2 Possible Reasons Intel might be withholding Q9450

[wwswimming]

now that the numbers for Phenom power consumption are out,
AMD's case for buying a Phenom are pretty low. if there's
an 40 watt difference between it and the Q6600 - that's
about 1 kilowatt hour a day. 20 cents a kilowatt hour is
pretty normal in the US, our power company has a 3-tier
power consumption where they really kill you if you use
more than a certain amount, 10, 20, or 30 cents a kilowatt
hour, depending.

anyway, that's $6 a month, $72 a year. $144 for 2 years.

basically, if Intel were to come out with the Q9450 for
the low $300's, in first quarter 2008, well, AMD's "great
green and white hope" won't help them become profitable
again. it looks like AMD has another round of AMD R&D to do.

now, didn't intel have some kind of anti-trust litigation ?

i wonder if by postponing the Q9450 for 3 months,
Intel creates a fair argument for themselves in case
AMD or anybody does litigate against Intel on anti-
trust grounds. "hey, we withheld the Q9450 for 3
months so we could milk the Q6600 and not put too
much pressure on AMD."

by withholding the release, to some degree, Intel
neutralizes possible unfair trade practices/monopoly
type of accusations.

having watched various Apple, Microsoft, Intel,
AMD, AT&T lawsuits in the media over the years, it's
hard to keep them straight.

still, i wonder if there's some legal thinking that
went into the decision to withhold the Q6600.
in addition to the financial thinking, "hey, we
can just keep it in cruise mode and we'll make
loads of money".

from the shareholder point of view, and considering
that there's companies like Milberg Weiss Lerach
that would be happy to sue on behalf of the share-
holders for an alleged possible Intel management
mistake, withholding the Q9450 is good legal-wise
AND profit-wise for Intel.

but it's bad for us :-(

 

[o1die]

I think Intel just wants to postpone more R&D for awhile; why shouldn't they? Selling old technology as long as prices hold is more profitable.

 

[Idontcare]

Perosnally I think people are way overthinking this whole thing. Waaayyyy too many conspiracy theorists out there.

Intel needs to deploy their limited supply of 45nm chips to wherever there is the greatest demand for these 45nm, less in supply than their 65nm older sibling, chips.

Intel is not losing sales (or profits) by not immediately releasing 45nm quads to the dekstop market. Intel is surely capacity constrained at 45nm...it behooves Intel (because it is in their shareholders best interest to do so) to allocate more of their 45nm chips to the market segments that will pay the most for them.

That would be Apple and the Server markets. AMD can still make a case in some of the server markets that their IMC and hypertransport on X2's and Phenoms are still more performance/watt than Clovertown (65nm quad Xeons) for those workloads that drive >2 sockets per system. These markets are hungry for high-FSB low-power Harpertowns (45nm Xeon quads).

If Intel has limited 45nm capacity...and are wanting to increase their market share in the server space...why would they lessen the supply of 45nm quad Xeons just to replace an existing and still performance competitive quad core Kentsfield for desktop?

Supply versus demand...it really need not be any more complicated than this, and Occam's razor should apply here.

 

[21stHermit]

Originally posted by: Idontcare
Perosnally I think people are way overthinking this whole thing. Waaayyyy too many conspiracy theorists out there.

Have to agree. But conspiracy sells.

Intel needs to deploy their limited supply of 45nm chips to wherever there is the greatest demand for these 45nm, less in supply than their 65nm older sibling, chips.

Intel is not losing sales (or profits) by not immediately releasing 45nm quads to the dekstop market. Intel is surely capacity constrained at 45nm...it behooves Intel (because it is in their shareholders best interest to do so) to allocate more of their 45nm chips to the market segments that will pay the most for them.

That would be Apple and the Server markets. AMD can still make a case in some of the server markets that their IMC and hypertransport on X2's and Phenoms are still more performance/watt than Clovertown (65nm quad Xeons) for those workloads that drive >2 sockets per system. These markets are hungry for high-FSB low-power Harpertowns (45nm Xeon quads).

If Intel has limited 45nm capacity...and are wanting to increase their market share in the server space...why would they lessen the supply of 45nm quad Xeons just to replace an existing and still performance competitive quad core Kentsfield for desktop?

Supply versus demand...it really need not be any more complicated than this, and Occam's razor should apply here.

One could make the same general case for dual core, currently AMD can't compete with current 65nm C2D's yet the replacement 45nm dual cores appear to be on track for a Jan 20th. release.

Therefore, I suspect their is a modest engineering reason for the modest 45nm Quad delay.

 

[Idontcare]

Originally posted by: 21stHermit
One could make the same general case for dual core, currently AMD can't compete with current 65nm C2D's yet the replacement 45nm dual cores appear to be on track for a Jan 20th. release.

Therefore, I suspect their is a modest engineering reason for the modest 45nm Quad delay.

The most compelling technical/engineering argument I have heard thus far is the purported "4-layer PCB" stability issue when combined with high-FSB quads.

I find this particular angle to be compelling because it does not require server quads to have the same high-FSB problem as most server boards are 6 or 8 layer PCB's AND even Kentsfield quads (the B3's at least) had singificantly little FSB overhead when they were released to the desktop.

My B3 QX6700 does not like a FSB >333MHz, and I tested it in both Striker and the Evga boards.

So I could see where in Intel's situation, if they were going to run into a design-margin issue then it would most likely be on the desktop (cheaper constructed mobo's versus server markets) and it would be with the high-FSB quads (history of 65nm quads supports this margin concern).

However until a reputable source actually confirms anything I will pretty much pass on even reading too many thread topics regarding "Intel is teh bad guys in withholding 45nm chips"...

 

[wwswimming]

The most compelling technical/engineering argument I have heard thus far is the purported "4-layer PCB" stability issue when combined with high-FSB quads.

is the 4 layer PCB you're referring to the one that
the CPU chip (monolithic quad core) or chips (2 dual
core) are mounted on ?

 

[taltamir]

its called making money. They will release it when the competition justifies it. Either the competition from AMD, or the competition from their last gen products.

 

[Cattlegod]

not a chance. that is not how economics works.

 

[Xvys]

This is all about increasing the value of the Q6600's already in the pipeline. Intel could of released the Penryn Q's in January and be forced to slash the price of the Q6600 to make them competitive. Due to the lack of competition, Intel can now sell millions of the remaining Q6600 for perhaps $200 instead of $150.

 

[BonzaiDuck]

I'd like to see people more reserved in "playing the conspiracy card."

Intel is a "dominant" firm just short of being part of a duopoly. Therefore, competition is limited and comes from "concentrated" -- even single -- player(s).

So they can plan their release of newer product within more wiggle-room than "many-buyers-many-sellers" would allow.

they can also price-discriminate, and they have done it before.

Even though we might just look at things from standpoint of dominant-firm/oligopoly (or duopoly), throwing considerations about older motherboards and pissy-fit-feuds regarding nVidia are also possible.

IdontCare raises the possibility of technical issues per bus-noise above 1,333 FSB. Maybe -- maybe not.

It all depends on what other things "were done" with the QX9650, or planned for the 9450 and other models.

If they can afford to wait for the expected release and revenue flow from the Penryns, they have time to clean up technical issues as well.

 

[arcas]

A lot of people seem to have forgotten what it was like 16-17 years ago when Intel was the main player in the x86 game. Intel was content to release one or two incremental speed improvements per year and stuck with architectures as long as possible (the 386 and 486 were 'king' for 4-5 years each. compare this to the Pentium-Pro which lasted 2 years, the Pentium-II/III lasted 3 years combined). The difference: competition. It wasn't until AMD and later Cyrix and IBM started producing x86 clones that were roughly comparable to Intel's stuff that the race began and prices fell.

(I'll just focus on the AMD-Intel leapfrogging but Cyrix and IBM played a role in the 486/pentium days)

AMD releases their 40mhz 386 which is roughly as fast as the slowest 486s at the time. Not long afterwards, Intel soon introduces the DX2 486 line...

AMD releases the K6 in spring 1997 which was generally a match for the Pentium MMX. Intel answers a few months later with the Pentium-II.

AMD releases the K6-2 in late 1998 which was a match for the Pentium-II. In Feb 1999, Intel introduced the Pentium-III.

AMD introduces the Athlon in 1999 which generally beats the P-III and wins the race to 1GHz. Intel answers with Pentium-4 in late 2000. AMD counters with Athlon-XP/MP in 2001.

AMD introduces Athlon64 in 2003/2004. Intel counters with Prescott, realizes that's a dead-end and uses ideas from the Pentium-M to come up with the Core/Core2.

etc, etc.

 

[Idontcare]

Originally posted by: arcas
A lot of people seem to have forgotten what it was like 16-17 years ago when Intel was the main player in the x86 game. Intel was content to release one or two incremental speed improvements per year and stuck with architectures as long as possible (the 386 and 486 were 'king' for 4-5 years each. compare this to the Pentium-Pro which lasted 2 years, the Pentium-II/III lasted 3 years combined). The difference: competition. It wasn't until AMD and later Cyrix and IBM started producing x86 clones that were roughly comparable to Intel's stuff that the race began and prices fell.

(I'll just focus on the AMD-Intel leapfrogging but Cyrix and IBM played a role in the 486/pentium days)

AMD releases their 40mhz 386 which is roughly as fast as the slowest 486s at the time. Not long afterwards, Intel soon introduces the DX2 486 line...

AMD releases the K6 in spring 1997 which was generally a match for the Pentium MMX. Intel answers a few months later with the Pentium-II.

AMD releases the K6-2 in late 1998 which was a match for the Pentium-II. In Feb 1999, Intel introduced the Pentium-III.

AMD introduces the Athlon in 1999 which generally beats the P-III and wins the race to 1GHz. Intel answers with Pentium-4 in late 2000. AMD counters with Athlon-XP/MP in 2001.

AMD introduces Athlon64 in 2003/2004. Intel counters with Prescott, realizes that's a dead-end and uses ideas from the Pentium-M to come up with the Core/Core2.

etc, etc.

Not disagreeing with the many valid points you make here...but I would add (as someone who does remember those days) that while it was true the total number of speedbins available to the market at any given time were fairly small (3-4 speedbins) the performance range they spanned was pretty breathtaking.

486-33MHz...and the next chip was 66MHz (2X the clockspeed!) and then the next chip was the 100MHz (another 1.5X clockspeed!)

We don't have speedbins anymore where the performance gap between two successive chips are so substantial. Could you imagine Intel (or AMD) releasing a 2GHz and then NOT releasing anything else until they could release a 4GHz chip?

That would be crazy exciting, and that is how it was back then. The mid-step chips, the 40MHz ones and the 75MHz and 80MHz x486 chips did not come out until later (they backfilled the series).

 

[lopri]

Wait till Nehalem and its accompanying chipsets debut. Those will give us a clearer picture how the CPU market (meaning prices to us) pans out. It's a very likely possibility that we won't get such bargains that we get now, but Intel can't just do what they used to do 15 years ago. Computers are now a commodity - back then it was a luxury. If I venture to guess, the market will be kinda divided to a more distinguishable segments. Say today one can pick up a $200 CPU and get the performance of a $1K CPU - this may no longer be the case. If I venture to guess, there will be very performant sub-$200 CPUs, but there will also be $1K CPUs that a $200 CPU can't match no matter what.