Embedded Processor Watch

Editor: Jim Turley

In This Issue:

• - Welcome to Embedded Processor Watch!
• - Editorial: Can Palm Pilot Overcome Its Success?
• - MIPS IPO Reveals 76% of Revenue From Nintendo
• - Demise of Somerset Splits PowerPC Vendors
• - Matsushita Licenses ARM7TDMI
• - About Embedded Processor Watch

Welcome to Embedded Processor Watch!

Embedded Processor Watch is an independent journal dedicated to reporting and analyzing advances in microprocessors for embedded systems, information appliances, consumer electronics, industrial control, and real-time systems.

Our editorial staff specializes in 32-bit & 64-bit microprocessor hardware. We also cover CPU support logic and digital signal processor (DSP) chips. Embedded Processor Watch does not accept advertising and is not affiliated with any microprocessor vendor.

Privacy notice: We will not make your information available to anyone else. We may send you information on other MicroDesign Resources publications, services, and conferences from time to time. If you do not wish to receive such mailings, send a request to cs@mdr.zd.com.

Anyone may subscribe to this newsletter and receive timely news and information on microprocessors. Only the editors of Embedded Processor Watch may post messages. To write to the editor, Jim Turley, please address your e-mail to: "JTurley@ba.mdr.zd.com"

The Editorial View: Can Palm Pilot Overcome Its Success?

The New Market for PDAs Feels Like Macintosh vs. Windows All Over Again

Palm Pilot has single-handedly reinvigorated the market for PDAs. With more than two million units sold, it's far and away the most popular product in this turbulent category. But Pilot's success attracted Microsoft's attention. Even now, "palm-size PCs" from legions of Microsoft licensees have begun to wage cruel war on Pilot. The battle smacks of one from recent history: Macintosh versus Windows. Will the pint-sized battle play out the same way?

The parallels are eerily familiar. On one side, we have a startup company driven by a dedicated leader with a singular vision. From Palm Computing (now part of 3Com) Jeff Hawkins takes the role of Steve Jobs. His product is similar to, but significantly different from, others in its category. It emphasizes ease of use. Its hardware and software were developed together. It even uses a Motorola 68K chip. The product develops a grass-roots following of happy, satisfied (not to say fanatical) users. Possession becomes a lifestyle statement; its use a passive-aggressive jab at the presumptive market leader.

On the other side, we have the acknowledged irresistible force in the industry. Its product is later to market. The user interface adopts many of the innovator's features, layered on top of a general-purpose kernel. Hardware design is left to low-margin vendors. Sales channels are wide open. Name recognition attracts the business community. Camp followers on both sides gird for battle.

There's talk and speculation about Pilot's future now that Windows CE has entered the PDA market. Some see Microsoft's entry as the death knell for Pilot, Psion, TI's Avigo, and other "nonstandard" platforms. Much of the debate has taken on the missionary accent of zealotry, with partisans on a sacred quest to defend their chosen technology.

Pilot has advantages both for independent software developers and for users. Any Pilot program runs on practically any Pilot. It is a single, software-compatible system. Unsophisticated users can buy or download Pilot applications and use them right away.

Not so for the palm-sized WinCE units. NEC's MobilePro runs different software than Philips's Velo, which runs different software than Casio's E-10, and so on. These units have different hardware designs that are not always hidden under the API, not to mention different microprocessors. With Windows CE, the API might be the same, but the binaries are different for each system. This fragments the market for independent software and confuses consumers.

If 3Com is careful, Pilot can avoid this problem for years. Palm started at the very bottom of the 68K food chain (a 68EC000 core at 16 MHz), so there's plenty of headroom for binary-compatible growth. Motorola could upgrade the part with an '040 core or some flavor of ColdFire (provided certain 68000 instructions were reinstated) and remain totally binary compatible with the first Pilot.

It's clear that Windows CE can expand into very different kinds of applications, and thus become the single operating system around which an OEM standardizes. It may not be as ideal as Pilot for this category, but it clearly has broader uses. And, hey, it's from Microsoft. For companies with their eye on more than just PDAs, Windows CE is the safer choice.

In the Mac/Windows battle, it was the cheap hardware and copious third- party software that carried the day, not the user interface. Users may have preferred the Mac's design but they chose a PC because of its cheaper price, faster rate of product evolution, and greater selection of products.

Palm could repeat this example or learn from it. The company can pursue either margins or market share, but not both. Keeping its own prices down and licensing PalmOS would likely fill the market with cost- competitive spinoffs. The increased market share could overwhelm Windows CE almost before it gets started. Squeezing the maximum profit from each unit, as Apple did, would leave the back door open for Microsoft's minions to overtake Pilot's lead.

As long as PDAs remain simple items--and Pilot has proven the wisdom of this attitude--user interface will prevail over software availability. A PDA is frequently a personal choice, not a company dictate. If the initial "out of the box experience" isn't a happy one, users won't care about the choice of third-party applications. Consumers will judge on usability, where the Pilot excels.

I think Pilot and its successors will be the most popular PDA for some years to come. Palm-sized PCs will retain a smaller portion of the PDA market: more successful than the dreadful HPCs but not as popular as Pilot. Outside the PDA market, Windows CE will be hugely popular among OEM developers and appear in a range of very different products.

Pilot appeals to users, while Windows CE (if not the palm-sized PCs themselves) appeals to developers. In its niche, Pilot will reign supreme. --Jim Turley

Mario Makes Millions for MIPS

Prospectus Reveals Nintendo Brings In 76% of MIPS Technologies Revenue

As Silicon Graphics prepares to spin off its MIPS Technologies design department (see Microprocessor Report 4/20/98, p. 1), the company's prospectus reveals a company that's heavily dependent on one customer. According to documents filed with the U.S. Securities and Exchange Commission (SEC), MIPS garnered 76% of its total 1997 revenue directly or indirectly from Nintendo, up from 67% the previous year. These statements paint a financial picture of a company whose success has been virtually subsidized by one customer, and whose future financial condition could be fundamentally altered if Nintendo fails to select MIPS for its next-generation game consoles.

MIPS earns money from Nintendo in three ways. First, NEC pays a royalty to MIPS for each VR4300 chip it sells to Nintendo. This royalty is based on the VR4300's selling price, rather than a flat per-unit fee. Second, until the end of 1997, MIPS collected a per-unit royalty from Nintendo for its work on the system's graphics coprocessor. This royalty reached a contractual cap in 4Q97, which explains the nearly $2 million dip in income at the end of that year. Finally, MIPS collects a royalty for each Nintendo 64 game cartridge, regardless of its manufacturer.

There appears to be no intellectual property embodied in the game software or cartridges; MIPS may have just negotiated a game royalty as a way of sharing in Nintendo's success. Together, these direct fees accounted for 76% of MIPS's overall income in the last six months of 1997.

As an independent company, MIPS will no longer be designing workstation processors; those employees who are (and their expenses) will stay with SGI. This reduction in headcount and the disproportionate decrease in expenses should help the new company stand on its own.

An independent MIPS will definitely be smaller. Many key employees have already left. Sandcraft, QED, ArtX, and Tensilica, for example, are all staffed or founded by former MIPS engineers. These startups have found-- sooner than Silicon Graphics apparently did--that the future of MIPS lies in embedded systems and not workstations.

All other things being equal, if the Nintendo royalty stream dried up today, MIPS would still collect about $1 million per month from chip sales from its other licensees. If the sales of, say, handheld or palm- sized PCs picks up, that number would increase. That's not a bad revenue stream for an intellectual property firm, but it would force massive layoffs and curtail development work. On its own, MIPS Technologies may be much smaller, but still profitable, with some very lucrative potential upside.

Demise of Somerset Splits Embedded PowerPC Vendors

With the sun setting on Somerset, the once bucolic design center jointly founded (and funded) by IBM and Motorola (see Microprocessor Report 6/22/98, page 4), the course of embedded strategy from these two companies will take a turn. IBM has essentially ceded the midrange of the embedded PowerPC market to Motorola, leaving the latter company to focus on communications controllers while IBM pursues ASIC business.

With Somerset, IBM had a three-pronged strategy that included servers, midrange processors (primarily for Apple), and its 40x series of ASIC cores. Without Somerset, IBM will have no midrange processor cores to replace the existing 740 and 750 (G3) series. Instead, IBM's embedded design center near Raleigh (North Carolina) plans to "cherry pick" CPU designs from the company's three high-end design centers in Rochester (NY), East Fishkill (NJ), and Austin (Texas).

Motorola will continue using Somerset to develop new midrange embedded processors, especially new cores for its line of integrated communications and networking chips. While IBM fine-tunes its cores for ASIC development, Motorola will produce packaged parts. Thus, the two companies will split along custom/commercial lines, as well as over microarchitectural implementations.

For the short term, both companies still have access to the 603e/604e and 740/750 processor cores and will likely use them in ASIC designs (IBM) and custom controllers (Motorola). Beyond about 1999, however, the two companies will be working with separate cores. In general, we expect IBM's processors will have somewhat lower performance than Motorola's, which will follow the previously published PowerPC roadmap more closely (see Microprocessor Report 8/26/96, p. 12). Even with different cores, software compatibility should not suffer after the two companies part company.

Matsushita Licenses ARM7TDMI

Adding another feather to what is becoming a full headdress, ARM Holdings, plc (the new name for the newly public Advanced RISC Machines) has signed Matsushita, a Japanese consumer-electronics conglomerate as its newest licensee. Matsushita becomes the 27th ARM licensee to be identified and the 31st overall by our count.

Matsushita's stated intent in signing the ubiquitous license agreement is to use the ARM7TDMI core in two kinds of products, GSM cellular telephones for the Asian market and "multimedia Internet appliances." Matsushita controls the Panasonic, Technics, and Quasar brand names, all popular and well-known to consumers around the world. Matsushita's Internet appliances could appear as a Panasonic television or set-top box, for example.

An ARM7TDMI is certainly underpowered for any serious multimedia processing, so Matsushita will probably pair the core with another media processor and relegate the ARM to control functions, a la Samsung's ill- fated MSP (see Microprocessor Report 6/2/97, p. 11). Matsushita may develop the media processor on its own, or it may try to license one of the many existing designs.