Advanced Micro Devices officials for the past couple of years have been touting the chip maker’s ambidextrous computing strategy, the idea of offering both its traditional x86 processors and embracing the ARM architecture to give system makers and users options in the chip platforms they use.
Over the next two years, the company—which is in the middle of multi-year transformation of its business—will tighten that embrace of ARM, with plans for a framework dubbed “Project SkyBridge” that will more closely merge the two chip architectures. That will include making the x86 and ARM chips pin-compatible, meaning they will both be able to run on the same motherboard wilth few tweaks.
Over the next few years, x86—which run in the bulk of servers and PCs—and ARM’s low-power architecture, found in most mobile devices like smartphones and tablets, will be the dominant chip platforms, Lisa Su, senior vice president and general manager of AMD’s Global Business Units, said during a Webcast press conference in San Francisco May 5. The combined total addressable market for both x86 and ARM this year will hit $80 billion, and grow to more than $90 billion by 2018, according to AMD.
AMD will be the leader of that market, Su said.
“The bridge to the future is having both ARM and x86 in [AMD’s] portfolio,” she said.
Cloud computing, hyperscale data center environments, greater mobility, big data and other computing trends are forcing changes throughout the IT industry, including in the chip business that for the past decade has been dominated by Intel and its x86 architecture. ARM’s rise has presented Intel with a challenge not only in the booming mobile chip space, but now with plans to push into the niche ultradense server space.
IBM and four other vendors last year launched the OpenPower Foundation, a group looking to leverage a model similar to ARM’s—which designs chip architectures that it licenses to vendors like Samsung and Qualcomm—to expand the reach of Big Blue’s Power architecture beyond IBM’s servers and into other devices and systems. IBM executives see the consortium as a way of building a hardware and software ecosystem around Power to make it an alternative to Intel’s x86 chips.
In a blog post April 28, Gordon MacKean, chairman of the OpenPower Foundation, showed a photo of a motherboard server from Google running on a Power8 chip from IBM.
AMD officials announced in 2012 their intention to build ARM-based server chips, and in January introduced the Opteron A1100 “Seattle” chip and a software-developer kit. The company demonstrated a server powered by Seattle, a 28-nanometer system-on-a-chip (SoC) with eight ARM Cortex-A57 cores, and running a Linux variant from the Fedora Project.
Starting next year, AMD will begin rolling out new 20nm accelerated processing units (APUs)—which feature CPUs and GPUs integrated on the same chip—and SoCs that will offer pin-compatible x86 and ARM chips. The 64-bit ARM SoCs also will be based on the Cortex-A57 architecture and will be AMD’s initial chip to offer Heterogeneous System Architecture (HSA) features that will make it easier to move workloads between the CPU and GPU. It will be a platform for Android workloads. The x86 variant of the SkyBridge effort will include AMD’s next-generation Puma+ cores. The SkyBridge chips also will include AMDs Graphics Core Next technology, HSA features and Secure Technology.
In 2016, AMD will begin introducing high-performance, low-power “K12” chips based on ARM cores developed in-house by the chip maker. AMD will take out an ARM architectural license to develop the K12 chips.
AMD’s ambidextrous computing strategy is part of a larger push by the chip maker to move into several potential high-growth areas and reduce its reliance on a PC market that has seen more than two years of declining sales. When CEO Rory Read took over AMD 2011, PC chips accounted for about 95 percent of the company’s revenues. By the end of 2015, these new growth businesses—including dense servers, ultraportable devices, embedded solutions and semi-custom chips—will make up more than 50 percent of AMD’s revenues.
During the press conference, Read said there is “no doubt AMD is transforming. We’re building a differentiated AMD.”
Seattle is not AMD’s only use of ARM technology. The chip maker recently announced low-power x86 “Beema” and “Mullins” APUs aimed at notebooks and tablets. The chips also include what AMD officials call the platform security processor (PSP), based on the Cortex-A5 architecture from ARM that includes ARM’s TrustZone data security technology.