AMD's first ARM-based server chip is set to start sampling in "a few weeks." The news comes from Andrew Feldman, GM of the company's Server Business Unit, who unveiled the Opteron A1100 series at the Open Compute Project summit in San Jose today.
Code-named Seattle, the chip comes in configurations with four or eight CPU cores based on ARM's 64-bit Cortex-A57. The top model has 4MB L2 and 8MB L3 caches, and its clock speed will be at least 2GHz. Interestingly, Seattle boasts a configurable dual-channel memory controller that works with DDR3 or DDR4 memory at up to 1866 MT/s.
On the connectivity front, the ARM-based Opteron has eight lanes of PCI Express 3.0. Eight 6Gbps SATA ports provide storage, and there are dual 10-Gigabit Ethernet controllers onboard. The chip also includes dedicated co-processors to assist with encryption and compression duties. It doesn't appear to have any Radeon-derived compute goodness, though.
After introducing Seattle, Feldman turned his attention to the Opteron A1100 Series developer kit, which includes a microATX motherboard loaded with a Seattle chip, four registered DDR3 DIMM slots, eight SATA ports, and dual PCIe slots (in an x8/x0 or x4/x4 configuration). Also on the menu: UEFI-based firmware and a Fedora-based Linux distro. Interested developers should be able to get their hands on the dev kit soon.
Every tech presentation needs One More Thing, and Feldman had something up his sleeve—or rather, in his pocket. He pulled out an expansion card-sized device that squeezes an eight-core Seattle server into a chassis compliant with the Open Compute Project's Common Slot specification. Feldman said the IP behind the server-on-a-stick design would be released to the Open Compute community. He also predicted that ARM-based processors would make up 25% of the server market by 2019. AMD "will be the leader in ARM CPUs for servers," he pledged.