PNNL

The research of the Advanced Memory to Support Artificial Intelligence for Science (AMAIS) project, sponsored by the Advanced Scientific Computing Research (ASCR) program in the Department of Energy (DOE), focuses on new memory systems applicable to scientific computational research. These new memory systems are needed to enable the convergence of first-principles scientific modeling/simulation with AI data driven science. The resulting requirements increase demand on memory capacity to satisfy the large volume processed by large scale AI boosted simulations. In addition, the expanded memory capacity requirements should not negatively impact memory bandwidth and latency to the point that high performance computing for memory-bound algorithms becomes impractical. 

AMAIS is giving special consideration to Fabric Attached Memory (FAM), a system architecture that allows large data sets to be stored in memory that is accessible to all compute nodes in a cluster. The architecture is well suited to handle large data sets that are too big to fit into a single node's local memory and that cannot be easily partitioned without incurring substantial computational costs. 

Therefore, resulting memory architectures introduce additional tiers in the memory hierarchy that require revisiting architectural models and refactoring algorithmic kernels, performance analysis, compilers and runtime systems and cybersecurity.  

Related Projects

TEC⁴

The Transferring Exascale Computational Chemistry to Cloud Computing Environment and Emerging Hardware Technologies (TEC⁴) project’s goal is to accelerate the transition from the basic research associated with developing and implementing electronic structure methods to their widespread use in solving complex challenges in industrial chemical sciences.