Hewlett-Packard Enterprise has revealed its first Arm-based server for high performance computing, Apollo 70. The machine is built around Cavium’s 64-bit ARMv8-A ThunderX2 chips and will become available in 2018.
The company has also launched Apollo 2000 Gen10 server nodes for entry-level x86-based HPC systems, and Apollo 4510 Gen10 storage systems for big data analytics, both available immediately.
The new range was presented at the Supercomputing ’17 conference in Denver, Colorado.
Once more unto the breach
HPE claims to be the leader in HPC, quoting research by IDC that puts its market share at 36.4 percent.
Apollo 70 is intended to introduce more choice in this space and offer an alternative to systems based on Intel’s Xeon CPUs. According to HPE, the specialized server was built for memory-intensive workloads and delivers up to 33 percent more memory bandwidth than today’s “industry-standard servers.”
Apollo 70 builds upon the work done by Project Moonshot – HPE’s original attempt to introduce Arm into the data center, as part of a new class of devices called ‘microservers.’ In 2013, Moonshot was billed by HP executives as the most important server invention in a decade – but it never made it as a general-purpose computing platform.
At the same time, there has been a surge of interest in Arm as a platform for HPC. For example, Japan’s 1,000-petaflop next generation supercomputer, known as Post-K, will be powered by Arm CPUs. The EU’s Mont-Blanc Project is about to release its third prototype, featuring 3,000 Arm cores and UK’s upcoming Isambard supercomputer will feature more than 10,000 Arm cores.
“HPC and AI applications are not exclusive to big research organizations and corporations; they can drive efficiency and innovation in the day-to-day business of every company,” commented Bill Mannel, vice president and general manager of HPC and AI Segment Solutions at HPE.
“Today, HPE is augmenting its proven supercomputing and large commercial HPC and AI capabilities with new high-density compute and storage solutions to accelerate market adoption by enabling organizations of all sizes to address challenges in HPC, big data, object storage and AI with more choice and flexibility.”