Road to Exascale Ends With Big News

For the last 10-plus years, our industry has been navigating the road to exascale. For the United States, that road hit a major milestone. Cray will deliver the nation’s first exascale supercomputer to the U.S. Department of Energy’s Argonne National Laboratory in 2021 in partnership with Intel. Named Aurora, the Argonne system will be comprised of more than 200 Cray® Shasta™ system cabinets, our unique software stack optimized for Intel architectures, the Cray® Slingshot™ interconnect, and Intel innovations in compute processor, memory, and storage technologies. As a subcontractor to Intel, our part of the program contract is valued at more than $100 million ― one of the largest in our company’s history. Are we proud? ... [ Read More ]

Critical Role of HPC Storage in Autonomous Vehicle Development

Self-driving cars were once only the stuff of science fiction. But thanks to recent computing advances, they are becoming more of a reality every day. Still, a number of issues need addressing before fully autonomous vehicles share our roadways. And one of the largest is capturing and processing the data generated by fleets of training cars. This massive amount of data creates an unprecedented storage challenge that autonomous vehicle development companies cannot overcome using their existing enterprise storage architectures. Luckily, here at Cray we’ve made it our mission to provide technology that solves the world’s most complex computing challenges, so this problem is right in our wheelhouse. First, a little background. Driving a ... [ Read More ]

Autonomous Vehicle Development Drives Extreme Storage Requirements

We can probably all agree that driving a car today is a radically different experience than it was just a few years ago. The next generation of vehicles – self-driving cars – are poised to take modern automotive features a quantum leap forward and provide totally different ways to get from point A to point B. One of the reasons fully autonomous vehicles are still a few years off is that designing an AI model that can safely and accurately drive a car is turning out to be an incredibly complicated task. The two main challenges are simulating every possible driving scenario that could occur so the model can be trained to identify the appropriate reaction, and handling the vast volume and variety of data generated by fleets of training ... [ Read More ]

How Hyperparameter Optimization Improves Machine Learning Accuracy

In January, a team of Cray developers and researchers published a paper, “Recombination of Artificial Neural Networks,” on arXiv.org, highlighting the hyperparameter optimization (HPO) capability Cray announced in November. We cover their findings in this blog post. Using a variety of high-performance computing systems and neural network models, the Cray team demonstrated that the hyperparameter optimization capabilities introduced in the Cray® Urika®-CS and Urika®-XC AI and analytics software suites improve the time-to-accuracy as well as final accuracy of machine learning models trained on Cray systems. The table below, excerpted from the paper, highlights the improvements achieved using Cray’s HPO capability across a range of ... [ Read More ]

Deep Learning at Scale Using Cray Distributed Training

This article was written by the following Cray and NERSC contributors: Steve Farrell, Machine Learning Engineer; Thorsten Kurth, Application Performance Specialist; Jacob Balma, Performance Engineer; Peter Mendygral, Performance and Software Engineer; Nick Hill, Software Engineer. Deep neural networks (DNN) are revolutionizing science across many domains including high energy physics, cosmology, biology, and climate. As the field of deep learning advances, DNN architectures grow more sophisticated and capable of solving complex tasks in scientific problems such as classification, regression, and simulation. Training and evaluating such models requires increasingly large datasets and computing resources. Through the NERSC Big Data ... [ Read More ]