Heterogeneous Memory

Heterogeneous memory research focuses on optimizing the performance and energy efficiency of computing systems by leveraging different memory technologies with varying speed, capacity, and cost, such as DRAM and persistent memory. Current efforts concentrate on developing algorithms and architectures that intelligently manage data placement across this memory hierarchy, particularly for computationally intensive applications like recommendation systems and large language models. This work is crucial for improving the scalability and efficiency of machine learning, enabling faster training and inference, and reducing the energy consumption of data centers and edge devices.