day for A-3s.
Processors go through steppings on their way from
prototype to production – A-Zero, A-1, A-2, A-3, B-Zero,B-1 and so on. In the process, their masks are
changed and their errata cleaned up. Intel generally
likes to reach B-Zero before declaring volume pro-duction
and recently met with disaster when it
changed that formula in its race to leapfrog AMD.
Just look at Coppermine mess. Coppermine was
reportedly an A-3 at introduction.
Valley scuttlebutt says Intel has reached a B-Zero-step
Itanium internally, even a B-1, but that those
achievements still reportedly failed to yield the
required speeds. So here’s Intel’s dilemma. Does it
pull out all the stops and Jerry rig the workarounds
needed to meet its commitment to ship even though
Linux is the most appealing operating system it can
come up with? Or does it hold fire, fix the problems
with Itanium in an orderly manner and risk missing
its long-promised, long-delayed delayed rendezvous
with destiny? Bearing in mind, of course, that that’s
just the kind of thing Wall Street can be particularly
cruel about even if Itanium isn’t supposed to be any
great revenue producer this year. And if Wall Street
isn’t feeling particularly vindictive, there’s always
Sun, even if its next chip is ridiculously late too.
Intel has reportedly kicked the problem upstairs to
its executive staff – the hallowed name of Andy
Grove is whispered – who, it is said, will be given
various scenarios to pick from including admitting
publicly that Itanium production has slipped from Q3
to Q4 at best.
Personally we’d bet they opt for the workaround
route to save face and aim to announce systems
and production chips, say, oh, maybe, in mid-October
at the e-business solutions shindig in San
Francisco that Intel’s got on the calendar. Intel
spokesmen are already laying the foundations by sug-gesting
that it wouldn’t be the least bit unusual for
customers to buy – as in pay for – systems running
“eval” or “pilot” NT software, comparing it to the
kind of big-ticket factory equipment that you’d buy
one of and run through its paces before you’d buy a
whole swat of the stuff. That kind of positioning buys
them time. They’re already saying that the preview
version of NT64 that Microsoft delivered this week at
its Professional Developer Conference in Orlando,
Florida, as we said it would (CSN No 356), is good
enough for pilots.
Meanwhile, maybe Andy and his boys can sort out the
little Cygnus problem currently besetting the
so-called former Trillian Project, now known as the
IA-64 Linux Project after a run-in with the
trademark police. Seems Cygnus, now a Red Hat
holding, is dragging its feet on GCC compilers on
the theory that Intel should be paying it for the
work while Intel is of an entirely different opinion.
Sources close to Trillian claim the compiler issue -
and let us remember how compiler-dependent
Itanium's EPIC architecture is - is the project's
biggest, single speed bump. There's little anybody can
do about SMP stability, for instance, in the absence
of final silicon.
Itanium development has taken so long that there's
not a single OEM left doing its own Itanium
two-way or four-way workstations or servers. They
all got so sick and tired of waiting for the chip to
arrive they chucked internal development. Anything
fielded, aside from the rare Itanium eight-way or
the Unisys 16/32-by, will be re-branded Intel widgetry. |
