Perlmutter (supercomputer)

Perlmutter is a petascale supercomputer located at the National Energy Research Scientific Computing Center (NERSC) at Lawrence Berkeley National Laboratory. Named after Nobel laureate Saul Perlmutter, it is one of the world's fastest supercomputers, designed to accelerate scientific discovery across a wide range of disciplines.

Overview:

Perlmutter is a heterogeneous system comprised of two main partitions: a GPU partition and a CPU partition. The GPU partition is based on the NVIDIA A100 Tensor Core GPUs and is designed for artificial intelligence, data analysis, and simulation workloads. The CPU partition, on the other hand, utilizes AMD EPYC processors and is optimized for traditional scientific computing tasks.

Hardware Architecture:

• GPU Partition: This partition features thousands of NVIDIA A100 GPUs interconnected by NVIDIA's NVLink technology. This provides high bandwidth and low latency communication crucial for deep learning and other GPU-accelerated applications.
• CPU Partition: The CPU partition consists of a large number of AMD EPYC processor-based nodes. These nodes are interconnected with a high-speed network to enable efficient communication for distributed computing.
• Storage: Perlmutter is equipped with a substantial amount of storage capacity. A significant portion of this storage is dedicated to a flash-based all-NVMe file system called "Durable Flash Archive" or DFA, providing extremely fast I/O performance for data-intensive workloads.
• Interconnect: A high-performance Slingshot interconnect network links all nodes within Perlmutter, allowing for rapid data transfer and communication between different components.

Scientific Applications:

Perlmutter is used by researchers to tackle complex problems in diverse fields, including:

• Cosmology: Simulating the evolution of the universe and studying the distribution of dark matter and dark energy.
• Climate Modeling: Developing more accurate climate models to predict the impacts of climate change.
• Materials Science: Designing new materials with specific properties for a variety of applications.
• Fusion Energy: Simulating plasma behavior to advance the development of fusion energy.
• Bioenergy: Optimizing the production of biofuels and bioproducts.
• Drug Discovery: Accelerating the identification of potential drug candidates.

Impact:

Perlmutter enables researchers to perform simulations and analyze data at unprecedented scales, leading to new scientific insights and breakthroughs. Its advanced capabilities contribute significantly to the advancement of scientific knowledge and address some of the world's most pressing challenges.

NERSC:

Perlmutter is a key resource within NERSC, which provides high-performance computing resources and expertise to researchers funded by the U.S. Department of Energy's Office of Science.