HPE to release massively scalable systems to power AI
Hewlett Packard Enterprise (HPE) is getting ready to release three new High Performance Computing systems optimized for massively parallel processing and AI-related tasks.
Aimed at research and national security organizations that need to crunch huge amounts of data, the new HPE Apollo 6000 Gen10, HPE SGI 8600 and HPE Apollo 10 Series systems will be available worldwide from July 2017.
The HPE SGI 8600 is a liquid-cooled, petascale system, based on the legacy SGI ICE XA architecture, that HPE says is built to solve the most complex scientific, engineering and national security challenges.
The HPE SGI 8600 scales to more than 10,000 nodes without the need for additional switches and supports arrays of liquid-cooled Nvidia Tesla GPU accelerators, linked via high-speed NVLink interconnects.
The HPE Apollo 6000 Gen10 System is the latest version of HPE's platform, redesigned to offer more than 300 teraflops per rack, for higher rack-scale efficiency and better price-performance. These improvements over its predecessors have been realized by reducing latency and increasing IOPs performance, as well as by reducing power consumption and cooling requirements.
The system also includes HPE's "silicon root of trust", which prevent servers from booting using any firmware that doesn't match the fingerprint built into the system's integrated lights-out chip.
The HPE Apollo 10 Series encompasses systems optimized for entry level Deep Learning and other AI applications, and that are designed to be relatively simple to manage and deploy.
Within the series, the HPE Apollo sx40 System is a 1U dual socket Intel Xeon Gen10 server with support for up to 4 NVIDIA Tesla SXM2 GPUs with NVLink. Meanwhile, the HPE Apollo pc40 System is a 1U dual socket Intel Xeon Gen10 server with support for up to 4 PCIe GPU cards.
Alongside the new systems, HPE has also upgraded its Performance Software Suite - which brings together a range of utilities for provisioning, managing, optimizing and monitoring high performance computing systems.
HPE continues to invest in memory-driven computing research, with the hope it will boost enterprise applications, and last week was awarded a US research grant to continue work on developing an exascale supercomputer called The Machine.