Managed-Retention Memory: A New Class of Memory for the AI Era

TL;DR

This paper introduces Managed-Retention Memory (MRM), a novel memory class optimized for AI inference workloads, addressing limitations of High Bandwidth Memory (HBM) by balancing density, read bandwidth, and energy efficiency.

Contribution

The paper proposes MRM as a new memory solution tailored for AI workloads, leveraging workload-specific trade-offs to improve performance and cost-effectiveness over traditional memory types.

Findings

MRM offers improved read bandwidth and energy efficiency for AI inference.

MRM demonstrates potential to utilize SCM technologies effectively.

Analysis shows MRM can outperform HBM in key AI workload metrics.

Abstract

AI clusters today are one of the major uses of High Bandwidth Memory (HBM). However, HBM is suboptimal for AI workloads for several reasons. Analysis shows HBM is overprovisioned on write performance, but underprovisioned on density and read bandwidth, and also has significant energy per bit overheads. It is also expensive, with lower yield than DRAM due to manufacturing complexity. We propose a new memory class: Managed-Retention Memory (MRM), which is more optimized to store key data structures for AI inference workloads. We believe that MRM may finally provide a path to viability for technologies that were originally proposed to support Storage Class Memory (SCM). These technologies traditionally offered long-term persistence (10+ years) but provided poor IO performance and/or endurance. MRM makes different trade-offs, and by understanding the workload IO patterns, MRM foregoes long-term data retention and write performance for better potential performance on the metrics important for these workloads.