By Jack Wells
Here at the Oak Ridge Leadership Computing Facility (OLCF) in East Tennessee, deploying the next top supercomputer for open science is akin to an ambitious hike in the Smoky Mountains: once one towering crest is reached, the next one appears within sight.
Just 18 months after the OLCF brought Titan—then the fastest supercomputer in the world—to full operation for users in May 2013, we announced a contract with IBM to create the next big machine: Summit.
Summit will expand on Titan’s groundbreaking hybrid architecture to deliver several times the computational power of the 27-petaflop Titan.
Navigating the peaks of launching new massive machines requires planning, ingenuity, and a certain affinity for risk. It’s the latter quality that steered the OLCF toward IBM’s data-centric approach to computing. The partnership allows for the evolution of Titan’s heterogeneous architecture, which integrates novel graphics processing units (GPUs) and conventional central processing units (CPUs) at unprecedented scale.
Through initiatives like the Innovative and Novel Computational Impact on Theory and Experiment, or INCITE, program, our users have employed GPUs to tackle some of the world’s most pressing challenges at scales and speeds that were previously prohibitive.
Recent scientific accomplishments achieved under INCITE include advances in plasma physics led by C.S. Chang of Princeton Plasma Physics Laboratory, human skin modeling led by Michael Klein of Temple University and Proctor & Gamble, and mapping of the Earth’s interior led by Jeroen Tromp of Princeton University. These projects leveraged Titan’s GPU accelerators to achieve greater simulation speed and increased fidelity, reaching solutions in less than a quarter of the time—and using a fraction of the energy—it would have taken on a CPU-only architecture.
Jointly managed by the OLCF and Argonne Leadership Computing Facility (ALCF), INCITE is currently accepting proposals for 2016 from US and non-US based researchers for projects that require leadership computing resources. The submission process is now open and continues through June 26.
With Summit, the role of GPUs is evolving and will dramatically improve data movement between Summit’s NVIDIA Volta GPUs and IBM POWER CPUs
IBM’s experience in HPC and focus on minimizing data movement and energy consumption will play a crucial role in Summit’s success. Work has already begun in anticipation of Summit’s 2018 arrival with OLCF staff members familiarizing themselves with the machine’s next-generation software and hardware, including IBM’s Elastic Storage System, using small-cluster test systems.
Titan continues to demonstrate the value of GPU accelerators and advance science. Summit is the next step in GPU integration and the next peak in HPC. It will undoubtedly lead to even faster scientific research and results.
Oak Ridge National Laboratory is supported by the US Department of Energy’s Office of Science. For more information about INCITE or to submit a proposal, click here.