On the eve of this week's biannual Intel Developer Forum, Intel announced the second delay of its latest wireless LAN silicon, highlighting the company's struggles to firmly establish a foothold in the wireless arena.
The conference, in San Jose, is expected to draw more than 4,000 hardware designers and software developers who specialize in Intel-based products. Twenty-five sessions, spread over at least four IDF tracks, will cover everything from global spectrum policy reform to power management technologies for mobile WLAN devices.
A troop of Intel executives is expected to unveil details about an array of wireless and mobile technologies, such as:
* The next generation of Pentium M mobile processors.
* A new mobile chipset.
* Research on creating seamless roaming for mobile users over different Internet connection zones.
* A broadband demonstration using 802.11 WLAN and 802.16 fixed wireless technologies.
* A road map for Intel's future 802.11a, 802.11a/b and 802.11g WLAN chips.
But Intel will have to update that road map to get around the pothole that just opened up.
Intel earlier this year said it would ship in the third quarter a combination 802.11a/b WLAN chip, so that a Centrino notebook user could connect to either a 5GHz, 54M bit/sec 802.11a network or a 2.4GHz, 11M bit/sec 802.11b network. But last week, the company confirmed the chip will ship later in 2003, the second delay this year. Intel's first 802.11a chip is due out soon. An Intel spokesman says the 2.4GHz 54M bit/sec 802.11g chip will be ready by year-end. Intel has said it will ship a chip that can handle 802.11a, 802.11b and 802.11g by June 2004.
Rivals lead the pack
But all of those products are months, even years, behind Intel's WLAN rivals. Two of them, Atheros Communications and Broadcom , timed announcements for this week of their latest WLAN products, alongside of which Intel's look pale. Both vendors' chips are used widely wherever WLAN vendors want more than 802.11b, which is all Intel currently supports.
Atheros' new chipsets, one supporting 802.11a/b/g, the other for 802.11b/g, feature advances that cut power consumption by 60 %, and roughly double the ranges of the previous products. In one 802.11g test, range increased to 2,400 feet compared with about 800. Both chips also have a special, proprietary feature to reach a throughput of up to 90M bit/sec, compared with roughly 22M to 27M bit/sec in a standard 802.11g network. But this requires Atheros chips on each end of the connection.
Broadcom will unveil two WLAN chipsets in its AirForce line, also supporting 802.11g and 802.11a/b/g. Broadcom says the chips use 80 % less power than the WLAN chips that are part of Intel's Centrino notebook package. (For columnist Scott Bradner's take on Broadcom's news, see page 24.)
The WLAN chips used in Centrino are not Intel products. Centrino, announced with great fanfare in February, is a collection of products. Those products include Intel's well-regarded Pentium M processor, an Intel memory controller chipset and a WLAN interface cobbled together with three chips from four companies: Koninklijke Philips Electronics NV for the radio, Texas Instruments for the baseband and a joint Intel-Symbol Technologies project for the MAC chip.
In June, Intel acknowledged that some Centrino notebooks with Intel's ProSet software were causing VPNs to crash. Corporate users were advised to remove a specific piece of the ProSet software to ensure smooth operation.
"Most Intel Centrino Mobil Technology-based systems shipping today have (this piece) removed," an Intel spokesman says. "VPN software verified with Centrino will not experience this issue."
"Intel really isn't in the WLAN chip business . . . yet," says Jeff Thermond, vice president of the home and wireless networking business unit at Broadcom. "They're a buyer of piece parts from outside Intel, acting like a contract manufacturer. That will certainly change. The mystery to us is why it's taken them this long."
"We've taken a focused approach," says Jim Johnson, vice president of Intel's wireless network group. "We're delivering not only performance, but also standard networking capabilities for management and security. It seems that cadence (of product introductions) is working for us. Would I like it to be faster? Absolutely."
Almost every major notebook builder has adopted Centrino, which would seem to be a major victory for Intel. But a closer look shows a more complex arrangement.
Dell introduced Centrino notebooks early this year on its Latitude line for enterprise users and Inspiron for consumers. "Centrino is our default," says Steve Macon, wireless product marketing manager for Dell. "It's included standard in the system price for all these (products)."
Yet about half of all Latitude Model D notebooks bought by enterprise customers ship with a wireless adapter based on Broadcom chips. These customers choose an alternative 802.11b/g card (also included in the standard system price), or they pay an extra $69 for an 802.11a/b/g card.
"Investment protection is a key factor for some enterprise accounts," Macon says. "They can deploy one (802.11a/b/g) card with existing 11b nets, and then migrate these to 11g or 11a." The "combo card" ensures the notebook, unlike a Centrino client, can work with any of these.
In digital cellular, Intel's market share is miniscule. "They've made a lot of noise, and a couple of product announcements," says Mike Yonker, CTO for wireless terminals at Texas Instruments, which is the market leader. "It really boils down to: How many phones can you go out and buy today based on Intel technology?"
Intel attacks the cellular market with its XScale technology, based on a unique variant of the StrongARM microprocessor. Earlier this month, Intel announced a single, integrated chip that combines the XScale processor with a signal processor, Flash memory and SRAM for phones on GSM/EDGE networks. Earlier this year, a version of this chip was released for GSM/General Packet Radio Service networks.
"All that we do is based on the convergence of voice and data," says Hans Geyer, vice president of Intel's PCA Components group. "Applications drive data use over these nets: It's pictures, video and games that fill up the (broadband) pipeline. Cell phones are becoming small computers."
Geyer is dismissive of suggestions that Intel is at a disadvantage in cellular markets. "In this change from voice to voice-plus-data, all of us are starting from scratch," he says. "This market (voice-plus-data) didn't exist three years ago, and the rules and requirements are completely different (from voice-only cellular)."
Intel's resources for such a massive effort are formidable. For the past three years, the Intel Communications Fund has worked through $500 million set aside for investing in early stage companies. Last year, it set aside $150 million of that just for WLAN investments. So far, it has given $35 million of that to 15 companies working on network-to-network wireless roaming, back-end provider billing, low-power radios and cellular gaming applications, says John Hull, the fund's director.
In each of the past two years, despite the economic implosion of the high-tech industry, Intel has invested more than $4 billion in research and development. CTO Pat Gelsinger oversees 75 laboratories and 7,000 researchers. Despite stunning losses in the two key groups where much of the new wireless products are located (Intel doesn't break out revenues and losses in more detail), the company is forging ahead.