Marvell chairman, president and CEO Dr. Sehat Sutardja took the keynote stage yesterday at the ARM Technology Conference and told the audience how Marvell and ARM came to be partners. It’s an instructive story, because it shows how Sutardja thinks. This story suggests—at least partially—how Sutardja has pushed Marvel to become the semiconductor power that it now is. Sutardja and his partners founded Marvell in 1995. Back then, the major microprocessor architectures in vogue were the x86, PowerPC, and MIPS. The x86 fueled the rise of the PC. A PowerPC heart beat inside of Apple’s Macintoshes and the architecture owned the networking and communications landscape. MIPS had made huge inroads in high-volume game consoles.
And then there was ARM. In 1995, the ARM core had barely penetrated the mobile phone handset market. The whole idea of SoCs was just forming and most ASICs did not include processors because there weren’t nearly enough gates available on chip. The RISC-versus-CISC wars were raging and everyone, it seems, was trying to replace the PC processor because of the attractive sales volumes and unit pricing. The PC market was juicy and tempting indeed.
And then there was the ARM core. Originally, “ARM” was a three-letter acronym for “Acorn RISC machine.” It was a simple 32-bit RISC processor based on the RISC work at the University of California at Berkeley that had been developed as a CPU for Acorn’s personal computers, sold almost exclusively in the UK during the latter half of the 1980s. In the late 1980s, Apple was casting about for a processor core to use in the design of the Apple Newton PDA and found ARM. However, the ARM processor wasn’t a core back then, it was a chip. Apple engaged Acorn and VLSI Technology to develop the original ARM processor into something that could be used in the Newton. The net result of that work was a licensable processor core and a company named ARM.
The ARM processor core ended up in cell phones because it was small and sufficiently capable but in 2000, when Sutardja started to look at processor cores to put into Marvell’s disk-controller chips, the ARM core was only running at 15-20MHz. Sutardja remembers thinking that the ARM processor was underpowered but that it had potential. Marvell became the second company to obtain an architectural license for the ARM processor, after Digital Equipment Corp.
Since then, Marvell has shipped more than one billion chips containing ARM processors. At first, it was disk controllers, designed at a time when disk controllers contained no processors. Now they all do. Smart is better than dumb—that was the real message within Sutardja’s keynote. Now, Sutardja foresees ARM processors in pretty much everything Marvell designs—chips for mobile devices, communications and network infrastructure equipment, cloud computing, information appliances, and smart furnishings.
“Today is the era of superintegration,” said Sutardja. He foresees chips with multiple cores and accelerators. “These chips all tend to look the same,” he said. “They just vary in peripherals.” Marvell has all of these pieces, said Sutardja, and can assemble them in various configurations for customers. Here, Sutardja’s vision for Marvell echoes the EDA360 concept of SoC Realization, where processors, accelerators, memories, peripherals, and interconnect are plug-together parts that can be readily combined and recombined to produce appropriate designs in silicon.
“There’s still lots to be built,” said Sutardja. “Life is mobile and there’s no reason to make a dumb smartphone because the component savings is minimal.” This idea echoes the EDA360 concept of apps-driven software differentiation. “Everything else will be smart too,” continued Sutardja. You don’t have to have dumb TVs, dumb set-top boxes, or dumb cars. Consumers are rapidly becoming accustomed to smart products and they are beginning to prefer them, to expect them.
Sutardja concluded: “Now, we can build chips to make these products smart at low enough cost.” I think you’ve just heard Marvell’s product strategy for the foreseeable future.