Marvell debuts passive optical network chips

For a long time, I was wondering the decision of PMC buying Passive, a PON chip maker. Even though the Passive division did not bring in as much revenue as we expected, but it seems it’s the right acquisition after all. If Marvell and Broadcom are trying to get into the same market, it must be a good market to be in. With the Passive acquisition, PMC is ahead in the game and becomes the market leader of the moment.

by Rick Merritt, EE Times, 02/02/2010
Broadcom, Cavium may be next to jump into PONs

SAN JOSE, Calif. — Marvell is making its first foray into home broadband terminals, launching a new family of chips for passive optical networks. The news comes as the number two player in the PON market, startup Teknovus Inc., is expected to be acquired by Broadcom Corp. or Cavium Networks.

“We didn’t play in DSL and cable set-top boxes but we are jumping in now with the inflection point of optical broadband coming to the home,” said Nafea Bshara, chief technology officer of Marvell’s enterprise business unit.

The timing is good. Market watcher Infonetics Research estimates the worldwide market for PON systems will hit $4.5 billion in 2013 driven by the rise IPTV systems such as those of AT&T and Verizon and rapid rollouts of PON-based services elsewhere, especially in China.

Marvell will roll out four devices this year under the Avanta brand–including Gbit, Ethernet and 10 Gbit/second Ethernet PON chips–leveraging a wide range of its in-house silicon blocks. The $3 billion Marvell hopes to outgun PMC-Sierra which shipped about half the $140 million in PON chips sold in 2008, according to market watcher The Linley Group (Mountain View, Calif.).

“Carriers such as China Telecom are understandably nervous” depending on PON chips from a handful of relatively small chip makers with revenues typically well below $200 million, said Bshara. “They need a credible company to come and support this market,” he said.

Marvell quietly acquired Iamba Networks (Cupertino, Calif.) in a deal that closed in March. It has married Iamba’s Gigabit PON technology with Marvell’s own internally developed Ethernet PON designs to create dual-mode products.

“This is unheard of in the industry today and will drive costs down as it drives economies of scale up,” Bshara said.

The Marvell SoCs will pack plenty of punch. They include the company’s ARM V5-compliant Armada 300 processor at data rates ranging from 1.2 to 2 GHz.

Some of the chips will also integrate a six-port Gbit Ethernet switch that supports Audio-Video Bridging, IP multicasting and a draft Energy-Efficient Ethernet standard. In addition, some will build in a version of Marvell’s Prestera programmable packet processor to handle IPv6, power-over-Ethernet and voice over IP.

The networking blocks leave the Armada processor available to handle third-party apps and carrier services such as data backup and parental controls. Marvell has shipped about 10,000 developer kits for the Armada CPU.

The initial chips will be made in 55nm process technology, consume 3.2W max and fit into a 19x19mm BGA. Some of the devices shipping later this year will be made in a 40nm process.

The chips “sound pretty impressive with greater integration than anything that else on the market,” said Jag Bolaria, analyst with The Linley Group. “PMC-Sierra will now have some real competition from a credible long term supplier,” he added.

The Gbit EPON standard is strongest in Asia with deployments in China, Korea and Japan, said Bolaria. The 2.5 Gbit/s GPON is being used in the U.S. by Verizon.

Late this year, Marvell will roll out a 10G Ethernet PON chip. It is designed for a smaller market of systems that deliver links that fan out to a neighborhood or a large apartment building.

Marvell’s Bshara said he believes he has an edge over competitors such as Broadcom or Cavium believed to be eyeing an acquisition of startup Teknovus which sold an estimated $25 million in PON chips in 2008.

“I am not surprised [Teknovus is] trying to shop themselves around,” said Bshara. “But whoever picks them up will be two years behind us to do the integration,” he added.

“For Teknovus, an exit at this point is good as the market heats up and its technology has the most value,” said Bolaria. “Teknovus does not have the resources to develop the complete client platform including a CPU, Ethernet chip, Wi-Fi, and etc.,” he added

Leave a Reply