Researchers at Oak Ridge National Laboratory will have a new supercomputer to work with in 2017 when IBM delivers Summit, a hybrid CPU/GPU computer. It will have at least five times the performance of Titan, the largest supercomputer currently housed at ORNL.
Summit will use an open-architecture technology called OpenPOWER Foundation that is being developed by the major vendor, IBM, and component suppliers NVIDIA and Mellanox. The supercomputer will run software that includes IBM XL, NVIDIA, and PGI environments supporting OpenMP and OpenACC programming, and IBM HPC software including Linux, Platform Computing LSF scheduler, resource manager, system management, and a GPFS parallel file system.
Watch a video on the Summit supercomputer below, via Engineering TV:
From a hardware perspective, Summit builds on the hybrid multi-core architecture the lab pioneered with Titan. “The large, powerful nodes allow applications to achieve very high performance without having to scale to hundreds of thousands of Message Passing Interface tasks,” says director of the Summit project at the OLCF Buddy Bland. “The combination of very large memory per node and the powerful IBM POWER and NVIDIA processors provides an ideal platform for data analysis as well as computation.”
It will feature more than 3,400 nodes, each with:
- Multiple IBM POWER9 processors and NVIDIA Volta GPUs.
- CPUs and GPUs connected with high-speed NVLink.
- Large coherent memory: More than 512-GB of combined DDR4 and high bandwidth memory, all directly addressable from the CPUs and GPUs.
- An additional 800-GB of NVRAM, which can be configured as either a burst buffer or extended memory.
- Over 40 teraflops peak performance.
The areas and problems that will be explored with Summit include:
Combustion science: Understanding combustion so that the efficiency of internal combustion engines can be increased by 25-50% with lower emissions by using advanced fuels and new, low-temperature combustion concepts.
Climate-change science: Understanding the dynamic ecological and chemical evolution of the climate system with uncertainty quantification of the effects on regional and decadal scales.
Energy storage: Exploring chemical reactions at the atomic and molecular level required to design new materials for energy storage and engineer safe, large-format, durable, rechargeable batteries.
Nuclear power: Simulating reactor-scale operations to determine safe, increased nuclear fuel burn times, power upgrades, and reactor lifetime extensions, and thereby reduce the volume of spent fuel.