Marvell's plan: Enabling 'next billion' connected device users

Leveraging its homegrown ARM-based Sheeva CPU processors, Marvell Technology Group's co-founder and CEO has a clear vision: Enabling the next billion users of connected devices.
In today's electronics industry, Marvell is not alone in its quest. Nor is its idea necessarily new.

Both Nvidia and Freescale Semiconductor, for example, are betting big on netbooks that run Linux on an ARM processor. They hope to stir up competition for--if not an end--the current Windows-Intel dominance of the notebook PC market.

In the lead up to Taiwan's Computex show in June, Nvidia, with its Tegra SoC (based on ARM11), and Freescale, with its i.MX51 CPU (ARM Cortex-A8), are racing to pick up design wins from bigger computer OEMs and ODMs.

Meanwhile, Texas Instruments, which is pushing its ARM-based OMAP processors, is counting on cellphone and consumer electronics OEMs familiar with ARM to become a disruptive force in the notebook PC market.

Marvell, on the other hand, scoffs at netbooks. The company has set its sights on the non-PC, non-netbook markets.

Marvell CEO Sehat Sutardja said during a recent interview with EE Times: "A lot of people tell us, 'Oh, you are doing ARM because you want to replace PCs.' But they are wrong."

High-tech companies tend to chase existing markets like PCs, said Sutardja. "But why should we when the next billion [users] is much bigger than the PC market?"

The Marvell CEO stressed: "PCs will be always PCs," quickly adding: "You don't want to throw away the PC's legacy overnight." But next-generation Internet devices "shouldn't pretend to be a PC."

Jim McGregor, chief technology strategist at In-Stat, agreed. Companies developing "those devices that look, smell and act like a PC" will be "fighting an uphill battle against Intel and Microsoft," he said.

Marvell's mission is to insinuate its Sheeva CPU into a number of connected consumer devices. The company's strategy includes Sheevas designed into digital photo frames, used to upgrade feature phones into smart phones and powering set tops, Blu-ray recorders and other connected devices that "allow consumers to absorb content," according to Sutardja.

In a race to the next billion devices

"The next billion people will use new information displays and 'connected' consumer devices in places they've never imagined before," Sutardja predicted. "In kitchens, bathrooms and laundry rooms, for example, where you used to think it's silly to use your PC."

Many share Sutardja's optimism. In-Stat's McGregor said, "Marvell's strategy represents the lowest path of resistance and highest probability of success for their technology and market position."

But industry observers aren't necessarily sold on Sutardja's prediction that "the same engine" will be used in a race to the next billion devices.

Sutardja claimed, "Those days are over when OEMs chose different processor cores for different consumer products." ARM-based Sheeva processors, loaded with interfaces, audio, graphics, Ethernet, 3G and Wi-Fi, for example, will become an alternative to a variety of SoCs that dominate and fragment today's consumer devices, he said. "Not using the same engine isn't making [an end system] any cheaper."

A healthy dose of skepticism is warranted, however.

First, the consumer market is notoriously broad and fragmented. Second, Marvell's consumer device gambit is still largely unproven.

While acknowledging that Marvell's new Sheeva products "can compete in set-top box, Blu-ray and picture frames," Linley Gwennap, president of market researcher Linley Group, said: "These markets are currently dominated by MIPS and PowerPC processors."

McGregor called Sutardja's "same-engine" theory "not practical" and "not likely anytime soon." That's because different devices have different performance, power, form factor, thermal and other design constraints. "It would be nice to have one processor that could do it all, but it is just not practical, and will likely not be practical for some time," he said.

McGregor added: "Using different processors can aid the OEMs in adding value by adding differentiation to their products. Just look at how many ARM cores and processors are available. STMicroelectronics and TI have even announced multicore processors based on the Cortex-A9 architecture."

Marvell's competitors are beginning to stir. "In the Shiva-class, TI is the only one shipping in volume, with their OMAP3, in the ARCHOS 5 PMP," said Will Strauss, president of Scottsdale, Ariz.-based market researcher Forward Concept. "The OMAP3 has also been announced in the upcomming Palm Pre. Qualcomm's Snapdragon has been announced in a Toshiba cellphone that hasn't yet shipped."

Added Strauss, "No others are yet on the street."

Why Marvell has a chance

Marvell has more than a few advantages, starting with its ARM architecture license. It's one of few companies in the world that can legally design and sell ARM-compliant CPUs.

Most in the industry have yet to recognize Marvell as a full-fledged, high-performance processor company. Marvell is widely known as a leading supplier of disk-drive controllers, Ethernet chips and mobile Wi-Fi chips in which ARM-based CPUs are used to control the flow of data.

In reality, its new Sheeva family of ARM CPUs are not the first Marvell has implemented on its own, higher performance architecture. Many of Marvell's homegrown ARM cores are embedded in custom processors. Architectural details have never been disclosed.

Marvell, for example, has optimized its own dual-core ARM over the last four years, said Sutardja. Further, Sheeva is not Marvell's first gigahertz-level ARM processor. Sutardja said Marvell first used its gigahertz ARM processor several years ago in custom-designed printer engines.

"We've developed our own compilers, debuggers and verification tools. They have been optimized. Our chips have been tested by big customers," he claimed.

"Marvell has the oldest and the most mature, high performance ARM architecture, developed on its own, already shipped in hundreds of millions of units," Sutardja claimed.

Linley Group's Gwennap said Marvell's strength is in its economies of scale. "Marvell was able to spread its CPU design costs over the 300 million units it shipped last year, therefore keeping its costs low," he said.

Android factor

Marvell contends that the Android platform will play a big role in spreading Sheeva CPUs in divergent consumer electronics devices.

In Sutardja's mind, Internet-connected consumer devices will share many of the feature sets available in smartphones. "Connected digital photo frames, for example, are in many ways like smartphones without the function of making voice calls," he said.

If software developed for a smartphone platform like Android can be easily ported to another consumer device, it will assure that smartphones continue to grow in volume and morph into many other devices, the Marvell CEO said.

Marvell's strategy does not diverge much from those of the leading mobile handset vendors. Kai Oistamo, Nokia's executive vice president, said in an interview last fall that it seeks to be not just another cellphone company but a much more broadly defined mobile computing company.

Does this mean that mobile handset companies are aggressively pursuing the development of mobile computing and consumer devices? Or are they just confused?

Clearly, not every mobile phone company is on the same page. Sutardja said, "Some of them are working on them [new non-mobile handset devices], while others have not made up their mind."

Marvell's strategy to leverage the smartphone platform to expand in the CE market makes sense, yet the company will still have to battle TI for ARM-based application processors. "Due to its strength in smartphones, TI has more than 50 percent share, compared to about 20 percent share for Marvell," Gwennap said.

While the movement toward Android is growing, especially among Japanese consumer electronics vendors, the market has not yet signaled whether Android can become a cross-platform environment for connected consumer devices beyond smartphones.

With or without Android, some see that software support for Web content may remain a nagging problem for any devices based on Sheeva or other ARM processors.

Gwennap said that for devices that need to display Web content (including many of the newest set-top boxes and Blu-ray players), Intel processors support a broader range of Internet plug-ins and other software. "In contrast, devices based on Sheeva or other ARM processors will not be able to display all Web pages properly."

However, he added, "The amount of Web software available on ARM continues to increase, and consumers may not be willing to pay a big premium for an Intel-based system if the ARM-based system can handle the most-popular Web sites."

Will Marvell's consumer gamble pay off?

While acknowledging Marvell's ability to compete in the CE market, In-Stat's McGregor called the company "a late comer" and "still a smaller player," compared to companies like Samsung, Broadcom, TI, Freescale and Qualcomm. "In this space, business relationships are critical, and sometimes more important than a superior product. Thus, Marvell must continue to opportunistically capitalize on any opportunity that develops," he concluded.

Gwennap concluded that Marvell has developed an excellent CPU that stands out from other RISC and x86 (Intel) designs in performance, power and cost.

However, he noted, "To win consumer designs, Marvell must take the next step and deliver complete application-specific solutions based on the Sheeva CPU. Only then can we see if this strategy will succeed."

BY Junko Yoshida

Copyright © 2009 TechInsights, a Division of United Business Media LLC All rights reserved.



Copyright 2008-2009 Daily IT News | Contact Us