SAN FRANCISCO—Intel officials next year will introduce a low-power version of its powerful Xeon E3 server chip as it looks to stave off competition in the data center from Advanced Micro Devices as well as ARM and its list of partners.
The company already offers systems-on-a-chip (SoCs) for dense, low-power servers based on its Atom platform; Intel last year launched Centerton, this year will release the 22-nanometer "Avoton" and "Rangeley" chips, and in 2014 will unveil the 14nm "Denverton." However, the company also will roll out an SoC version of its "Broadwell" Xeon processor, which will include such integrated features as fabric, I/O and accelerators for servers, storage devices and networking hardware, according to Diane Bryant, senior vice president and general manager of Intel's Data Center and Connected Systems Group.
Jason Waxman, vice president and general manager of Intel's Cloud Platform Group, said the new Broadwell SoCs will not compete with the company's Atom platform, but will more likely be "an extension of where Avoton plays."
Avoton is the upcoming 22-nanometer version of Intel's Atom-based server chip, due out later this year.
Intel released the news at a conference here July 22, where Bryant and other executives gave more details about the company's strategy for a data center environment that is undergoing rapid change brought on by such trends as cloud computing, mobility and big data.
The Atom and upcoming Broadwell SoCs are aimed at organizations—like cloud service providers and Web 2.0 companies such as Facebook and Amazon—that are looking for small, high-performance and energy-efficient servers that can be used in their huge data centers to help process massive numbers of small workloads. System makers like Hewlett-Packard and Dell already have projects under way to develop such microservers.
At the event, Waxman showed off an Avoton SoC, which includes eight cores and offers integrated Ethernet and support for up to 64GB of memory. The Avoton SoCs will be used in servers, while the Rangeley chips will be aimed at networking devices. Both will be part of Intel's C2000 product family.
Officials with ARM and its partners—including Samsung, Nvidia, Calxeda and, starting next year, AMD—believe that ARM's highly energy-efficient architecture, which powers most smartphones and tablets, can fit well in such microservers. Some vendors, like Calxeda, already offer server chips based on ARM's current 32-bit designs. However, next year, ARM's 64-bit ARMv8 architecture—which also will feature such capabilities as greater virtualization support and memory—will begin appearing in systems.
ARM on July 22 announced an expanded partnership with Oracle to further optimize Java Platform, Standard Edition for its 32-bit designs, and support Java SE in the upcoming 64-bit architecture.
AMD in May announced its x86-based Kyoto Opteron chip for the microserver market, and next year will begin offering products based on ARM's architecture as well.
The evolving Atom and Xeon products are part of Intel's larger push to shift away from its traditional process of introducing new general-purpose processors that OEMs put into their servers. Instead, the chip maker is delivering a broad portfolio of offerings optimized for particular workloads based on their needs for compute performance and power efficiency. In addition, Intel also is looking to expand its reach beyond servers and into all facets of the data center, from storage to networking to software.
"Our goal is [to ensure] that all workloads, regardless of what they are, run best on Intel architecture," Bryant said.
Intel executives argued that Intel offers a greater range of products, better software compatibility and more software tools to organizations than ARM can.
Charles King, principal analyst with Pund-IT Research, said that ARM and AMD are formidable foes and that Intel will need to work hard to keep them at bay. However, Intel's argument of compatibility and familiarity—if it can offer a wide range of products that meet the needs of organizations, why would they want to switch silicon architectures and have to deal with new operating systems, software tools and applications—is a strong one.
"From an architectural standpoint, it makes sense," he told eWEEK. "Whether the market agrees with me, that's another point."