This guest post from Intel explores some of the company’s offerings at this week’s ISC 2018 High Performance conference in Frankfurt, Germany.
It’s a fascinating time for High Performance Computing. The convergence of AI, analytics, and simulation and modeling is transforming systems and users. So, there’s a lot to talk about it at ISC 2018, including announcements from Intel on next-generation Intel® Omni-Path Architecture and Intel® Select Solutions for Professional Visualization, to Machine Learning Day with Pradeep Dubey, and demos, presentations, and panel discussions in Intel’s Collaboration Hub (booth F930). Check out the full schedule of Intel presentations at ISC 2018.
A New Intel® Select Solution for HPC—Software-Defined Visualization
As we move toward the Exascale era, advances in computing have presented a new challenge for the industry: our ability to generate data is increasing faster than our ability to consume and analyze it. This week at ISC, Intel announced Intel® Select Solutions for Professional Visualization, an easily-deployed, Intel-validated system that is purpose-built for the demands of today’s most complex HPC visualizations. Intel Select Solutions for Professional Visualization allows the platform’s memory to do graphical rendering of large datasets in real-time. A greater footprint of available memory makes these Intel solutions more ready for the larger data sets in HPC workloads than architectures with smaller, captive memory pools.
Next Generation Intel® Omni-Path Architecture Coming
Intel® Omni-Path Architecture (Intel® OPA) fabric, the leading 100 Gbps interconnect on the Top500, is deployed worldwide today for a wide range of HPC and deep learning applications. The next-generation of 200 Gbps Intel OPA is coming in 2019. It will be interoperable with existing Intel OPA installations. Find out more at the Intel booth.
Special Presentation: Supercomputing Circa 2025, Dr. Raj Hazra (Monday, June 25, 6pm – 6:45pm)
Dr. Hazra will explore how the convergence of AI, analytics, and simulation and modeling will require new system architecture approaches for HPC.
Industrial Day, Dr. Marie-Christine Sawley, Chair (Tuesday, June 26)
Learn how leading enterprises drive change in their IT environment to address future business needs and how they measure progress toward IT optimization.
Machine Learning Day
On Scale-out Deep Learning Training for Cloud and HPC, Pradeep Dubey (Wednesday, June 27, 1:45pm – 2:15pm)
Scaling Stochastic Gradient Descent is still a challenging problem. We describe the Intel® Machine Learning Scalability Library and present proof-points demonstrating scaling deep learning training on 100s to 1000s of nodes across Cloud and HPC systems.
Collaboration Hub
In Intel’s Collaboration Hub, you can catch up on the latest HPC and AI innovations and information with insightful talks and panel discussions covering a range of compelling topics, including the convergence of HPC and AI, HPC in the Cloud, HPC for manufacturing, and the newest technologies and solutions. Here’s a sampling:
- Towards a Modular Supercomputer Architecture for HPC and Data Analytics from Jülich Supercomputing Center
- Panel: AI in HPC moderated by Binay Ackalloor, Intel Director, Artificial Intelligence and Machine Learning
- Panel: HPC in the Cloud moderated by Bill Magro, Intel Fellow and Chief Technologist
- Scaling Up Deep Learning by SURFsara
View Booth Demos Showcasing the Latest Technologies
- HPC and AI in Action
Intel architects and developers are always innovating to improve performance and outcomes in HPC. Demonstrations will showcase exciting developments in several fields. Check the schedule in the booth for times. - Artificial Intelligence
Watch how Intel® Xeon® Scalable processor-based servers achieve dramatic improvement in segmentation of tumor profiles in brain images. - Software-Defined Visualization
See exciting new software visualization solutions for HPC highlighting OSPRay, AI Denoiser and OpenFOAM*. - Application Acceleration with Intel® FPGA
Learn how FPGAs offer higher performance and lower power than GPUs with demos of various types of acceleration. - Distributed Deep Reinforcement Learning
See comparisons of various training times utilizing TensorFlow optimized for Intel® Xeon® Scalable processors.
