March 17, 2022
Staff Accomplishment

Ang to Support DOE Panel on Industry Collaboration

Ang will be part of a virtual discussion regarding AI hardware

CENATE image

PNNL’s collaboration with Micron aims to enhance artificial intelligence applications on high-performance computers such as those pictured here.

(Composite Image by Timothy Holland | Pacific Northwest National Laboratory)

On March 18, Pacific Northwest National Laboratory (PNNL) Chief Scientist for Computing James (Jim) Ang will participate in a virtual discussion led by the Department of Energy’s Office of Science on “Driving U.S. Competitiveness and Innovation: A New Era of Science for Transformative Industries.” Ang will support Micron Technology Corporate Vice President of Advanced Computing Solutions Steve Pawlowski in highlighting PNNL’s collaboration with Micron, a domestic memory and storage semiconductor company, to a U.S. Senate Committee on the United States Innovation and Competition Act.

PNNL and Micron collaborate on a project called Advanced Memory to Support Artificial Intelligence (AI) for Science (AMAIS). Computer Scientist Andrés Márquez from PNNL’s Advanced Computing, Mathematics, and Data Division serves as the lead principal investigator for the AMAIS project. “Scientific computing is often hampered by the lack of memory capacity, bandwidth, and speed of current devices,” said Márquez. “With the AMAIS project, we hope to overcome this ‘memory wall’ by developing a memory platform for new devices and improved organization and orchestration of existing ones.”

Researchers on the project use a hardware–software codesign strategy to improve the efficiency of AI applications on high-performance computing systems. The PNNL side of the collaboration primarily focuses on five main thrusts related to near-memory computing: high-level synthesis to enable the automatic generation of custom hardware accelerators, performance analysis to identify and ameliorate detrimental memory access patterns, software stack scale-up and scale-out strategies to exploit parallelism, and the development of application kernels that exercise these new technologies.

Márquez notes that this project could yield benefits across different research fields. “AI is becoming integral to many research areas,” said Márquez. “Everything from computational chemistry and molecular dynamics to modernization of the power grid would benefit from a memory boost.“

The AMAIS project is sponsored by the Advanced Scientific Computing Research program within the Department of Energy.