[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Lucent Technologies & Sun Microsystems



>	And Intel! If the NC model based on a virtual machine gets the
market, Intel will loose the monopoly on CPUs!

There is virtually no likelihood that the NC model will succeed.  It's too 
incompatible, too limited, and the much-ballyhooed cost savings simply aren't 
enough to justify the dead-end, limited-function, crippled/brain-dead NC 
machine.

>	And I don't think they [Intel] can compete with other designs.

That's preposterous, take a look at Intel's earning statements and their 
production and R&D capacity and then try to say again (believably) that any of 
their competitors are going to put Intel out of the business.  I think that 
competition by AMD and Cyrix will siphon off (just) enough money to force Intel 
to not be *too* abusive.

The place where CPU competitors *can* make a significant inroad, however, is in 
dedicated niche markets that Intel hasn't targeted.  For example, the new 
Quantum "bigfoot" hard disk drives have *the* simplest drive control electronics 
board that I've *ever* seen... and I suspect that a good part of the reason for 
that is the "Lucent" chip as one of the (only three!) ICs on the board.  I 
presume that this part is one of Lucent's new DSPs, which I suspect does damn 
near everything to run this new (and VERY cost-effective) drive in software.

Another example of a niche (but still VERY large) CPU market is the ubiquitous 
Rockwell chipsets used in nearly every modem and fax machine out there.

Still another (upcoming) huge-but-niche market is the new Texas Instruments DSP 
(which executes up to 1.6 Billion instructions per second, going to 2 Billion 
shortly) and which costs (in quantity) less than $100.  One-tenth the 
transistors of a Pentium (simple but fast since it does the parallelization and 
optimization at COMPILE time, then uses a VLIW where each word contains as many 
as eight separate instructions which can all be guaranteed executable in 
parallel), runs on only 2.5 volts (two flashlight batteries!) and capable of 
doing a 1024-point Fast Fourier Transform in just 70 microseconds!  Texas 
Instruments is betting a bundle on this chip, and I suspect based on what I've 
seen it's going to be an enormous hit... unbelievably fast, and very, very 
cheap.  (Anyone here know if Inferno has been brought up on the TI DSP yet?)  Of 
course, this chip isn't intended to replace the Intel-family machine in PCs.

As for beating Intel at *INTEL's* game, I wouldn't count on that... we've seen 
huge press blitzes over the years about Sparc, HP/PA, Alpha, MIPS, PowerPC, and 
God knows how many other CPUs that were going to "revolutionize" the 
architecture of the personal computer, and none of them have gone anywhere of 
any consequence (other than to near-oblivion, in most cases).

Even if some tiny company were to manage to create a CPU just as good (and 
marketable!) as Intel's... they probably (a) couldn't produce it in 
Intel-equivalent volumes, and (b) would be so outspent in R&D by the enormous 
financial capacities of Intel that any later-generation product would have Intel 
regain (or at least come damn close) the upper hand.

Brings to mind a story about a fellow who was digging around in the company 
archives at the Coca-Cola headquarters in Atlanta, and managed to find an old 
scrap of paper which laid out the "famous secret formula" for the company's 
flagship (which he then published in his book).  Of course the company wasn't 
happy about that, but one of their executives pointed out to the author:

  "Look... today's your lucky day.  I have in my hand the NEW, accurate, 
complete formula for [modern] Coca-Cola.  I'll give it to you!  ...  Now, what 
are you going to do with it?  Are you going to build a factory to manufacture 
it?  Where do you think you're going to find all the growers and other suppliers 
to provide you with all the required ingredients?  Are you going to set up a 
network of thousands of your own bottlers and distributors around the world?  
And you can't call your product 'Coca-Cola', of course..."

The bottom line is that even making an equivalent product, EVEN IF you have cost 
advantages, there's a long way to go between achieving that and really seriously 
putting Intel into a precarious situation.

Gordon Peterson
http://www.computek.net/public/gep2/