Broadcom, the leading supplier of 802.11g chips for consumer Wi-Fi devices, is claiming equipment made by trailing rival Atheros Communications causes an enormous degradation in the speed of nearby 802.11b and 802.11g networks.
The culprit: the 108 megabit-per-second Turbo mode in Atheros' Super G feature set, according to Broadcom. Super G is built into Wi-Fi adapters and gateways from D-Link and NetGear. Only devices from each of the companies can work at the higher speed with similar devices from the same companies; they are currently not interoperable. Also, the standards group Wi-Fi Alliance has not certified Atheros' Super G.
"We've done the testing with both the NetGear product and the D-Link product, and proved that it is bad neighbor technology," says Jeff Abramowitz, Broadcom vice president of marketing. Instead of the standard 54-Mbps throughput, nearby 802.11g networks even on different frequencies may deliver as little as 1 Mbps, he says.
Atheros dismisses Broadcom's claims. In its tests, the Turbo mode causes no more performance problems with other devices on the same or non-overlapping channels than a comparable 802.11g device, says Craig Barratt, Atheros president and chief executive officer.
Because Broadcom's claim should be testable and reproducible, Broadcom representatives say the company will release its methodology and results for verification. They plan to demonstrate the claimed performance problem at the Comdex trade show in Las Vegas next week.
As a market leader without a proprietary double-speed mode, Broadcom's public airing of their concern and their tests could raise eyebrows. Atheros has gained on Broadcom in market share, especially in dual-band hardware that supports both 802.11a and 802.11g in the same adapter or access point.
Conversely, Broadcom supplies 802.11g chips to the leading consumer manufacturers, including Cisco's Linksys and Apple Computer, and its customers have the most to lose if Atheros's technology does interfere with plain 802.11g. Broadcom's customers might also face eroding customer market share faced with a 108-Mbps alternative to standard 802.11g at 54 Mbps.
D-Link spokespeople say internal tests of the company's XtremeG products, which use the Atheros chip, show only the same amount of degradation on neighboring networks that any nearby Wi-Fi network would cause. NetGear declined to comment for this story.
The Wi-Fi Alliance has certified D-Link's XtremeG PC card as an 802.11g product, and the access point is still working through the certification process.
NetGear does not list Wi-Fi certification on its site for this equipment, but displays its own seal for the device.
The Wi-Fi Alliance, while unwilling to directly address Broadcom's claims, plays a role in ensuring that certified products interoperate and don't cause harmful effects on other Wi-Fi certified products.
"If a product did not interoperate because in proprietary mode it broke the interoperability, that product shouldn't pass the certification program," says Frank Hanzlik, managing director for the Wi-Fi Alliance. "We could do an investigation and call for a repeal."
Repeated attempts to contact the IEEE 802.11g task group and the 802.11 public affairs group requesting comment were not returned.
However, Gartner analysts had cautioned about potential incompatibilities between emerging wireless technologies, shortly after 802.11g's debut. The research firm said products that are not certified by standards organizations may prove to have interoperability problems with other 802.11g devices, as well as with older 802.11b wireless LAN technology.
Atheros's Super G technology comprises three parts, only one of which is at issue. One element is frame bursting, in which the shorter packets designed for the slower 802.11b network are rebuilt into longer packets that can dramatically increase network efficiency, according to Barratt. All major Wi-Fi chipmakers are offering frame bursting, which is an element in the upcoming 802.11e standard for improving multimedia streaming and voice over IP service.
Super G also includes an element of compression in which two Atheros adapters use lossless methods to compress redundant data, Barratt says. Because frame bursting rewrites packets into a mode that any 802.11b or 802.11g device can receive, and compression is used only when available in both a gateway and an adapter, neither of these elements have any impact on devices using chips from other manufacturers.
Turbo, the third component of Super G, requires some engineering cleverness, however, that directly affects other devices, although the scale of this effect is the matter in dispute.
In the 2.4-gigahertz frequency band allotted for unlicensed use in the United States, the 802.11b and 802.11g standards have 11 channels, which are staggered and overlap. The Turbo mode of Super G uses channels 5 and 6 simultaneously to achieve its double speed, according to Barratt. That should leave channel 1 almost entirely and channel 11 clearly out of contention for the same frequencies.
However, Abramowitz offers spectrographic analysis and bandwidth throughput charts that demonstrate that Turbo activated in its continuous or static mode caused significant frequency overlap and bandwidth degradation in channels 1 and 11, as well as the channels 5 and 6.
Atheros also offers a dynamic version of Turbo G in which the device cycles through its faster mode and 802.11g mode to find compatible 802.11b and 802.11g devices. The dynamic mode is even worse, as it degrades performance in nearby networks for a few minutes, then lifts that degradation, and then returns, Abramowitz says.
The NetGear gateway currently supports only static mode, although the company promises on its Web site to offer an upgrade to dynamic in early 2004. D-Link offers both static and dynamic Turbo.