It's in the Chips
By Kevin Werbach, Mon Jul 08 00:00:00 GMT 2002

Computing and communications functions move increasingly closer, and Intel plans on putting both in the same silicon.


The convergence of computing and mobile communications is usually understood at the level of applications. Laptops are sprouting wireless cards, as mobile phones deliver data services such as Web access, networked games and personal information management. It's almost as hard to find a PC today without a modem port as it is to find a mobile handset without a CPU, memory and a display screen. Yet down at the hardware level, communications and computing functions are still rarely integrated.

Intel has other ideas. The chip giant foresees a day in the not-too-distant future when every chip can have an integrated radio, without significant increases in cost.

Think about that for a second. Intel shipped hundreds of millions of microprocessors last year. Those volumes give it massive economies of scale on everything it produces. And it's deadly serious about doing for RF what it did for the PC. At Intel's spring analyst meeting, President and COO Paul Otellini boasted that the company had 1500 engineers working on wireless technologies.

Intel's interest in wireless shouldn't come as a surprise, says Kevin Kahn, Intel Fellow and director of the company's Communications Interconnect Technology group: "The long-term themes within the electronics industry have been the rise of computing and then the rise of connected computing. It's almost inconceivable today to talk about any serious amount of computation and not talk a about connectivity."

Intel executives are quick to distinguish the vision statement of a radio on every chip from specific production plans. What gets integrated into what products at what time is a business question. PDAs and mobile handsets call for smaller form factors and lower power consumption than laptops, for example. The key point is that it will be as easy to add communications functions to a device as today it is to add more computing power.

Intel is particularly excited about WiFi and other unlicensed wireless technologies, because that's where it sees the strongest resonance with its familiar PC industry. Says Kahn: "If you look historically at our industry, one way of looking at it is as a sequence of battles beyond chaotic and orderly things. In almost every instance that I can think of, the chaotic thing won." The PC beat the mainframe and Ethernet beat centralized LAN protocols. Now, WiFi is challenging top-down wireless data technologies such as 3G. If history is any guide, the messy bottom-up approach will win. In addition to building WiFi chips and devices, Intel is lobbying the US government to provide more flexibility and spectrum for unlicensed wireless technologies.

A tale of two processes


The involvement of Intel, and of other companies coming from the PC semiconductor world, heralds significant changes in communications. Radios and chipsets for wireless devices often involve specialized analog components, special-purpose signal-processing chips and exotic materials. "Traditionally, radios have been one of the last bastions of the art, if you will, as opposed to the engineering of chip design," observes Kahn. On the PC side, vendors have perfected commodity CMOS processes for driving costs down and performance up, resulting in the familiar curve of Moore's Law. Today's 2 Gigahertz Pentium 4 will go to 3 Gigahertz tomorrow, because it follows well-established manufacturing and miniaturization techniques.

To get a sense of how the relentless cycles of the computing world are coming to wireless, consider the development path of wireless LAN chipsets. It was only three years ago that the first 802.11b devices hit the market, and last fall that vendors such as Atheros went into production with higher-speed 802.11a chips. From initial price points in the hundreds of dollars, 802.11b cards for laptops can now be had for $50, and 802.11a cards for only slightly more. Those price decreases have been achieved through a combination of higher volume production and denser integration of functions onto fewer chips.

And that's just the start. There is already a startup - Bermai - demonstrating a single-chip CMOS 802.11a implementation that can be integrated with existing 8021.11b designs. A chip that supports 802.11a, 802.11b and wide-area protocols such as GSM and GPRS can't be far away. On day one, that chip will be too expensive for most devices. But the wireless superchip will get cheaper over time. At some point, the premium for including it in any device that needs one of the supported protocols will evaporate. Moreover, as the components become cheaper and smaller, they need not be separate chips at all. They can be components of CPUs and other powerful chips.

Flexibility wins


Integrating radios into chips is more than just an engineering accomplishment. It has profound consequences for the devices and services that make use of those chips. The most obvious advantage is price. When the addition of wireless communications to a device adds negligible cost to the device, there's no reason not to do so. Eventually, predicts Kahn, "communications is going to become essentially free."

Another advantage of building RF capabilities into CPUs is that wireless devices will have newfound smarts, because they will be able to take advantage of the computational power of the microprocessor. They will be able to sense and adapt to whatever wireless networks are within range. Such flexibility initially adds costs. If the goal is to build a radio that handles one frequency band and protocol, the best solution may be a hard-wired, special-purpose chip. Move to two radios, then three, and the advantage begins to dissipate. At some point, the flexible solution always wins. And not just because there isn't space in a handset for four radio chipsets. Volume is everything in chip production, because fabs and equipment represent the bulk of the costs. A chip that goes into 100 million devices may be cheaper than one built for only 10 million, even though it's more complex inside.

Europeans are accustomed to a single mobile standard, GSM. As new wireless data technologies become more important, having devices able to handle multiple bands and protocols will be essential. Already, 802.11a devices must be flexible enough to support the different regulatory contours of the 5 Ghz spectrum band in different parts of the world.

Going forward, no one system is likely to provide the best, most affordable connectivity everywhere. 3G will offer wide-area coverage, but it will be expensive and its data bandwidth will be limited. GSM and GPRS will be important transitional technologies, especially in the US, where 3G spectrum hasn't even been assigned. 802.11 offers cheap, high-bandwidth data connections, but limited range and interference potential. And before long, there will be at least four standard 802.11 protocols with different characteristics that a device manufacturer may want to support. Finally, ultra-wideband technology holds great potential, though it has only recently been approved for limited commercial use in the US due to interference concerns.

There will be a need for devices that can use any of these protocols as needed, giving the user the best or cheapest connection at a given moment. The only way to do so is to build radios agile enough to adapt, which means they must have access to significant computing power.

Going horizontal


The convergence of computing and communications also poses a business model challenge to wireless operators and vendors. In hardware businesses, governed by Moore's Law, today's top-of-the-line device is tomorrow's entry-level and the next day's junk. Users generally pay the costs of their devices up front, and upgrade every few years when the new generation becomes too compelling to pass up.

In contrast, communications businesses follow what one sharp-tongued observer called Moron's Law. Once services are rolled out, they evolve slowly if at all. Wireline voice phone service is basically the same product today as it was fifty years ago, with minor improvements such as touchtone and call waiting. Wireless services are somewhat better, driven by competition and technological improvement. Still, 2G digital services have been on the market for the better part of a decade, with packet data connections just now starting deployment. Handsets are replaced with some frequency, but the underlying networks follow an infrastructure model. Carriers invest large sums building out their networks, and then recoup their expenses over time through usage fees.

Because it has depended so much on specialized hardware and single-purpose networks, the communications industry has traditionally been vertically integrated. On the other hand, the PC business is horizontal: there are separate markets for chips, systems and software. Whether a PC uses an Intel CPU or one from AMD, it can run the same applications. This industry model has promoted innovation and competition at every level.

The hardware elements of communications have already started to go horizontal. Handset vendors are outsourcing production to the same contract manufacturers that are prominent in the PC industry. Switching equipment increasingly uses general-purpose semiconductors. Once connectivity can be integrated into devices at marginal cost, though, the possibility of an entirely different communications industry arises.

Imagine that every laptop, every PDA, every home media server is also an agile communications device, able to connect to any available network. In such a world, paying a carrier for access to a single network, with a limited choice of services and hardware, will seem archaic. There will still be services businesses linking together these devices and, more important, the user data that flows across them. But they won't look much like the integrated communications carriers of today.

When, not if


The integration of wireless and computing is a matter of when, not if. "From where I'm sitting it looks fairly inevitable," says Kahn. "Don't get me wrong; there are some hard problems to solve. But none of them look insurmountable. It's just a matter of engineering to solve them."

Kevin Werbach is an independent technology analyst and consultant.