PNNL and Micron Unveil Crete: A Revolutionary AI-Focused Computer with Massive Active Memory

Revolutionary AI-Focused Computer System Unveiled

The Department of Energy's Pacific Northwest National Laboratory (PNNL) and Micron have jointly developed a groundbreaking computer system called Crete, designed specifically for AI science applications. This innovative system, which came online on August 12, 2025, boasts an impressive 15 terabytes of active memory co-located with the system processors 1

1

2

Unique Memory Architecture

Crete's memory-to-processor configuration is unparalleled within the DOE laboratory complex and the high-performance computing community. The system utilizes Compute Express Link (CXL), an open industry standard, to connect memory with the processor via an I/O switch developed by XConn Technologies 1

1

Andrés Márquez, the AMAIS initiative lead and PNNL computer scientist, explained, "We are exploring novel memory technologies to address applications and problems that are constrained by memory shortcomings that manifest in the form of capacity, system bandwidth or sharing" 1

1

Innovative Memory-Processor Integration

Source: Tech Xplore

The Crete system cleverly combines two types of memory architectures:

1. Tightly coupled: Supported by Micron Registered Dual In-line Memory Module
2. Loosely coupled: Via custom Micron CXL controller boards

This unique configuration offers unprecedented opportunities for data-driven scientific computing, including AI applications 1

1

2

Massive Memory Capacity

To put Crete's memory capacity into perspective, its random-access memory is equivalent to 240 high-end laptops, each with 64 gigabytes of memory, running simultaneously and communicating in real-time 1

1

Advancing AI-Driven Research

Mark Helm, senior fellow of Micron's Pathfinding and Strategy team, emphasized the system's potential: "The Crete system represents a bold leap forward in redefining how memory and compute can work to unlock scientific discovery. We believe that the future of AI-driven research depends on architectures that break free from current memory constraints" 1

1

2

Related Stories

US National Labs Launch Massive AI Supercomputing Initiative to Maintain Global Technological Leadership

Phison aiDAPTIV and Intel tackle local AI's memory problem with 26B-parameter models on 16 GB RAM

SK hynix and TetraMem unveil in-memory computing chip that processes AI directly inside memory

Breakthrough in AI Memory Technology: ECRAM's Hidden Mechanisms Unveiled

Applications and Accessibility

Crete is now available to DOE Office of Science national laboratory users through the Advanced Memory to Support Artificial Intelligence for Science (AMAIS) initiative. The system is particularly suited for applications requiring more memory than typically available in current computing architectures 1

1

2

James "Jim" Ang, chief scientist for computing at PNNL, highlighted the system's immediate applications: "We have computational chemistry applications that are going to use this larger memory space right away. We're excited to find out what Crete can do with generative AI analyzing scientific data, as opposed to large language models and text" 1

1

2

Future Implications

The Crete testbed is expected to push the boundaries of CXL technology capabilities and shape the future of scientific AI. It aims to provide a launchpad for a new generation of algorithms that will accelerate discovery in chemistry, advanced materials, and molecular biology 1

1

2

Márquez concluded, "The memory mechanism enabled by the XConn memory switch allows memory that is sitting outside in the network to be easily accessed by multiple hosts at the same time, passing information without going to the hard drive for large workloads. That's a big differentiator. Nobody else is doing that" 1

1

2

As AI continues to evolve rapidly, Crete represents a significant step forward in integrating AI with simulation and data analysis workflows, potentially revolutionizing scientific research and discovery.