On Thu, 25 Nov 2004 07:05:00 -0500, Tony Hill
This is logical, but purchasing decisions sometimes are based on
emotion. Just go to your local Ford dealership and look at the GT
proudly displayed front and center. How many of those does the dealer
actually sell? My bet is - low single digits, most likely 0. How
much it helps to move the lowly Focuses, Escapes, and Rangers? So
much that it makes sense to Ford to sell the GTs at a loss
(considering development costs, Ford will probably never see any
profits off it). Who in his right mind (I mean regular mortal
taxpayers, not some heirs or jackpot winners) would splurge $139,995
on a Ford? Same goes to P4EE+SLI. While only few will be actually
sold, they'll give Intel bragging rights and sell many more lower-end
systems.
That only works to a certain extent. The mainstream market (ie no one
in this newsgroup! :> ) doesn't get all that emotional about PCs like
they might about something like a car. In fact, most people won't
even see the numbers, let alone care much about them.
Being technically capable is one thing. Having the license is the
other. I am almost sure NVDA patented the technology and trademarked
SLI, or could do so to prevent others from using it. Besides, it may
be easier, faster, and thus economically sensible, to just copy
existing technology rather than trying to develop it from scratch.
ATi and VIA are also planning on doing SLI chipsets, so there's
nothing too specific to nVidia here. The smarts are ALL in the video
card, from the motherboard side of things it's pretty much just two
PCI 16x connectors. If Intel really wanted to do this, I'm quite
certain that they could, regardless of anything nVidia might say.
Worst-case might be that the term "SLI" might be trade-marked, in
which case they would have to come up with some other name.
I suppose nVidia could go to great lengths to try and prevent their
video cards from working in SLI setups on non-nVidia platforms, but
that would be utterly moronic. Considering nVidia makes probably a
hundred dollars or more on EACH of the video chipsets in an SLI setup
(assuming a top-end card, the only place it really makes sense), as
compared to pennies on a motherboard chipset, they would have to be
completely stupid to try such a thing.
NVDA has cornered the high-end AMD market and has good positions in
midrange AMD. But P4 chipsets is quite a different game.
In the high end, NVDA will have to compete against INTC on both
performance and customer loyalty. Intel admittedly knows a thing or
two about making high-performing chipsets, especially using their
proprietary FSB, DDR2, and PCIe. Matching it, while not impossible,
would be a tough job. As for customer loyalty and brand power, Intel
is in the league of its own.
Traditionally there's been good reason to stick with Intel chipsets
when buying Intel processors. Even today I wouldn't bother with a
non-Intel chipset if I were to buy an Intel processor, with the
possible exception of ATI (particularly in laptops).
In lower end, it will be a fierce competition on price against VIA,
SIS, ATI, and once again Intel.
As mentioned previously, VIA is kind of hurting on the Intel side of
things while SiS is stuck in the very bottom-end of things. If you
look at HP, for example, they don't sell a single non-Intel chipset
for any of their desktop systems using Intel processors. ATI,
meanwhile, seems to be really focusing on mobile chipsets (and doing
reasonably well at such).
The real competition on the low-end, as you correctly point out, is
Intel itself. An i845GV or i865GV chipset sells for DIRT-CHEAP,
there's no way nVidia is going to make any money by undercutting them
on price there.
Judging by what we see in AMD chipset
market, price is not the selling point of Nforce chipsets. Besides,
low end usually demands for integrated graphics, and Nvidia just
decided to drop the graphics from AMD line of chipsets. Will it
re-introduce the graphics into P4 line, considering that P4
northbridge must have its own memory controller unlike A64 one, and P4
FSB is more complex than hypertransport bus?
The P4 bus does lend itself rather better to integrated graphics when
compared to the Athlon64. It would seem to make sense for nVidia to
have some sort of integrated memory controller on their P4 chipsets.
Could it be that newer
Nvidia graphics core takes too much of the valuable die real estate,
or produces too much heat to be easily integrated into the
northbridge, or both, and the older one (GF2MX) just does not make the
cut by today's standards?
The integrated graphics of the nForce2 (basically a GeForce4 MX) is
probably still reasonably competitive when compared to Intel's i915G
integrated graphics, let alone the comparatively weak i845G or i865G.
However you may be on to something on the die-size thing. According
to Anand Tech's review of the ATI integrated chipset, the integrated
video portion takes up almost 90% of the chipset.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2269&p=16
Now, obviously integrating video allows you to sell chipsets for a
higher price, perhaps making it worthwhile, but a die that is 8 to 9
times bigger is going to cost you a good bit extra.
Still, the more I look at this, the more I think that nVidia will
mainly stick to a couple niches. In particular I suspect they'll
target laptops and workstations first and foremost, not worrying so
much about desktops, except perhaps on the top-end of the enthusiast
market, which isn't that far removed from workstations (at least from
the chipset perspective).
