Re: Curly soon to be out of a job
From: Bill Todd (billtodd_at_metrocast.net)
Date: 02/06/05
- Next message: Tom Linden: "Re: Curly soon to be out of a job"
- Previous message: zetasum: "One Small step one infinite leap"
- In reply to: Rob Young: "Re: Curly soon to be out of a job"
- Next in thread: Tom Linden: "Re: Curly soon to be out of a job"
- Reply: Tom Linden: "Re: Curly soon to be out of a job"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Sun, 06 Feb 2005 16:34:08 -0500
Rob Young wrote:
> In article <f-WdnfsHEfh77pjfRVn-3w@metrocastcablevision.com>, Bill Todd <billtodd@metrocast.net> writes:
>
>>Rob Young wrote:
>>
>>>In article <ZYCdnWemnaTOiJjfRVn-oQ@metrocastcablevision.com>, Bill Todd <billtodd@metrocast.net> writes:
>>>
>>>
>>>>Rob Young wrote:
>>>>
>>>>
>>>>> That paper has been around for quite some time. It is a great
>>>>> paper. But alpha is no more and it isn't as if Itanium is
>>>>> a dog.
>>>>
>>>>That's not the issue, Rob: the paper in question never asserted that
>>>>Itanic was a dog, just that its potential was unimpressive compared with
>>>>Alpha's - an assertion which certainly seems to have been borne out over
>>>>time.
>>>>
>>>
>>>
>>> True - if you quit measuring time. According to Paul, Montecito
>>> puts IPF ahead of Power5 - even Power5+.
>>
>>Which I fully expect will just be another in his long string of overly
>>optimistic projections. Hardly a good record on which to base your own
>>prognostications.
>>
>
>
> The problem with that view of course is that Montecito performance
> is already out in the wild. Certainly not overly optimistic,
> but tangible.
Oh, really? I must have missed the relevant benchmark data, but will be
most interested in seeing it.
Surely that's what you're referring to, right? Not just another set of
performance 'projections'?
>
>
>>>>Paul has some experience in the industry (though primarily in memory
>>>>rather than processor manufacturing) and a while ago parlayed that into
>>>>a position of prominence at a moderately obscure Web site. However, his
>>>>ego far exceeds his competence, and since not long after the Alphacide
>>>>he has consistently been significantly over-optimistic in predicting
>>>>Itanic's performance relative to its competition.
>>>
>>>
>>> But performance is getting consistently better and Montecito
>>> will obtain much greater performance at 100 Watts to boot.
>>
>>Well, duh: funny how the passage of time and the advance of technology
>>seem to have that effect on things. Just *how much* better has always
>>been the real question, and so far Itanic has always come up short.
>>
>
>
> But as the stock companies say: "past performance is no
> guarantee about future performance" or something to that effect.
> In the case of Itanium, this is a good thing.
>
> http://biz.yahoo.com/prnews/041216/sfth046_1.html
Er, aren't you repeating what we just saw in another post? Why don't
you instead try addressing the response I gave to it (i.e., the fact
that SGI will have to sell a comparable system *every 2 or 3 days* to
generate the revenue that Alpha systems were generating for Compaq
before the Alphacide)?
...
>> Remember that Sun is the only major OEM
>>
>>> that hasn't adopted IPF - they're going to be feeling pretty
>>> lonely and left out in a year or so.
>>
>>Or pretty smug at not having wasted the kind of resources on a sinking
>>architecture that HP and (though to far lesser extents) other Itanic
>>OEMs did.
>>
>
>
> But here the point appears to be lost on you.
Very little is lost on me, Rob. I just don't swallow every ridiculous
assertion that gets floated on the Web without analyzing it.
When you tick
> off all the OEMs that have or are adopting IPF:
>
> IBM, HP, Bull, Unisys, NEC, Fujitsu, SGI, Dell, etc.
JF has already observed that suggesting that anyone other than HP and
SGI has actually 'adopted' Itanic rather than is merely keeping their
options open in case it might actually prove worthwhile is rather a stretch.
>
> Sun is missing. Now I suppose someone could view it from your
> perspective in that Sun is the wise one. But the evidence is
> heavily tilted in the other direction, Sun is the silly one.
So far, the evidence is that Sun is the wise one: *no one* save HP and
SGI has generated Itanic sales that register *at all* when compared to
the business they obtain from other platforms. HP is a special case in
having a captive clientele (at least those who have decided to stick
with their current OS platforms and who thus *have* to begin the
transition to Itanic or face the extinction of the hardware they're
running on), and SGI is a special case not only for the same reason but
because they, unlike *anyone* else, have managed to get something
resembling leading performance out of Itanic for their rather
specialized applications (a tribute far more to the surrounding
infrastructure than to Itanic itself: after all, they achieved much the
same result using the aging MIPS platform).
>>>>> This is the wrong question. The question to ask that SKHPC answers
>>>>> is how much was lost on each Alpha CPU manufactured? It is quite
>>>>> obvious that IBM micro is losing money on every CPU they manufacture
>>>>> as IBM micro consistently turns in a loss.
>>>>
>>>>A paper loss, Rob. Just as Alpha generated significant profits
>>>>*overall* for Compaq at the system level (even after taking those
>>>>per-chip 'losses' into account), so POWER generates similar gravy for IBM.
>>>
>>>
>>> This is correct. But by the time you shift the paper profit
>>> to cover the paper loss - at the end of the day you have a much
>>> less profitable business teetering on a money losing business.
>>
>>Much less profitable than what, Rob?
>
>
> Do the math. If you are losing hundreds of millions for
> maintaining your own CPU infrastructure, billions:
>
> Included in income from continuing operations for full-year 2002 are after-tax
> charge of $433 million related to the PwCC integration and after-tax charges of
> $1.1 billion associated with the Q2 [2002] realignment of the Microelectronics
> division and productivity actions.
>
> And even post-realignment:
>
> April 16, 2004 (8:23 AM EDT)
> IBM Reports $150 Million Loss In IC Unit
>
> By Mark LaPedus, Silicon Strategies
>
> Armonk, N.Y. -- IBM Corp. on Thursday (April 15) disclosed a $150-million loss
> within its former technology unit, mostly due to ongoing chip yield issues and
> a drop in intellectual property revenues.
>
> The former technology unit includes IBM's Microelectronic Division, which
> provides ASIC products, ICs and foundry services. Recently, IBM combined its
> separate technology and systems business units into one operation.
>
> The Microelectronics Division has been in the red for some time. During a
> conference call with analysts, John Joyce, senior vice president and chief
> financial officer for IBM, said the former technology unit lost $150 million on
> a pro- forma basis in the first quarter of 2004 alone.
>
> ---
>
> You need to sell a ton of kit just to make up for the loss
> on the CPU infrastructure side.
And IBM does precisely that, plus a ton of services as well - all
generated by its leading processor architecture.
So no matter how much you try to isolate the fabrication process from
the direct consequences it leads to, when you take it all together IBM
makes a *lot* of money (billions of dollars) from the POWER
architecture. Whereas Intel has *lost* comparable amounts of money on
Itanic and will likely *never* make it back, because unlike IBM Intel
can't leverage whole-system profits even if Itanic eventually *does*
become something resembling a success for system houses (a possibility
which of course is still *very* open to debate).
But you know this, you're just
> being *clever*!
>
>
>>Intel has sunk $billions and so far only received $millions in return?
>
>
> Not according to their financial statements.
Really? Please provide the relevant numbers: specific amounts spent on
Itanic, and specific profits returned from it.
Because the number of Itanics that Intel has *produced* is insufficient
to have generated anything like $1 billion in profit (even assuming that
they hadn't given a great many of them away at fire-sale or even lower
prices as inducements), whereas a couple of years ago the NYT (among
others) estimated that something like $5 billion had already been sunk
into the project - a number which, of course, would be considerably
higher by now.
But the difference
> between Intel and IBM is obvious. One is a merchant CPU vendor,
> the other is a house CPU vendor.
Save for the minor fact that IBM is a *system* house while Intel merely
sells chips. You just don't seem to have any comprehension of the
consequences and multiple dimensions of that difference.
Intel can, and likely will, continue to make oodles of money on the
absurdly high-volume commodity chips which currently sustain it. But
Itanic seems unlikely ever to fall into that category now, and since
that was the rationale Intel used to justify undertaking its development
in the first place Itanic's future seems, shall we say, delicately
balanced at best.
IBM, meanwhile, will continue to make oodles of money out of systems -
more than enough to continue its development of POWER as long as the
POWER architecture gives it the significant edge that it currently does
(and seems likely to continue to even after your vaunted Montecito
appears, Mr. DeMone's opinion notwithstanding).
>
>
>>Being less profitable than that would be quite a challenge. In fact,
>>most corporations couldn't afford to be even significantly *more*
>>profitable than Itanic. The real question, now that Intel has been
>>forced to come up with a 64-bit x86 extension, is just how much longer
>>they will choose to be that unprofitable in an increasingly unpromising
>>cause.
>>
>
>
> Ha. And when Intel reports Itanium profitable, I guess it will
> be crow eating time?
In the unlikely event that Itanic ever gives Intel back all the money
that has been spent on developing, manufacturing, and promoting it (plus
interest, of course) - or even can reasonably be considered likely to
repay those expenditures based on its *current* generated profit and
expected lifetime (not rosy future projections thereof) - then I'll be
more than willing to eat some crow, even though merely breaking even
after such a monumental effort is not likely what Intel had in mind when
it undertook it.
...
>>You just don't get it, Rob (not that this is anything new):
>>
>>For low-end, high-volume servers, there's x86-64. For higher-end,
>>low-volume servers there's whatever you're currently using (assuming
>>your vendor is willing to continue selling it to you, as most
>>non-brain-dead vendors seem to be) - because the cost of the CPU in
>>those servers is such a minor component of their overall cost of
>>ownership that any savings from using a merchant CPU would be lost in
>>the noise (whereas the costs of *moving* to another hardware or software
>>architecture would be significant indeed).
>>
>
>
> You say all that, skipping over the fact that IBM is losing
> big time in their micro-electronics division.
You keep saying that, Rob, while studiously ignoring its silliness. I'm
really beginning to believe that you're a write-only 'bot.
Now you
> trot out x86-64 as it is some sort of edge. Here's something
> to consider, what percentage of the server market is Opteron?
You seem to be confusing Opteron with x86-64, Rob. Itanic's problem is
that this (admittedly so far relatively minor in terms of actual market
share) threat to Xeon has prompted Intel to produce 64-bit-capable Xeons
as well - making a very tough Itanic sell in the low end (and even some
of the mid-range) even tougher.
Just keep whistling past the graveyard, Rob: it's what you're best at.
- bill
- Next message: Tom Linden: "Re: Curly soon to be out of a job"
- Previous message: zetasum: "One Small step one infinite leap"
- In reply to: Rob Young: "Re: Curly soon to be out of a job"
- Next in thread: Tom Linden: "Re: Curly soon to be out of a job"
- Reply: Tom Linden: "Re: Curly soon to be out of a job"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
|
