Updated: AMD (s amd) is debuting a new lineup of chips for servers, including its first ever ARM-based part. For AMD, which built its business around a license of Intel’s (s intc) x86 architecture, building an ARM part is both a declaration of independence and a necessary step for the chipmaker struggling to reinvent itself.
The chip firm is introducing three new parts for servers. One is a more muscular core aimed at the traditional enterprise computing and high-performance-computing market. AMD dubs this chip Warsaw.
The other two parts, Seattle and Berlin, are less powerful and designed more for webscale workloads that can be parallelized. Berlin is a smaller chip that uses an Atom-like CPU core with AMD’s APU graphics core, while Seattle is the 64-bit ARM-based chip that could bring AMD out from Intel’s shadow. The Seattle chip will be in servers in the second half of 2014 and have eight and eventually 16 cores using the ARM Cortex A-57 design that can run at up to 2 GHz per core.
Andrew Feldman (pictured above), the corporate vice president and general manager of AMD’s server business unit, and a speaker at the GigaOM Structure event happening this week in San Francisco, says it’s clear that ARM will have a place in the data center, but he understands that he needs to make a case for AMD as a viable builder of ARM-based chips for that market.
He thinks that AMD’s experience building server chips will help it, as will his experience building out networking fabrics at SeaMicro, the company AMD purchased last year as part of a bid to get into the more power-efficient and dense server architectures. Feldman and I have discussed the changes in the data center — from small workloads to the need for power efficiency — for the last three years. It’s nice to see this vision closer to playing out in the mass market.
It’s too early to say how AMD will stack up against the myriad other vendors trying to build out ARM-based chips for the data center — or even how it will continue to stack up against Intel. But the new chips are a clear attempt to bring AMD into the new era of cloud computing. It’s a welcome step.
This story was updated Tuesday t 6am PT to correct errors about the chips names and cores.