News around the Linux on Power Community
Michael Gschwind 060001TJ2H 9,398 Views
Michael Gschwind 060001TJ2H 8,592 Views
Supercomputing 16 had so many exciting announcements and events around AI and Deep Learning on Power, it was hard to not miss some of the highlights. As I previously shared, SC16 started with the launch of PowerAI on Monday morning. PowerAI and the new GPU enterprise server were favorite attractions of SC16 attendees at the large, centrally located IBM booth during the conference with large crowds interested in learning more about the PowerAI and our custom-designed GPU enterprise server “S822LC for HPC”.
But Deep Learning innovation on Power did not with IBM’s PowerAI offering. True to the collaborative spirit of OpenPOWER, we had been working with many partners to create a broad ecosystem around PowerAI and Deep Learning, several of who were present as IBM guests at the booth. We launched the Julia language on Power at Supercomputing 16 with our partners at Julia Computing. The combination of Julia and Power provides a perfect combination for Deep Learning, as Julia Computing CEO Viral Shah explained to HPCwire: “Using IBM’s Power platform with NVIDIA GPU accelerators increased processing speed by 57x – a dramatic improvement. IBM Power provides 2-3x more memory bandwidth combined with tight GPU accelerator integration to create a high performance environment for deep learning with Julia.” At the IBM booth, Julia Computing demonstrated how to take advantage of the unprecedented power of Julia and IBM’s GPU-accelerated Power servers with “Deep Eyes”, an exciting public health application built on Power, Julia and the MXnet framework that can diagnose eye disease with a cheap camera and automated diagnose that can refer affected patients to specialists for treatment.
NIMBIX unveiled their Jarvice high-performance cloud platform on Minsky, with a “push to compute” interface to create a Minsky-powered high performance PowerAI instance at the push of a button.
Other exciting AI and Deep Learning demos included realtime video analysis for image segmentation using a CAPI-attached FPGA accelerator and a realtime recommendation engine with AI using the S822LC for HPC system.
To get started with your own Deep Learning applications, learn more about the S822LC for HPC. Share your ideas on how to use Deep Learning and the Power of the S822LC for your application use cases in the comments section below.