Intel: The chipset is the product

  • Thread starter Thread starter Robert Myers
  • Start date Start date
Tony Hill wrote:

[SNIP]
Intel solution. For the moment their only solutions in this market at
the e7505 chipset (limited to 533MT/s bus speeds for now at least) and
the i875P (no PCI-X support and limited memory capacity for a server).

IBM and Unisys build their own chipsets, and they are pretty meaty. I
would like to see a comparison between one of them fancy 4P in-house
chipset boxes and an Opteron box. IBM reckon they can go to 32 way,
believe it when I see it I guess. :P

Cheers,
Rupert
 
Tony Hill said:
Take a look at the 2-way and greater servers from all the major OEMs.
HPaq doesn't have a single Intel chipset in the bunch, all Serverworks
for 2 and 4 way with their own customer job for 8-way setups. IBM is
pretty much the same story. Dell, forever the Intel stalwart, has
something like 1 or 2 of their 2-way servers using Intel chipsets, but
the bulk use Serverworks and all of their 4-way servers are
Serverworks chipsets.

However now Intel has declined Serverworks license for future
chipsets, meaning that all of those servers from all of the major OEMs
need to switch to an Intel chipset for future designs. What's even
worse though, there is no Intel chipset for them to switch to! Intel
has yet to release a 4-way (or greater) chipset for their P4-style
Xeons.

In short, Intel is largely shooting themselves in the foot. Their
performance in the 4P server market absolutely stinks vs. the Opteron,
largely because they are limited to 4 processors sharing a 400MT/s
bus. They can't increase that, not because they don't have the
processors for it but because they don't have their own chipset and
refuse to let Serverworks build one for them. Even in 2-way servers,
where the margin by which the Opteron beats them is slightly less
embarrassing, they're still stuck at a 533MT/s bus speed and forcing
all their customers to trash existing designs in favor of an untested
Intel solution. For the moment their only solutions in this market at
the e7505 chipset (limited to 533MT/s bus speeds for now at least) and
the i875P (no PCI-X support and limited memory capacity for a server).

Well, as you pointed out, even though they are denying Serverworks any
further licenses for Xeon, IBM and HP don't use Serverworks for their 8-way
and greater chipsets, they use their own. So perhaps there is still a couple
of chipsets for Dell to use if they want 4-way or greater Xeons -- they'll
have to buy them from HP or IBM. :-)

Yousuf Khan
 
Effective for consumer chipsets, sure, but this whole discussion
started with the high-end server chipsets where Intel has been failing
miserably for 5 years and is now looking to become the ONLY supplier
in the business.

I'll admit I'm speaking from a consumer perspective and don't really
pay much attention to the high-end chipsets. From the perspective of
the original Grantsdale thread, and Intel publicizing the chipset,
which is traditionally only geek turf, the strategy comments are still
valid, I believe.

This bunch knows much more about server chips and chipsets than I do,
so I just read those threads and absorb the info.


Neil Maxwell - I don't speak for my employer
 
...long ago in a Galaxy far, far, away. Intel went into the
chipset business to promote PCI (and because the rest sucked
bilge-water). They then stumbled badly too (RZ1000?).

This is the "guarantees compatibility and reliability" part; they
could make their own decisions about how to make it if they were
unhappy with what was available. As I said, they made some mistakes,
and doubtlessly will make more in the future, just as all chipset
manufacturers do. They had their successes as well, such as the
430HX, which was quite a good chipset for its day.
Work*ED* is the operative. Scaring away chipset builders is
*DUMB*, I don't care who your employer is. There is no money to
pump here. ...you're already doing all you can.

Exactly the point. There's no money there, and if Intel's management
is dumb, it's a very profitable form of dumb. Therefore, there must
be some other logical reason why they do this, and I believe it's as I
have stated.

Intel is notoriously paranoid about hitching their wagon to other
people's non-commodity horses. The last time they did this was with
Rambus, and look what happened. Again, without a guaranteed supply of
quality chipsets, sales of CPUs could suffer. It's a tradeoff, like
all business decisions, and their direction is pretty clear.

In any case, their system still seems to be working well; the 865/875
chipsets are, by all reports, robust, reliable, and good performers,
if more expensive than the competition. I don't know what proportion
of Intel-based systems have Intel chipsets, though. Certainly, many
of the most popular P4 motherboards are built on these chipsets.

Likewise, some AMD systems get a bad rap due to cruddy chipsets. It's
not the CPU's fault, but the non-technical people don't differentiate
between them; they have no idea whether there's a Via or SIS inside
there, just that the system is unstable.
That was true ten years ago. Hell even graphics chips were done
on antique lines ten years ago. Things change. Modern chipsets
have the same issues as processors today.

I didn't say anything about antique lines. I don't know about other
companies, but where I work, chipsets are made in older factories with
more depreciated equipment (as is flash), some of which are still
making CPUs. They're still very modern, but the cutting-edge 300mm
lines are primarily cranking out cutting-edge CPUs. This is basic
resource management; it only makes sense from a cost-per-die
perspective, particularly on low-margin products.
"Effective" is a fungible term. Building chipsets isn't
profitable, though *sometimes* necessary. Pissing those off
building the unprofitable parts for *YOU* is *DUMB*.

Effective from the point of view of allowing them to sell the CPUs
without reliance on external vendors, as I stated. This is a primary
goal, as far as I can tell (but I have no insider information on
this). It's certainly not about making money; I'd be surprised if an
entire year's chipset production makes as much as a few weeks worth of
CPUs. I'll say it again - it's a tool, just like any tool, with costs
and tradeoffs.

There are alternative chipsets available, of course, and they're
mostly cheaper, though I prefer the mature versions of the Intel
chipsets for stability. YMMV, as always, and there are lots of other
choices for everyone.
Of course you Intel folks may have some other motive.

What other motive do you believe Intel has, beyond selling as many
CPUs as they can?

Why do you imply that I have some sort of motive or influence beyond
working for a large corporation when I'm trying to discuss things
rationally and from my perspective as a hobbyist? Do you really
believe that all 80K+ people who work at Intel are corporate clones,
pushing the company agenda at all costs?

I understand that this is holy war material for many people, and I'm
not interested. If you want to contradict my conclusions with
experience or data, I'm ok with that. If you want to throw stones for
emotional reasons, I'll pass, thanks anyway.


Neil Maxwell - I don't speak for my employer
 
Neil said:
This is the "guarantees compatibility and reliability" part; they
could make their own decisions about how to make it if they were
unhappy with what was available. As I said, they made some
mistakes, and doubtlessly will make more in the future, just as
all chipset
manufacturers do. They had their successes as well, such as the
430HX, which was quite a good chipset for its day.

Ok, now how does this relate to telling ServerWorks to "go screw"?
The fact is that Intel is no better than the rest (remember the
TX?). Intel has made more mistakes than the law should allow!
Fortunately, for their employees, they have tons of cash and can
afford to piss off their customers. If you think that situation
can last...
Exactly the point. There's no money there, and if Intel's
management
is dumb, it's a very profitable form of dumb. Therefore, there
must be some other logical reason why they do this, and I believe
it's as I have stated.

Chipsets are *NOT* a "very profitable form of dumb". It's simply
*DUMB* to produce what others can do easily and won't contribute to
the bottom line. Sell the antique line-space to the competition!
BTW, chipsets, like graphics widgets, aren't so much the users of
garbage processes anymore. These things are getting as
sophisticated as the processors.
Intel is notoriously paranoid about hitching their wagon to other
people's non-commodity horses. The last time they did this was
with
Rambus, and look what happened.

That was predicted *here*, if you care to read the archives.
Absolute arrogance is a guarantee of disaster, sooner or later.
Fortunately the RMBS debacle was sooner.
Again, without a guaranteed
supply of
quality chipsets, sales of CPUs could suffer. It's a tradeoff,
like all business decisions, and their direction is pretty clear.

Certainly it's clear. The PHB's thought process is the only thing
that's murky. ...cut off your dick because the wife has a
headache?
In any case, their system still seems to be working well; the
865/875 chipsets are, by all reports, robust, reliable, and good
performers,
if more expensive than the competition. I don't know what
proportion
of Intel-based systems have Intel chipsets, though. Certainly,
many of the most popular P4 motherboards are built on these
chipsets.

I wish you well. Keep thinking sweet thoughts. Your leaders are
driving you off a cliff, me thinks.
Likewise, some AMD systems get a bad rap due to cruddy chipsets.
It's not the CPU's fault, but the non-technical people don't
differentiate between them; they have no idea whether there's a
Via or SIS inside there, just that the system is unstable.

<shrug> I never had any problems with Via chipsets (at least no
more than others). I do know enough to get appropriate drivers.
I've always preferred SiS though. Now with nVidia in the market
there are *many* alternatives to the Intel stuff.

If Intel thinks they can take the PC "private" again, they should
remember the RMBS fiasco. ...and Itanic! (though I"m certain they
still believe).

I didn't say anything about antique lines. I don't know about
other companies, but where I work, chipsets are made in older
factories with more depreciated equipment (as is flash), some of
which are still
making CPUs. They're still very modern, but the cutting-edge
300mm
lines are primarily cranking out cutting-edge CPUs. This is basic
resource management; it only makes sense from a cost-per-die
perspective, particularly on low-margin products.

So you're contradicting yourself?
Effective from the point of view of allowing them to sell the CPUs
without reliance on external vendors, as I stated. This is a
primary goal, as far as I can tell (but I have no insider
information on
this). It's certainly not about making money; I'd be surprised if
an entire year's chipset production makes as much as a few weeks
worth of
CPUs. I'll say it again - it's a tool, just like any tool, with
costs and tradeoffs.

So you admit that it makes good business sense to make widgets on
*expensive* tools that could be used for more *profitable*
products? Further, it makes good business sense to force a good
business partner out of your market, into the competition's, even
when you don't have a pot to piss in?

You Intel folks are strange.
There are alternative chipsets available, of course, and they're
mostly cheaper, though I prefer the mature versions of the Intel
chipsets for stability. YMMV, as always, and there are lots of
other choices for everyone.

Of course think as you do. Your paycheck comes from *INTEL*.
Sheesh!
What other motive do you believe Intel has, beyond selling as many
CPUs as they can?

What motive do you think M$ has? Think hard.

The *fact* is that Intel has tried to artificially segment and
restrict the market many times. Fortunately, their track record is
worse than M$'.
Why do you imply that I have some sort of motive or influence
beyond working for a large corporation when I'm trying to discuss
things

I don't think *you* have any influence. You're simply guided by the
"party-line". I started questioning party-line many years ago,
whin it didn't pass the laugh-test. You should learn the same
trick, but please do keep quiet at work. Independent thinkers
aren't wanted in such organizations. ;-)
rationally and from my perspective as a hobbyist? Do you really
believe that all 80K+ people who work at Intel are corporate
clones, pushing the company agenda at all costs?

No, in fact I know *one* who isn't. He's there because he's treated
quite well, and takes advantage of the situation. According to my
friend (his dad) he's not a happy-camper either. No, I just think
you've put the corporate blinkers on.
I understand that this is holy war material for many people, and
I'm
not interested. If you want to contradict my conclusions with
experience or data, I'm ok with that. If you want to throw stones
for emotional reasons, I'll pass, thanks anyway.

You're the emotional one! Please explain Intel's action against
ServerWorks in any rational language! They're pissed at BroadCom,
evidently so are going to saw off their own privates to make a
statement. Well, they tried the same thing with Itanic, and that's
worked just *soo* well.
Neil Maxwell - I don't speak for my employer

You seem to be doing a good job trying. (hint: do you know what a
signature separator is?)
 
Ok, now how does this relate to telling ServerWorks to "go screw"?
The fact is that Intel is no better than the rest (remember the
TX?). Intel has made more mistakes than the law should allow!
Fortunately, for their employees, they have tons of cash and can
afford to piss off their customers. If you think that situation
can last...

It doesn't relate to Serverworks at all; we were discussing the broad
chipset market and strategy, AFAICT. Fortunately for everyone, the
law has nothing to do with it, and the situation will either correct
itself or they will crash and burn. We'll just have to see how it
lasts.
Chipsets are *NOT* a "very profitable form of dumb". It's simply
*DUMB* to produce what others can do easily and won't contribute to
the bottom line. Sell the antique line-space to the competition!
BTW, chipsets, like graphics widgets, aren't so much the users of
garbage processes anymore. These things are getting as
sophisticated as the processors.

Again, I didn't mention "garbage processes"; these are the same
processes that were producing the last few generation's cutting edge
chips. They're just not on the edge any more. Typically, older
equipment gets updated and re-used to maximize capital life and
minimize disruptive upgrades to the fabs. This allows them to
implement their strategies of producing less demanding technologies at
reduced costs. Certainly the chipsets are sophisticated, but the
latest processors are further ahead, IME.

Really, it boils down to a strategic decision - rely on outside
manufacturers or do it yourself; each has their associated costs and
benefits. I've already been over the logic there, as I see it. You
obviously don't agree; we'll see if the strategy flops in the future.

BTW, I don't have much familiarity with server chipsets; are they as
low margin as mass-market chipsets? I'm sure the volume is far lower,
but if margins are higher, it changes the dynamics of the strategy.

On a side (but related) topic, Intel's never made money to speak of on
flash, yet they persist in it as a strategic direction. One side
effect of their cash flow is that they can invest in low-profit
segments if it fits into their long-term plans or enables their CPU
sales.
That was predicted *here*, if you care to read the archives.
Absolute arrogance is a guarantee of disaster, sooner or later.
Fortunately the RMBS debacle was sooner.

I don't need to read the archives, I was here, and in agreement. I
agree with you on this as well - corporate arrogance is very risky.
It was a stumble that would have killed many smaller companies.
Certainly it's clear. The PHB's thought process is the only thing
that's murky. ...cut off your dick because the wife has a
headache?

Sorry, I'm having a hard time mapping this to chipset strategy.
I wish you well. Keep thinking sweet thoughts. Your leaders are
driving you off a cliff, me thinks.

Well, they're still cutting the paychecks, and that's a Good Thing,
for me at least. If they collapse, I'll retire or find something else
to do, but I don't believe it will happen too soon. Time will tell.

In any case, it's not like a religious commitment or anything; they're
just a corporate employer.
<shrug> I never had any problems with Via chipsets (at least no
more than others). I do know enough to get appropriate drivers.
I've always preferred SiS though. Now with nVidia in the market
there are *many* alternatives to the Intel stuff.

Alternatives are a good thing. I don't have first-hand experience
with Via in recent years, but they seem to be the one everyone
complains about.
If Intel thinks they can take the PC "private" again, they should
remember the RMBS fiasco. ...and Itanic! (though I"m certain they
still believe).

Again, time will tell.
So you're contradicting yourself?

I don't think so. Typically, there are 3 generations of fabs running
high volume at any given time, and that's true today. Lower-end,
lower-margin products (including older CPUs) run on the older lines,
and the high-speed, high-margin stuff runs on the latest fabs. I'm
sure AMD runs in a similar fashion, though they have fewer fab lines,
so probably have less spread between the ends. I'm not real familiar
with their capacity or their fab designs, but I'm sure they don't shut
down their older capacity when new fabs come online.
So you admit that it makes good business sense to make widgets on
*expensive* tools that could be used for more *profitable*
products? Further, it makes good business sense to force a good
business partner out of your market, into the competition's, even
when you don't have a pot to piss in?

That decision has certainly been made. Wafer fab starts and product
mix are pretty flexible over the medium term. Demand for products vs.
potential margins vs. fab capability vs. die costs are all weighed to
determine how many wafers to start on what lines. Yes, you could make
more CPUs on the lines that make flash and chipsets, but this would
only make sense if you were unable to make enough on the other lines.

Of course, many of the older fabs are incapable of making the latest
CPUs, but will turn out flash and chipsets quite nicely, with very
good yields, since the performance was fine-tuned while they were
making high-margin CPUs under intense scrutiny.

I'm not a policy guy, but I'd have to assume that the most sacred cow
is the top few tiers of CPU; if demand can't be met, other areas will
be re-prioritized to meet it, since that's where the cash flow is.
I've seen this happen in the past.

Throw in the strategic decisions and contractual commitments for the
widgets, and it's only a difficult decision if you're short of fab
capacity. IME, Intel plans pretty hard to avoid such situations.

Again, choosing to make low-margin products to enable high-margin
products just a cost of doing business.
You Intel folks are strange.

All 80K of us, eh? I used to work at AMD. Did I only get strange
when I changed jobs? Will I become normal again if I quit? Does it
help that I have AMD CPUs as well as Intel?
Of course think as you do. Your paycheck comes from *INTEL*.
Sheesh!

I do have an ulterior motive, but it's pretty dependent on
cost-effectiveness and functionality. I don't have any Intel network
gear, nor do I buy cell phones based on whether or not they have Intel
flash in them (despite the fact that I manufacture flash). Working
for a company (any company) doesn't remove the capacity for
independent thought.
What motive do you think M$ has? Think hard.

You're the one that suggested another motive. I believe it's as I
stated. I'm not sure what Microsoft has to do with chipset marketing
motives.
The *fact* is that Intel has tried to artificially segment and
restrict the market many times. Fortunately, their track record is
worse than M$'.

They certainly have, and will continue to try in the future, I'm sure.
This doesn't change the fact that their chipset strategies seem to
have worked well for them historically. Whether it will bite them in
the future, as you suggest, remains to be seen.
I don't think *you* have any influence. You're simply guided by the
"party-line". I started questioning party-line many years ago,
whin it didn't pass the laugh-test. You should learn the same
trick, but please do keep quiet at work. Independent thinkers
aren't wanted in such organizations. ;-)

You've made a lot of assumptions about me, based on a post about
chipset history. What's my stance on HT, on frequency vs. model
numbers, on Itanium, and other "party-line" issues? It may come as a
surprise to you, but Intel does no mass employee indoctrination on
what kind of chips or designs are superior. Most people who work here
haven't got a clue about the details of CPUs and chipsets, IME.
No, in fact I know *one* who isn't. He's there because he's treated
quite well, and takes advantage of the situation. According to my
friend (his dad) he's not a happy-camper either. No, I just think
you've put the corporate blinkers on.

I'm unsure how you've come to this conclusion, and why you think my
situation is any different from your friend's son. I'm not spouting
pro-Intel sentiment here, just posting my analysis on chipset
marketing strategies (which I still believe are correct).
You're the emotional one! Please explain Intel's action against
ServerWorks in any rational language! They're pissed at BroadCom,
evidently so are going to saw off their own privates to make a
statement. Well, they tried the same thing with Itanic, and that's
worked just *soo* well.

As I've said, I have no expertise in server chipsets or their
marketing strategies. I don't believe I've defended Intel's strategy
in this area (since I rarely venture opinions on things I don't know
about). I simply posted my own perception of Intel's broader chipset
strategy (which is what started the thread), since no one else had
touched on that point. You then accused me of agendas based on my
email address. I fail to see any emotional points in my posts, but
feel free to point them out for me.
You seem to be doing a good job trying.

Heh! You don't seriously believe that my post on a rather obvious
aspect of historic chipset strategy is somehow advancing the purposes
of the megacorp Intel, do you?
(hint: do you know what a signature separator is?)

Any suggestions on spelling or grammar while you're at it?



Neil Maxwell - I don't speak for my employer
 
Robert said:
What you would attribute to cleverness on AMD's part, I would attribute
to necessity. Intel has the cash and the need to find applications for
what it does best (making silicon); AMD doesn't have the cash, and it
doesn't make silicon.

That's incorrectly stated. AMD does make silicon, of course, but it has
limited capacity and no need to look for things to do with the silicon
it can make.

RM
 
K Williams said:
Ok, now how does this relate to telling ServerWorks to "go screw"?
The fact is that Intel is no better than the rest (remember the
TX?). Intel has made more mistakes than the law should allow!
Fortunately, for their employees, they have tons of cash and can
afford to piss off their customers. If you think that situation
can last...


Chipsets are *NOT* a "very profitable form of dumb". It's simply
*DUMB* to produce what others can do easily and won't contribute to
the bottom line. Sell the antique line-space to the competition!
BTW, chipsets, like graphics widgets, aren't so much the users of
garbage processes anymore. These things are getting as
sophisticated as the processors.


That was predicted *here*, if you care to read the archives.
Absolute arrogance is a guarantee of disaster, sooner or later.
Fortunately the RMBS debacle was sooner.


Certainly it's clear. The PHB's thought process is the only thing
that's murky. ...cut off your dick because the wife has a
headache?


I wish you well. Keep thinking sweet thoughts. Your leaders are
driving you off a cliff, me thinks.


<shrug> I never had any problems with Via chipsets (at least no
more than others). I do know enough to get appropriate drivers.
I've always preferred SiS though. Now with nVidia in the market
there are *many* alternatives to the Intel stuff.

If Intel thinks they can take the PC "private" again, they should
remember the RMBS fiasco. ...and Itanic! (though I"m certain they
still believe).



So you're contradicting yourself?


So you admit that it makes good business sense to make widgets on
*expensive* tools that could be used for more *profitable*
products? Further, it makes good business sense to force a good
business partner out of your market, into the competition's, even
when you don't have a pot to piss in?

You Intel folks are strange.


Of course think as you do. Your paycheck comes from *INTEL*.
Sheesh!


What motive do you think M$ has? Think hard.

The *fact* is that Intel has tried to artificially segment and
restrict the market many times. Fortunately, their track record is
worse than M$'.


I don't think *you* have any influence. You're simply guided by the
"party-line". I started questioning party-line many years ago,
whin it didn't pass the laugh-test. You should learn the same
trick, but please do keep quiet at work. Independent thinkers
aren't wanted in such organizations. ;-)


No, in fact I know *one* who isn't. He's there because he's treated
quite well, and takes advantage of the situation. According to my
friend (his dad) he's not a happy-camper either. No, I just think
you've put the corporate blinkers on.


You're the emotional one! Please explain Intel's action against
ServerWorks in any rational language! They're pissed at BroadCom,
evidently so are going to saw off their own privates to make a
statement. Well, they tried the same thing with Itanic, and that's
worked just *soo* well.

You seem to be doing a good job trying. (hint: do you know what a
signature separator is?)


I think you need to get out more and see the reality. It doesn't look
like you have any experience of large corporations or the technical
issues discussed.
 
David said:
I think you need to get out more and see the reality. It doesn't
look like you have any experience of large corporations or the
technical issues discussed.

LOL[*]. ...and exactly what have you contributed here?

[*] How much more thirty years experience in system design and
processor development do I need to post to the Usenet?

What a maroon.
 
K Williams said:
How much more thirty years experience in system design and
processor development do I need to post to the Usenet?

Keith, pay attention! David specifically referred to **large**
corporations. Your experience at dinky little IBM doesn't count.
 
David said:
I think you need to get out more and see the reality. It doesn't
look like you have any experience of large corporations or the
technical issues discussed.

LOL[*]. ...and exactly what have you contributed here?

[*] How much more thirty years experience in system design and
processor development do I need to post to the Usenet?

Perhaps he judges by height? Or blood type?
What a maroon.

True.

/daytripper (who just loves good irony ;-)
 
Back
Top