BOSTON—ARM will take significant market share from Intel in the server chip space over the next few years, and Advanced Micro Devices will be a key vendor benefiting from the shift, according to the head of AMD’s server business.
In an interview with eWEEK here, Andrew Feldman, corporate vice president and general manager of AMD’s Server Business Unit, said the rapidly changing compute demands and decades of tech industry history are creating an environment that’s ideal for ARM’s strong ascent into the data center.
The rise of cloud computing, the rapid growth in the number of connected mobile devices hitting the Internet, and the demands of massive Web 2.0 companies like Google and Facebook are quickly changing the kinds of workloads that are running on servers, Feldman said. Data centers increasingly are supporting millions of users with servers running massive numbers of parallel workloads that need the type of small-core chips that ARM and its partners are manufacturing to power the bulk of the world’s smartphones and tablets. And that trend will only continue, he said.
Couple that with the history of the tech industry, where new smaller, more cost-effective technologies always beat larger legacy technologies—think x86-based servers winning out over mainframes and RISC machines, or x86 chips overcoming SPARC processors—and the idea of ARM’s low-power system-on-a-chip (SoC) architecture making significant inroads into the dominant x86 server chip market share makes sense.
“In the history of compute, smaller, more efficient, higher volume has always won,” Feldman said. “In my view, a 40-year record of unblemished success is something you have to pay attention to.”
In addition, thanks to the large number of chip makers that license ARM’s architecture, innovation in the ARM market is faster than in the x86 space, which is dominated by Intel, according to Feldman.
“[Server] customers are eager to see the vibrant ecosystem they’ve seen on the client side,” he said.
ARM officials have been talking for several years about moving their architecture into the server space to meet the increasing demand for more energy-efficient chips to power smaller, more dense systems that can process huge amounts of small workloads. Already some ARM partners—such as Calxeda and Marvell Technologies—already are offering server SoCs based on ARM’s 32-bit architecture.
But it won’t be until next year, when chips based on ARM’s upcoming ARMv8 architecture—which includes such server capabilities as 64-bit, improved virtualization support and more memory—begin appearing in systems that ARM servers will begin getting real traction in the data center. That will happen relatively quickly, Feldman said, predicting that by 2016, ARM could hold 20 percent or more of the server chip market.
“This new environment is going to have new needs, and the same-old, same-old will not work anymore,” he said.
AMD has revamped its server strategy to meet those needs. It will continue to innovate in its traditional x86 Opteron business, but also is making its own low-power microservers—thanks to its acquisition last year of SeaMicro—and working with the open hardware movement, as illustrated by its Open Compute 3.0 “Roadrunner” motherboard. In addition, beginning next year, it will begin making ARM-based server chips.
“We’ve relaxed our religious commitment to x86 and we’ve embraced ARM,” Feldman said.
AMD’s Feldman: ARM Will Quickly Gain Market Share in Servers
It will give AMD new business avenues to pursue, but it also will put the company into tighter competition with other ARM chip vendors. However, AMD has a history of making server chips, strong relationships with system OEMs and ODMs, and a large amount of IP that it can draw on, he said. In addition, AMD’s history is all about spending the necessary money and effort to come out with upgraded chips every 12 to 18 months, a costly and difficult proposition for smaller companies like Calxeda and Applied Micro.
“The CPU business is no place for small companies,” Feldman said. “It’s just too expensive.”
However, he acknowledged that larger ARM SoC makers Samsung and Qualcomm will be formidable rivals.
Intel is aggressively going after the microserver business with its low-power x86-based Atom platform. The company last year launched its “Centerton” Atom SoCs, and later this year will release the next generation, dubbed “Avoton.” In addition, while Hewlett-Packard is planning for low-power microservers powered by ARM chips from AMD and others as part of its Project Moonshot effort, the first systems from the initiative will be powered by Intel chips. HP officials have said a Moonshot cartridge powered by AMD’s upcoming low-power “Kyoto” chip will come out in the second half of 2013.
Intel officials, pointing to Avoton, have noted that the giant chip maker will be well into its second generation of Atom SoCs for microservers by the time the first ARM-based systems will hit the market.
However, Feldman said there is a lot of momentum behind the ARM server push. Not only does the computing environment and history favor it, but also systems makers like HP and Dell—both of whom are developing low-power servers that will run on ARM SoCs—want ARM to succeed in the data center, as do a rapidly growing group of end users, such as Facebook and Google.
He said he is excited to see what will happen over the next few years as ARM’s presence in the data center grows.
“I think we’re going to see unbelievable [stuff] with this,” Feldman said. “It’s going to be spectacular.”