Marvell Pours a WiFi Triple Shot
Firing a pre-CES warning shot, Marvell Technology Group Ltd. (Nasdaq: MRVL) announced today a supposed triple-strength 802.11n chipset, a step that was inevitable but is drawing some skepticism from others.
Marvell is all excited because the TopDog 11n-450, as it's called, boasts three transmitters and three receivers and a 3x3 antenna array, as opposed to the 2x2 arrangment found on many competing chips.
If those three radios deliver three separate WiFi streams, that's a 50 percent advantage, giving Marvell a claimed 450-Mbit/s throughput versus the 300 Mbit/s on certain other chips. The bigger throughput requires having Marvell's chip on each end of a connection.
Thus continues a war of escalation created by the way 802.11n works. The standard uses multiple-input/multiple-output (MIMO) antenna schemes to get its higher throughput and better range. Chip makers are working their way up to 3x3 formations, and the standard allows for a 4x4 as well, so Marvell's announcement is an understandable step.
Of course, the 450-Mbit/s throughput is just theoretical. Plenty of factors would shave the real figure downward: the overhead associated with protocols like TCP, the quirks of the physical surroundings, or the difficulty of processing high-definition video.
It's the first product to claim the 450-Mbit/s mark, but it's not the first 3x3 chipset. Atheros Communications Inc. (Nasdaq: ATHR) made that claim two years ago. The difference is that Atheros chooses to work with only two data streams.
"We wanted that extra antenna for redundancy," says Bill McFarland, Atheros's chief technology officer. "We did a lot of study on this, and the statistics are really not in your favor doing three streams in a 3x3 situation."
That's because wireless transmissions can be a crap shoot. In fact, McFarland likens it to rolling dice. Given three dice, it's easier to roll two sixes than to try to roll three -- you've got a backup in case one fails -- as any good Risk geek can testify.
So, Atheros isn't interested in matching Marvell's 450-Mbit/s ante. Neither is Broadcom Corp. (Nasdaq: BRCM), according to Mike Hurlston, vice president and general manager of that company's wireless LAN unit.
"Our view is that the most important thing to stimulate 802.11n demand is really cost. Where we've focused our energy is on bringing the cost down on a basic 802.11n offering," Hurlston says.
Rather than add transmission streams, Broadcom -- and Atheros -- have been more interested in boiling 802.11n down to a single chip. (See Broadcom Intros One-Chip 802.11n.)
Higher throughput could be of use in sending multimedia around the home -- one of the presumed drivers of future 802.11n demand. But Broadcom would rather revive the idea of using the 5 GHz spectrum of 802.11a.
"We're going to talk about that at CES. A number of our customers have introduced dual-band" products, Hurlston says.
In any case, more radios don't necessarily equate to higher bandwidth in real-world conditions.
"Just going to more streams and more radios doesn't necessarily help you get better worst-case performance, which is really what matters," says Steve Martin, vice president of engineering for systems vendor Ruckus Wireless Inc.
There's some self-interest there, as Ruckus claims its directional antenna technology can make wireless LAN more reliable. (Ruckus uses Atheros chips, incidentally.) Still, multiple streams create complexity, as WiFi transmissions "get finicky the more streams you have," Martin says.
Martin doesn't see much play for a 3x3 chip in enterprise or carrier settings yet -- but retail could be another matter, and might be where Marvell's customers will point the product first. "Consumers will buy that stuff because there's a bigger number on the box," he contends.
Marvell didn't return a call requesting comment.
— Craig Matsumoto, West Coast Editor, Light Reading