Rule Maker heavyweights Microsoft (Nasdaq: MSFT) and Nokia (NYSE: NOK) reported earnings last night and are on the move -- Nokia is up 28% and Microsoft is up 16% in afternoon trading. We'll take a look at Microsoft's earnings tomorrow and Nokia on Tuesday. Today, we'll be spending time on AMD and Intel.
In the past year, Advanced Micro Devices (NYSE: AMD) has been wildly successful in the consumer market. Nine of the top ten vendors use variants of AMD's Athlon chips in their desktop systems, and many estimates now say that AMD sells the majority of all processors in the consumer market. Yet this translates to a little less than 30% of the processor market as a whole, because AMD has yet to crack the business market.
Last I checked, the highest profit margin on any processor Intel (Nasdaq: INTC) sold came from its Xeon processors. The reason Xeons command such high prices isn't that they run faster than the lower-margin chips: the highest-end versions run at 550 MHz, almost half the speed of a standard Gigahertz Pentium III. The reason they fetch such high prices has to do with Symmetrical Multiprocessing (SMP), or the ability to use more than one processor in a machine.
High-end business servers need more processing power than even a 1000 MHz (gigahertz) chip can provide, and to get it they use machines with many processors in them. By choosing Xeons, they can use up to eight 500 MHz processors simultaneously, and 8 times 500 is quite a bit more than 1 times 1000, as is 4 times 700 MHz, another popular Xeon configuration.
The bottleneck when doing SMP is the ability to read data from memory (and write it back). Having eight processors twiddling their thumbs waiting for data to arrive is a waste of money. A Xeon is basically a Pentium III with much more on-chip cache (up to two megabytes, as opposed to the 1/4 megabyte in a gigahertz Pentium III). This helps Xeon stay at full speed in a multiprocessor environment rather than a situation of several processors trying to suck data from main memory through the same straw. For more information about on-chip cache, check out this recent Rule Maker story.
The ability to do SMP lies primarily in the motherboard. To make a motherboard, you need a set of supporting chips to do I/O and memory management and such. This is why AMD was so happy to finally demonstrate a prototype of its new 760MP chipset at Microprocessor Forum 2000 on October 10. Athlons have been out for a year now, but without a multiprocessor-capable chipset, nobody could make motherboards capable of accepting more than one Athlon at a time. Now they've demonstrated a chipset that can handle two, and hope to ship the new parts by the end of the year, which means actual dual Athlon systems will probably be out in the first quarter of 2001. Tom's Hardware has more information on this topic.
This puts price pressure on the low end of Intel's high end starting in Q1 2000, but doesn't cut into the market for four-way and eight-way Xeon systems yet. AMD will probably wait until its new 64-bit Sledgehammer processors start shipping before focusing on the higher-end configurations. Since it has to do a new chipset for new motherboards to support 64-bit computing, which is aimed at high-end servers anyway, it might as well build SMP support into that chipset from the beginning. This seems to be AMD's current plan, although plans have been known to change.
My research uncovered a ton of technical information about how Intel's and AMD's SMP approaches compare, (things like AMD licensing Compaq's EV6 instead of the Intel Multi Processor Specification, the importance of "Lightning Data Transport" to reduce pin count, and integrating L3 cache on-die into new motherboard chipsets). In brief: the new stuff is seriously cool, and what AMD is about to do leaves what Intel currently has in the dust. Then again, what Intel is about to do also leaves what Intel currently has in the dust.
Ever since the Athlon's introduction a year ago, AMD has been selling processors faster than it can build them. Going after the SMP market at this time won't help it sell more chips, only increasing its production capacity will do that. Yet supporting SMP puts price pressure on Intel's highest margin products, and inflicting pain on a competitor (distracting it and reducing the revenue stream with which it competes against you) is a valid business strategy. AMD is forcing Intel to compete on more fronts, drawing Intel's focus away from the desktop systems where AMD has been successfully converting customers in droves.
Just as importantly, SMP gives Athlon more legitimacy in the eyes of the business community. Without the ability to do SMP, many larger corporate vendors consider AMD systems to be toys, and won't even consider them for use in single processor desktop machines or workstations. AMD got this reputation years ago with its low-yielding K5 and K6 processor designs, and is still working to shake it. Intel's "Intel Inside" logo hasn't inspired much brand loyalty among price-conscious consumers, but it has given the company recognition and credibility that is vital when selling to businesses.
Then again, companies like Sun Microsystems (Nasdaq: SUNW) and IBM (NYSE: IBM) also share the high-end server space, using non-Intel processors. Intel doesn't have a lock on the high end, but it does have established relationships making it familiar to existing customers with the naturally conservative bent common to Fortune 500 companies. Most corporate information officers prefer to make purchasing decisions by pressing the "more of the same" button, so already being deployed is a big advantage. Until AMD has SMP systems that can handle the heavy-duty servers, it's a much harder sell even at the workstation level.
The computer manufacturer space is interesting as well. Ironically, the ongoing processor shortage plaguing the industry this year has actually made some Intel-only vendors cling even tighter to Intel. They play nice with the processor giant in order to work their way up the supply food chain, registering to become an Intel Product Dealer or even applying to become an Intel Premier Provider in hopes of getting first pick from an insufficient supply of chips.
This applies even more to the scarce supply of preproduction prototypes. It's hard to design an iTanium-based product without any iTanium chips on hand. Some smaller vendors who believe that Intel will remain king of the hill are afraid to even look at AMD, for fear of disturbing their relationship with Intel. Vendors who believe Intel will remain the core of their business often won't even evaluate non-Intel solutions, creating a self-fulfilling prophecy.
But mostly, this all boils down to market inertia. AMD currently has the better price-to-performance ratio with its Athlon product line, compared to the many incarnations of Pentium III. Intel is trying to pull back ahead before its customers fall out of the habit of buying Intel-branded products. AMD is trying to stay ahead long enough for that not to happen and is advancing and filling in gaps in its product line to make Intel's catch-up job harder.
The battle continues. Stay tuned for dispatches from the front.
Finally, today, if you've decided it's not too soon, or too late, to think about retirement, you're right. If you're looking for some guidance, consider signing up for The Motley Fool's Roadmap to Retirement Online Seminar. During this 12-lesson, four-week seminar, investors will figure out how much they need for retirement and how to build an investment strategy to get there. Enrollment ends November 2.
