Intel is expected to announce availability next week of a low-power Atom server chip code-named Avoton, which is likely to appear soon in systems such as Hewlett-Packard's Moonshot.
Avoton chips will become part of Intel's Atom C2000 product family and be used in high-density servers from multiple vendors, said a source familiar with Intel's plans.
Intel is expected to announce the chip at an event in San Francisco about hyperscale data centers. Other Atom chips expected to launch next week include the "Rangeley" processor for networking systems.
The Avoton chips will succeed Intel's first line of Atom S1200 server chips, code-named Centerton, which are used in HP's current Moonshot server. Avoton will be based on the Silvermont architecture, the first major architectural redesign for Atom in six years. Intel's upcoming tablet chip, code-named Bay Trail, is also based on Silvermont.
Intel has claimed Avoton will be up to seven times faster and four times more power efficient than Centerton, which shipped in December. Avoton will have up to eight cores, up to 64GB of memory, and include integrated Ethernet networking. Avoton will be manufactured using a 22-nanometer process. Actual figures for power consumption and clock speed have not yet been released.
Avoton is Intel's answer to an emerging threat from ARM in the market for high-density servers. ARM chips are based on a different instruction set and are found in many smartphones and tablets. HP has said it will also use ARM processors in Moonshot, and chip makers Calxeda, Advanced Micro Devices, Nvidia and AppliedMicro are expected to release ARM chips in the next year.
A collection of low-power chips packed into racks of high-density servers are considered optimal for workloads comprising a high volume of lightweight transactions, such as serving search results or social network posts. Intel has something of an early mover advantage along with support for 64-bit computing, but that's coming soon to ARM chips. Some other features in Avoton may lend themselves to slightly more heavyweight tasks, such as transaction processing for databases.
