New Itanium chips cost just $744

Tony Hill wrote:
Yup, Intel's only sold roughly 20,000 Itanium chips TOTAL since it's
introduction, that's counting every different model they've released.
Given a rough estimate of around $5 billion to develop, manufacturer,
test and market the Itaniums (this is probably a low estimate),
they're looking at a per-chip cost of somewhere on the order of
$250,000 per chip :

Given that intel came out and said (last year) that they spent over a
billion dollars (close to $2B) up to that point, I have to agree with
you that your measurement is pessimistic. You are suggesting they spent
more than $3B in a single year where they spent under $2B in the
previous 8 years.

Alex
--
My words are my own. They represent no other; they belong to no other.
Don't read anything into them or you may be required to compensate me
for violation of copyright. (I do not speak for my employer.)

"CJT" wrote in message
...
Rob Stow wrote:

Robert Myers wrote:

On Tue, 9 Sep 2003 22:22:36 -0400, Keith R. Williams
wrote:


In article ,
says...

On Tue, 09 Sep 2003 06:06:43 GMT, CJT wrote:

snip

Assuming they sell overseas, at what point (if any) does selling far
below cost become "dumping" and perhaps cause problems with, e.g.,
the WTO?


If manufacturers were not permitted to introduce new products and sell
them at a loss, we'd be stuck in a world with practically no new
products.


Indeed!


In order to show that Intel was "dumping" Itaniums, you'd have to show
that someone was losing business because of it, and that would be a
very tough sell. The only people who are going to buy Itaniums are
people who, for one reason or another, need a chip like Itanium.


1) That's not dumping (A US manufacturer cannot "dump")
2) Selling at a loss is not illegal
3) even if someone else loses money



I'm not an expert on it, but I believe that the EC has protections
against predatory pricing similar to those in the US. The US can't
stop others from dumping, but it can take retaliatory action in the
form of tariffs, and I assume the EC works the same way.

In any case, I don't think anybody is going to go to court in the EC
over Itanium.

Intel has very deep pockets, it has its heels dug in, and it intends
to force IA-64 onto the market one way or another. The chip they wind
up selling in large quantities may bear very little resemblance to the
original EPIC concept, but, barring the apocalypse, Intel intends to
make IA-64 its predominant instruction set one way or another.


The Itanic pricing might just possibly be evidence that Intel is
attempting to dump Itanics on the market. However, failing
spectacularly at an attempt to dump a product is not likely to
trigger punitive tarriffs from anyone :-)
Laughs - yes, tariffs - no.

Note also that not all anti-dumping laws are based on the US model.
In Canada, for example, our anti-dumping legislation can come into
play if one imported product is dumped on the market at the expense
of a competing imported product - even if there are no competing
domestic products. I vaguely recall this coming into play about
30 years ago when California sparkling wines were dumped on Canada
at the expense of Champagne and other imported sparkling wines.


It just seemed to me that selling at far below cost to capture sales is
a market distortion that might attract some sort of attention.


Every new processor is sold far below cost.
Particularly the very first unit sold.

--

... Hank

Hank: http://horedson.home.att.net
W0RLI: http://w0rli.home.att.net

On Wed, 10 Sep 2003 14:21:56 GMT, "Hank Oredson"
wrote:


"CJT" wrote in message
...

snip


It just seemed to me that selling at far below cost to capture sales is
a market distortion that might attract some sort of attention.


Every new processor is sold far below cost.
Particularly the very first unit sold.

It would be *really* interesting to know anything hard at all about
the what the cost of manufacture (not including R&D) for Itania would
be if Intel ever got up to decent size production runs (could we skip
the obligatory comments on that, just this once?).

Off-hand comments from someone who should know suggest that the heat
sink is a significant cost-contributor for Madison. Since Deerfield
consumes about 60 watts as opposed to the 160 watts for Madison, the
bottom line cost of manufacture for Deerfield has to be significantly
less than for Madison.

Intel can probably afford practically to give away Deerfields. It
might even want to, but irrespective of any laws, they can't afford to
distort the established market for their high-end Xeons, and I suspect
that's how the price was really set--just high enough so that
Deerfield isn't an attractive competitor for Xeon.

Aside from workstation users who really need the bigger address space
and who for one reason or another can't or won't buy an Opteron, the
only people who will be buying Deerfields are people who are buying
them for evaluation or development. I can't imagine that the buyers
who always want the best and the fastest would be interested in a
Deerfield.

RM

On Wed, 10 Sep 2003 10:59:18 -0400, Robert Myers
wrote:
On Wed, 10 Sep 2003 14:21:56 GMT, "Hank Oredson"
wrote:
Every new processor is sold far below cost.
Particularly the very first unit sold.

It would be *really* interesting to know anything hard at all about
the what the cost of manufacture (not including R&D) for Itania would
be if Intel ever got up to decent size production runs (could we skip
the obligatory comments on that, just this once?).

I would concur with Keith's earlier guess that it's probably around
$75. Maybe $100, but probably no more than that.

Off-hand comments from someone who should know suggest that the heat
sink is a significant cost-contributor for Madison. Since Deerfield
consumes about 60 watts as opposed to the 160 watts for Madison, the
bottom line cost of manufacture for Deerfield has to be significantly
less than for Madison.

Madison may be a hot running chip, but it's no where near THAT hot.
The top-end Madison at 1.5GHz and 6MB of L3 cache has a maximum power
consumption of about 110W. Intel currently specifies that
motherboards and power supplies should be capable of up to 130W to
support future Itanium2 chips, but none of the current chips consume
that much.

Deerfield, on the other hand, consumes 91W of power at 1.4GHz. The
Low Voltage Itanium2 (1.0GHz Deerfield, 1.5MB L3) is listed as having
a maximum power consumption of 55W in the datasheets and 62W
everywhere else (including on Intel's own marketing material), so I'm
not quite sure what's going on there. It could be that the 62W is a
maximum theoretical power consumption while the 55W is a maximum
real-world power consumption, or that the 62W number is leaving a bit
of wiggle-room for future processors.

Either way, when you get right down to it, I really doubt that the
heatsink plays much of a role to the cost of the processor. Even
quite a large heatsink is only going to cost you about $10 when you
get them manufactured under contract.

Intel can probably afford practically to give away Deerfields. It
might even want to, but irrespective of any laws, they can't afford to
distort the established market for their high-end Xeons, and I suspect
that's how the price was really set--just high enough so that
Deerfield isn't an attractive competitor for Xeon.

I'm not sure that Intel is too worried about stealing sales from Xeon
at this point in time. After all, the Xeon has a competitor that runs
the same software (Opteron), while if people switch to Deerfield, they
are then locked into IA-64 software which is Intel-only. However
Intel does still want to make some profit on this Deerfield chip,
since they haven't come anywhere close to recouping their costs on
IA-64 yet.

Aside from workstation users who really need the bigger address space
and who for one reason or another can't or won't buy an Opteron, the
only people who will be buying Deerfields are people who are buying
them for evaluation or development. I can't imagine that the buyers
who always want the best and the fastest would be interested in a
Deerfield.

The Deerfield is squarely targeted at the workstation market. I don't
think Intel has any plans to sell this chip in desktop systems for
Joe-six-pack. People will buy Deerfield for workstations if they need
the address space and either a.) the software is there for the Itanium
but not for Opteron, or b.) HP offers a more compelling Deerfield
workstation than what other companies are offering in Opteron
workstations.

--------------
Tony Hill
hilla underscore 20 at yahoo dot ca

On Wed, 10 Sep 2003 08:07:06 -0400, Alex Johnson
wrote:
Tony Hill wrote:
Yup, Intel's only sold roughly 20,000 Itanium chips TOTAL since it's
introduction, that's counting every different model they've released.
Given a rough estimate of around $5 billion to develop, manufacturer,
test and market the Itaniums (this is probably a low estimate),
they're looking at a per-chip cost of somewhere on the order of
$250,000 per chip :

Given that intel came out and said (last year) that they spent over a
billion dollars (close to $2B) up to that point, I have to agree with
you that your measurement is pessimistic. You are suggesting they spent
more than $3B in a single year where they spent under $2B in the
previous 8 years.

I've seen numbers that said they had spent $2 billion by 2001, but
this seems like a very low number to me, and I suspect it doesn't take
into count any costs associated with producing the chips, testing,
optimizing and especially the costs to market the processor. There
are a lot of associated costs, and I can't imagine that Intel managed
to fit all of those into only $2 billion in 7 years of development.
Of course, since 2001 they have spent more money still, I would guess
at a rate of about $1 billion a year.

Regardless, even if we assume that they've ONLY spent $2 billion on
the Itanium so-far, they still are not even remotely close to breaking
even on sales so far.

--------------
Tony Hill
hilla underscore 20 at yahoo dot ca

Tony Hill wrote:
I've seen numbers that said they had spent $2 billion by 2001, but
this seems like a very low number to me
??? $2B is not a small amount of money. Considering what most companies
spend on chip development $2B would be unmanageable. While intel is
horrendously wasteful, I can't think that $2B would not be a large
enough estimate.

and I suspect it doesn't take
into count any costs associated with producing the chips, testing,
optimizing and especially the costs to market the processor. There
are a lot of associated costs, and I can't imagine that Intel managed
to fit all of those into only $2 billion in 7 years of development.
Manufacturing is not that expensive (okay a few hundred thousand to a
million per stepping, but that won't add up much since they've only sent
Merced, McKinley, and Madison to manufacturing and all of these shipped
in a tenth the steppings that Pentium 4 did). Testing is included. I
don't know what optimizing you refer to, but payroll of compiler teams
and such was included. Marketting. You make me laugh! They've only
been advertising 2 years and they spent only $10,000,000 the first and
busiest year (compared to $2 Billion a year for Pentium 4).

Of course, since 2001 they have spent more money still, I would guess
at a rate of about $1 billion a year.
Where could they put the money? 2002 featured only one chip (a simple
shrink) going to manufacturing, very limited advertising, and fewer
projects under way than 2001.

Regardless, even if we assume that they've ONLY spent $2 billion on
the Itanium so-far, they still are not even remotely close to breaking
even on sales so far.
But this point is not in contention. It's a good thing some people have
long term vision.

Alex
--
My words are my own. They represent no other; they belong to no other.
Don't read anything into them or you may be required to compensate me
for violation of copyright. (I do not speak for my employer.)

"Alex Johnson" wrote in message
...

Manufacturing is not that expensive (okay a few hundred thousand to a
million per stepping, ...Testing is included.

Alex, the outsourced mask costs alone are $1 million for each stepping.
I suggest you revize your estimate upwards. WAY upwards!

"Felger Carbon" wrote in news:2TL7b.6423$PE6.2870
@newsread3.news.pas.earthlink.net:

"Alex Johnson" wrote in message
...

Manufacturing is not that expensive (okay a few hundred thousand to a
million per stepping, ...Testing is included.

Alex, the outsourced mask costs alone are $1 million for each stepping.
I suggest you revize your estimate upwards. WAY upwards!



That's $1M for a full mask set on 130nm. It was much cheaper on 180nm,
where most of the Itanium steppings have been, and typically you don't need
a full mask set per stepping of the same die.
So I'd say Alex was pretty close.

In article ,
says...
On Tue, 9 Sep 2003 22:22:36 -0400, Keith R. Williams
wrote:

In article ,
says...
On Tue, 09 Sep 2003 06:06:43 GMT, CJT wrote:

snip

Assuming they sell overseas, at what point (if any) does selling far
below cost become "dumping" and perhaps cause problems with, e.g.,
the WTO?

If manufacturers were not permitted to introduce new products and sell
them at a loss, we'd be stuck in a world with practically no new
products.

Indeed!

In order to show that Intel was "dumping" Itaniums, you'd have to show
that someone was losing business because of it, and that would be a
very tough sell. The only people who are going to buy Itaniums are
people who, for one reason or another, need a chip like Itanium.

1) That's not dumping (A US manufacturer cannot "dump")
2) Selling at a loss is not illegal
3) even if someone else loses money

I'm not an expert on it, but I believe that the EC has protections
against predatory pricing similar to those in the US. The US can't
stop others from dumping, but it can take retaliatory action in the
form of tariffs, and I assume the EC works the same way.

Perhaps, but we (I thought) were talking abut US "anti-dumping"
laws. The EU is certainly *different*.

In any case, I don't think anybody is going to go to court in the EC
over Itanium.

One never knows about the EU. ;-)

Intel has very deep pockets, it has its heels dug in, and it intends
to force IA-64 onto the market one way or another. The chip they wind
up selling in large quantities may bear very little resemblance to the
original EPIC concept, but, barring the apocalypse, Intel intends to
make IA-64 its predominant instruction set one way or another.

It *wants*. The market doesn't necessarily want what Intel
*WANTS* them to have. That was the fallacy of the Itanic five
years ago. MDR told everyone to close up shop. "No one could
compete against the Intel/HP juggernaut." Many believed it for a
time, but the proposition was simply silly on the face of it.

--
Keith

In article ,
says...
Robert Myers wrote:
On Tue, 9 Sep 2003 22:22:36 -0400, Keith R. Williams
wrote:


In article ,
says...

On Tue, 09 Sep 2003 06:06:43 GMT, CJT wrote:

snip

Assuming they sell overseas, at what point (if any) does selling far
below cost become "dumping" and perhaps cause problems with, e.g.,
the WTO?

If manufacturers were not permitted to introduce new products and sell
them at a loss, we'd be stuck in a world with practically no new
products.

Indeed!


In order to show that Intel was "dumping" Itaniums, you'd have to show
that someone was losing business because of it, and that would be a
very tough sell. The only people who are going to buy Itaniums are
people who, for one reason or another, need a chip like Itanium.

1) That's not dumping (A US manufacturer cannot "dump")
2) Selling at a loss is not illegal
3) even if someone else loses money


I'm not an expert on it, but I believe that the EC has protections
against predatory pricing similar to those in the US. The US can't
stop others from dumping, but it can take retaliatory action in the
form of tariffs, and I assume the EC works the same way.

In any case, I don't think anybody is going to go to court in the EC
over Itanium.

Intel has very deep pockets, it has its heels dug in, and it intends
to force IA-64 onto the market one way or another. The chip they wind
up selling in large quantities may bear very little resemblance to the
original EPIC concept, but, barring the apocalypse, Intel intends to
make IA-64 its predominant instruction set one way or another.


The Itanic pricing might just possibly be evidence that Intel is
attempting to dump Itanics on the market. However, failing
spectacularly at an attempt to dump a product is not likely to
trigger punitive tarriffs from anyone :-)
Laughs - yes, tariffs - no.

;-)))

Note also that not all anti-dumping laws are based on the US model.
In Canada, for example, our anti-dumping legislation can come into
play if one imported product is dumped on the market at the expense
of a competing imported product - even if there are no competing
domestic products. I vaguely recall this coming into play about
30 years ago when California sparkling wines were dumped on Canada
at the expense of Champagne and other imported sparkling wines.

Wierd! Is the purpose to float the importer of the turkey?
PLease, let the people decide what they want to buy. Wine isn't
really a commodity as, say DRAM is. I could care less what DRAM
I have (as long as it works). I do prefer some wines over
others. Selling under cost to gain market share isn't an unknown
strategy. Again I must ask, what is a silly law like this trying
to prove?

--
Keith