SAN JOSE, Calif. — ARM-based servers are coming, but even more slowly and perhaps more narrowly than once thought, according to companies that expect to build and use them.
"There will definitely be [ARM] server products shipping this year and a reasonable number next year, but it won't really begin to ramp until 2016," Forrest Norrod, general manager of Dell Inc.'s server group, told EE Times.
There's still plenty of work ahead to bring to ARM the wealth of server software that exists for the Intel x86, said Norrod and Wu Peng, a chief technologist for the datacenter group at the Chinese e-commerce company Alibaba. For example, Microsoft has yet to commit to supporting ARM with Windows Server.
ARM servers "need a lot of software preparation and lack system integration support," Wu said in an impromptu interview after a talk at the recent Flash Memory Summit. The 32-bit versions of ARM lack data management capabilities, but ARM's 64-bit performance "looks impressive."
Big data centers such as Alibaba will use 64-bit ARM chips as storage controllers first, Wu said. Intel already has a leg up here. Its Atom-based Avoton SoC is already in production as a storage server in some of China's data centers, he said.
In addition to rolling out lower-power Atom-based server SoCs, Intel has expressed willingness to customize its Xeon server processors for large data center customers. Wu said he met with Intel engineers in Portland, Ore., to discuss the possibility of making a lower-cost Xeon with a smaller L3 cache. "We have some scenarios that cannot benefit from a bigger L3 cache, so why introduce it?"
Next page: Turning on a LAMP