The Download: Feature Articles
NYU Expands Supercomputing Power with New HPC Cluster from AMD
By Keith Allison | June 8, 2020
The new cluster will be a powerful resource for NYU researchers fighting COVID-19
AMD and technology partner Penguin Computing Inc. announced the donation of a petascale supercomputer to NYU as part of the broad effort to combat COVID-19 and future pandemics. MIT and Rice University received a similar donation. AMD also announced it will contribute access to a cloud based system powered by AMD EPYC and AMD Radeon Instinct processors located on-site at Penguin Computing, providing remote supercomputing capabilities for selected researchers around the world.
“The COVID-19 pandemic has had a profound impact on higher education research, both in terms of its direction and the need for immediate results, so the timing of this donation is particularly fortuitous, and we’re tremendously grateful to AMD,” said Russel Caflisch, director of the NYU Courant Institute of Mathematical Sciences.
The computer is part of a larger initiative, the AMD COVID-19 HPC fund, which was established to provide research institutions with computing resources to accelerate medical research on COVID-19 and other diseases.
Caflisch said: “The computing resources donated by AMD will be put to use by NYU researchers from a wide range of disciplines in projects to address the many important facets of the COVID-19 crisis, including: discovery of drugs that may be therapeutic for COVID-19 and future SARS virus mutations, retrieval of relevant research results from the vast biomedical literature, analysis of medical imaging for screening of patients, and analyzing political attitudes and voting behavior in response to financial hardships.”
Power and Speed
The cluster consists of 20 compute nodes (servers), each equipped with an AMD EPYC Rome 7642 processor (having 48 processing cores), 512 gigabytes (GB) of host memory, 8 MI-50 32GB graphics processing units (GPUs) and 2 terabyte (TB) of local solid state disk (SSD) for data storage. In addition to the compute nodes, three nodes provide remote user access and cluster management. All cluster components are connected internally using the latest Infiniband network HDR technology providing a communication bandwidth of 200 gigabits per second (Gbps). The cluster can perform one quadrillion (1015) floating-point operations per second (a petaflop), requires over 60kW of power to run, and it will be deployed this summer in a heat-contained area in the new, energy efficient NYU research computing data center (RCDC).
This new computer, named Hudson, is the latest component to be announced in NYU’s expanding suite of high performance computing resources, which includes a powerful supercomputer, named Greene, as previously announced by NYU President Andrew Hamilton and scheduled to become operational in the fall, 2020. Both new supercomputers will be housed in a new state-of-the-art data center supported by a local, high-throughput, low-latency network dedicated to research.