Sehat Sutardja: An Engineering Marvell
Open up today's mobile electronic devices and you'll likely see the imprint of Sehat Sutardja, the beating heart of Marvell
Look inside an e-reader, game console, Blu-ray player, TV, or smartphone and odds are you'll find a cluster of chips designed by Marvell Technology Group Ltd., a 15-year-old Silicon Valley firm. Marvell crafts the CPU as well as the wireless transmitters and receivers, the digital signal processors, the video processors, and even the power management system. Its chips will likely appear in the wave of Android tablets expected to hit the market soon. And the One Laptop per Child organization just announced that its next generation of low-cost computers—the first to finally get below US $100—will rely on Marvell electronics as well.
But Marvell and its quietly intense founder and CEO, Sehat Sutardja, aren't exactly household names. That doesn't particularly bother Sutardja. If he's looking for affirmation of his enormous success, he need only consider the company's explosive growth, its lineup of innovative, in-demand products, its more than 5000 employees, its R&D and design centers in Canada, China, France, Germany, India, Israel, Italy, Japan, the Netherlands, Silicon Valley, South Korea, Singapore, Spain, Switzerland, and Taiwan, and its $3 billion in annual revenues. If Sutardja doesn't capture media attention like, say, Apple's Steve Jobs, well, he's okay with that.
But just as Jobs sets the vision for Apple, Sutardja pours his personality and passions into Marvell. And unlike Jobs, who was not the technological powerhouse of the two Steves that founded Apple, Sehat is an engineer to the core, a genius at circuit design for whom electronics is not only a vocation but also his only recreation.
So far, Sutardja, an IEEE Fellow, and his Marvell cofounders—his wife, Weili Dai, and his brother, Pantas Sutardja—have successfully managed the business as well as the technology. But in a sense it's been an off-Broadway production—it may be packing the house night after night, but it hasn't gotten a lot of media attention. Today, however, the company is trying to make the transition to a bigger stage, and Sutardja increasingly finds himself in the spotlight.
Born in Indonesia to Chinese parents in 1961, Sutardja had a fairly ordinary childhood for his time and place. "If it rains," he recalls, "you go out and dance in the rain. You see butterflies, you chase the butterflies. You see dragonflies, you chase the dragonflies. It was a simple life."
And then he discovered engineering, and his butterfly-chasing days were over.
It happened on a visit to Singapore, where his younger brother Pantas was living with their grandparents. Sehat was in sixth grade, Pantas in fifth. As boys will do, Sehat started poking through his brother's possessions, including some DIY books and magazines geared toward electronics hobbyists. He spotted a great project: building a Van de Graaff generator. He and Pantas bought some copper wire at a surplus store and managed to rig together a crude but functional machine.
Back in Indonesia after vacation, Sehat couldn't stop thinking about the generator. He began constructing a miniature version that he could use to shock people as a prank, scavenging gear from his parents' auto parts store. The device worked, but its copper contact points oxidized so fast he had to sand them down every few minutes.
After a little research at the local bookstore, he discovered that transistors could replace mechanical switches. "Okay," he remembers thinking, "I'll have to learn how to use transistors." Sehat found a radio repair shop where he could do just that. Within a year he received his radio repair license, a document he seems more proud of than his college diploma. (His wife carries it for him in her purse in case he wants to show it to people. Which he often does.)
Sehat continued to tinker with circuits, and he also kept abreast of new developments in the field. The most interesting news always seemed to be coming from the likes of Fairchild, National, Motorola, and Texas Instruments. His next step was obvious: He had to go to the United States.
One of the few people he knew in the States was the brother of a friend, who happened to be studying at the University of San Francisco, a private school. So Sehat applied there. He arrived in the summer of 1980, only to discover that the math and physics classes were repeats of what he'd already taken in high school. The college didn't even have an electrical engineering program, he was crushed to learn.
FAMILY MATTERS: Marvell's founders are actually one happy family. Sehat Sutardja [center] is the CEO; his wife, Weili Dai [left], and his brother, Pantas Sutardja, are also Marvell executives. The company reflects their passions; for example, the basketball court, pictured above, has special meaning for Dai, who played on an elite team as a child in Shanghai.
Another Indonesian friend was enrolled at Iowa State University, which did have an engineering school. Sehat transferred and then spent the next two and a half years racing through the curriculum, typically taking two extra classes each semester. During the summers of 1982 and 1983 he worked for IBM in Kingston, N.Y.
By the time he graduated in 1983, he knew a little more about U.S. universities and applied to the University of California, Berkeley, for a graduate program renowned in the analog circuit design community. His brother Pantas was also there, working on his undergraduate EE degree.
Sehat studied low-power, low-impedance amplifiers as part of a research program targeting the Integrated Services Digital Network (ISDN), which at the time was considered the next big thing in communications. He moved on to designing analog-to-digital converters for ISDN phones, again focusing on low-power devices because these phones would be powered through the phone network, not the electrical grid.
Even as a grad student Sehat made an impression, says Stephen Lewis, his office mate at Berkeley and now a professor at the University of California, Davis. "He's a perfectionist. For example, when we were students, we were building pipeline A-to-D converters. The traditional way of setting the gain of a switched-capacitor amplifier to two used two capacitors, one twice as big as the other. He figured out a way to do it with two identical capacitors, increasing the amplifier speed by increasing its feedback factor. We had a solution that worked, but he kept digging until he found a way to do it better."
Lewis recalls that Sehat revised his layouts over and over, trying to eliminate any errors that would prevent the capacitors from matching exactly. These errors wouldn't have caused the device to fail; they would have only reduced the performance slightly. Still, Sehat wouldn't let them go. It's that same doggedness and attention to detail, Lewis believes, that makes Sehat and his company so successful today.
Outside the lab and the classroom, Sehat didn't do much; if it wasn't related to engineering, he wasn't interested. He managed to meet the woman who became his wife anyway.
One day, Sehat and a friend visiting from out of town were getting on the elevator in the electrical engineering and computer science building. Already in the elevator was an undergrad named Weili Dai. Sehat's friend began chatting her up. Sehat said nothing, embarrassed by his friend's corny lines.
Dai wasn't terribly impressed either. "There were two guys," she says, "and one of the guys says, 'Hi, are you a computer science major?' I was like, 'Wow, this guy knows this building is computer science. Wow.' "
But she did remember the quiet guy, and he remembered her, enough to say hello when they ran into each other several weeks later. After a few more encounters, he invited her to study with him in his office. Neither liked bars, so for them this was a perfect date. By the time Sehat got his Ph.D. in 1988, the two were so serious that Sehat was unwilling to consider jobs outside the Bay Area.
He joined Micro Linear Corp., a fabless semiconductor company in San Jose. He worked on D-to-A converters and other chips for disk drives. But instead of using building blocks from the company's design library, as was expected, he designed every circuit from scratch, reducing the latency of the disk controller from 100 nanoseconds to just 10. His superiors doubted the designs would work, though, so Sehat found himself lobbying for his approach all the way up to the vice president of engineering, who eventually agreed to use the new circuits. They worked, and the controller ended up in Hewlett-Packard drives used by a wide variety of manufacturers.
Looking for more challenges, Sehat joined Integrated Information Technology in Santa Clara, Calif., where he designed circuits for digital video compression and decompression. The technology was novel but ended up in a dog of a product: the AT&T VideoPhone. It was time to move on, but Sehat didn't know what to do next. His wife did, though.
"Engineers always talk about technology," she says. "They figure out a way to design something, to make a product better." But Sehat was a little different, she says, for he also thought about how the product would fare in the market. When they were still in school, she had told him, "After your Ph.D., you get some real-world experience. But whenever you're ready, we're going to do a company of our own."
In 1995, he was ready. He asked his brother Pantas, who had recently left IBM's Almaden Research Center, to join the start-up. With some savings and money from Dai's parents, on 10 February 1995 the three incorporated Marvell—so named because they were planning to do marvelous things, and, they had noticed, a lot of successful companies' names ended with "el" or "ell," like Intel, Novell, and Nortel.
Sehat already had plans for the first product: a better read channel for disk drives. It sounds incredibly specialized and it is, but it's also one of the drive's key components. The read channel takes the analog signal coming from the magnetic head as it scans the disk, converts the noisy signal to digital, and puts that information out onto the bus that will take it to the computer. Existing read channels used a bipolar transistor on a complementary-metal-oxide semiconductor substrate (BiCMOS), but Sehat planned to use only CMOS. That way the channels could be manufactured by a chip foundry like the Taiwan Semiconductor Manufacturing Co., so Marvell wouldn't have to build its own fab. Using CMOS also meant that the device would consume less power. This would, however, present an engineering challenge: Existing CMOS read-channel designs were much slower than BiCMOS.
However, Pantas had worked on hard drive technology at IBM, and Sehat had experience in mixed signal chips—those that require combining analog and digital technology. Between them they thought they could design a read channel that could handle 180 megabits per second, putting them well ahead of existing 100 Mb/s BiCMOS products and, they projected, equal to the next generation of BiCMOS chips that would be coming from the established companies.
The traditional method used in read channels involved looking for peaks in the analog waves to define the digital bits. But as the bit density increased, it got harder to distinguish the peaks. The Sutardja brothers instead took multiple samples of the signal and then used digital signal processing circuitry to sort them out. This approach reduced noise and also sped up the bit rate.
"The Marvell founders would come to conferences, say they could do it in CMOS, but wouldn't reveal their secret sauce," says Stephan Ohr, an analyst with Gartner who covered analog devices at the time. "Years later we found out that they simply threw more gates at the problem, creating elaborate filtering circuits." The real breakthrough of this approach, says Ohr, was that it was scalable—they could make faster and faster devices simply by following the CMOS technology curve—and the approach could be transferred to other products.
Pantas defined the architecture, then moved on to simulate the digital circuitry; Sehat focused on the analog aspects of the design. The brothers hired four engineers to help out. By Christmas 1995 they had the first working chips.
"I'm a bit narrow-minded. I only see things
in terms of electronics"
Marvell now had a product but no customers. So Dai dialed information and got the number of every disk drive manufacturer she could think of. "Fortunately," she says, "in the storage market you only have about half a dozen guys."
The company sent samples to three drive companies. They all found the chip to be faster and cheaper and lower power—but, Sehat reports, they were reluctant to bet on a start-up.
Eventually, they convinced Seagate Technology to take a chance. Ken Burns, an executive at Seagate, told them that the company's next-generation drive would need a read channel at 240 Mb/s—could Marvell deliver?
Pantas recalls that, even though they had been designing to 180, the chip was already running at close to 240. They told Burns yes. In less than three months the Marvell team hit the 240-Mb/s mark, and Seagate became Marvell's first customer. "That changed everything," Pantas says.
Marvell now had an easier time getting its next customers, particularly in the Japanese market, says Sehat. "The Japanese customers cared more about the technology instead of the perceived risk of dealing with a small company." Today, in terms of units sold, Marvell has about 60 percent of the market for hard drive systems-on-a-chip.
"This little start-up, with one product line, put Texas Instruments out of the read-channel business," Ohr says.
Marvell then turned its sights on the Ethernet industry because, as with the disk drive read channel, Ethernet chips extract analog signals from a noisy environment. The company got a jump start in 2000 by acquiring Galileo Technology, a company that built controllers for data networking. Today Marvell is the second largest supplier of Ethernet transceivers and switches.
Next came Wi-Fi. Most Wi-Fi products at that time were cards with an array of components, but Marvell again blew the competition away when it announced single-chip Wi-Fi in 2005. The Wi-Fi market today brings in about 20 percent of Marvell's revenue.
These days, Marvell, along with many of its competitors, is building not just simple processors but what the industry refers to as platforms. That is, it puts together in one package most of the key electronic components for a device. Flint Pulskamp, an analyst with market research firm IDC, says designing platforms rather than single chips makes it easy for the customer and also presents a higher barrier to competition. Marvell now has platforms for smartphones, including Android devices, Wi-Fi networking, and Ethernet communications. Its platforms are also in two of the three top game systems (Sony's and Microsoft's) and a number of e-readers, including the original Kindle. Earlier this year, right after the company introduced a tablet reference design it calls Moby for use in $99 tablet computers, the One Laptop per Child association announced that it would use the Moby design in its XO-3 tablet. Next on Marvell's agenda: LED lighting, which requires a surprising amount of control electronics.
Just as Sehat had done as a teenage tinkerer, a grad student, and a young circuit designer, he's continued to push Marvell's engineers to rethink conventional designs to get a leap in performance. Sehat's Berkeley office mate Lewis believes that's how the company gets the edge on its competition, by making countless little things slightly better. "You put enough of those things together, and you have a product that is significantly better," Lewis says.
If Marvell's ascent sounds like the plot of a Disney musical, keep in mind that every good story has its moment of crisis. For the company, that point came in 2008, when the U.S. Securities and Exchange Commission accused the company of backdating stock option grants. The same charge was made against a number of tech companies at that time.
Jeff Palmer, vice president of investor relations for the company, explains it this way: "In the late '90s and 2000, when there was a land rush to hire people, potential employees asked for options at price points in the past. There is nothing illegal about granting stock options below market rate. The issue is if you don't document them in your [SEC] filings. This happened at many, many, many companies. We fought the charge, because we believed that we had done nothing fundamentally wrong, and none of the founders enriched themselves personally."
However, as chief operating officer, Dai was in the hot seat, dealing with issues she wasn't exactly comfortable with. "I'm a software engineer," she says. "I never even took a finance class. Finance, legal, all of those completely make a dark side."
Marvell eventually reached a settlement with the SEC. Without admitting or denying the allegations, it paid $10 million in fines, and Dai herself paid $500 000. In addition, Dai agreed to step down as chief operating officer; she now holds the hazy-sounding position of cofounder, vice president and general manager of communications and consumer business. She says she is not allowed to talk about the case but points out that the change in title hasn't changed what she does at the company, which she calls being the "caretaker" of the big family inside Marvell and its customers outside. "An unfortunate tsunami hit," she says, "but everybody knows who I am and what I do, and so life goes on."
Today Marvell has offices in 15 countries, and its cofounders are billionaires. Yet at heart, Sehat is still an engineer, and the corporate culture reflects that.
"We are an engineering-driven company," Sehat says.
"A nerdy company," Dai says.
You see that in the hiring. A vast majority have electrical engineering degrees—and not just those doing design work. Investor relations vice president Palmer, for instance, holds a BSEE from George Washington University. "It doesn't mean that someone who doesn't have an engineering degree could not be hired as a marketing person," says Sehat. "It'd just make it harder."
Sehat describes his management philosophy as a combination of two famous Silicon Valley styles: the "only the paranoid survive" mantra of Intel's former CEO Andrew Grove, and Hewlett-Packard's "management by walking around," which encourages executives to interact more with their employees.
Both styles come naturally to Sehat. "We've always been paranoid, from day one when we started the company," he says. "We always thought the other guys could [beat] us. So we needed to build something even better."
And his designers confirm that even now Sehat won't hesitate to roll up his sleeves and dive into the most difficult project. He keeps up on the technical literature and every week conducts random reviews of projects in the design stage. If his engineers tell him that a task will take two weeks, "Sehat will tell them it can be done in x number of hours or days," says Gani Jusuf, an early employee and now a vice president. "Newcomers think he is trying to put pressure on them, but he's not. It's that he is four steps ahead of everybody in quickness, and he can do it that fast."
And, says Jusuf, he "walks the walk. He may tell an engineer in a design review to change a layout. The designer will say, 'I can't do that in time.' And Sehat will go back to his office for a while, then come out with the new layout and say, 'Here you go.' "
Lately, though, Sehat finds himself being forced out of his EE comfort zone and moving onto a more visible stage.
"I never wanted to be involved in politics, because I thought it was not productive for me or the company," he says.
But the economic and environmental impacts of the semiconductor technology he has worked so hard to advance have started to worry him. "We can now build practically anything in a chip or two, at low cost, for less than the cost of maybe two cups of coffee," he says. "And every one of these devices will consume energy. In the next 10 years we'll be building 50 billion, 100 billion units of these new devices, all consuming energy. And we'll be in trouble very, very quickly."
As part of an industry that is creating this energy-sucking monster, he feels obligated to try and tame it. "We don't have 10 years to solve this problem," he says. "We don't even have 5 years."
Sehat is now a regular visitor to Washington, D.C., where he's been pushing energy conservation standards for small consumer electronics devices. Marvell this year helped write the Smart Electronics Act, HR5070, introduced by U.S. Representative Michael Honda, a Democrat who represents the Silicon Valley area. The bill calls for the U.S. Department of Energy and the Environmental Protection Agency to draft testing protocols for such devices. Marvell is also working with the California governor's office on a similar initiative for the state.
As you might expect, Sehat has already taken an exhaustive look at existing power supply technology, and he doesn't like what he sees. The design hasn't changed in decades, he notes, even though it's now possible to build in circuitry that would keep the voltage in phase with the current and thereby improve efficiency. "Technology has advanced so drastically over the last 20 years that we can implement this function practically for free," he says. Of course, he isn't blind to the business opportunities; Marvell is busy developing circuitry for such improved power supplies.
Sehat is also using his personal fortune to support technology education: "With the support of higher education diminishing, I'm afraid [the United States] will fall behind countries like China." He and Dai have donated more than $20 million to Berkeley to complete the new headquarters for the Center for Information Technology Research in the Interest of Society—otherwise known as Sutardja Dai Hall. Their two sons now pass it daily as EE students there.
But his riches haven't changed his lifestyle. He and his wife carpool to the office most days (admittedly in a Mercedes sedan); they don't own a yacht or vacation home. They eat most of their meals at work, where the cafeteria usually offers a Chinese meal along with more typical Silicon Valley fare. They work until 7 or so and then drive home together, typically talking shop.
In the evening, Sehat often relaxes by turning on the SyFy channel, which he half watches while sketching circuit designs. Or he might kick back with the latest IEEE publication, whether it's IEEE Spectrum or the IEEE Journal of Solid-State Circuits. And that is pretty much the extent of his leisure activities.
"I'm a bit narrow-minded," he says. "I only see things in terms of electronics."
Can that strict engineering approach continue to work for Marvell? It's not the scrappy little start-up anymore; it is now a formidable company with shareholders to please and quarterly targets to meet. For Marvell to continue to grow, says Sergis Mushell, an analyst with Gartner, the time has come for it to lead. He believes the company now needs to identify a killer app and open a new market on its own, rather than focusing on incremental changes to existing products. "That's not an easy thing," he says. "Many companies get to this point and struggle."
"Having a CEO so involved on the engineering side" may make it tough for original ideas to percolate through the organization, he adds. But so far, he says, Marvell is heading in the right direction. The company, with its platforms for tablet computers, smartphones, and e-readers, is trying to create additional markets for mobile, low-power processors.
And, while Sehat and his cofounders could certainly afford to retire tomorrow, Marvell remains their singular passion. Says Dai, "This is just the beginning for us. Our mission is long term. There's so much we ought to do. It's a 48-hour-a-day job still."
For Sehat, the need to keep creating semiconductor-based technology springs from something even more basic. "I don't know anything else," he says.
This article originally appeared in print as "Marvell Inside".