Chameleon: versatile and practical near-DRAM acceleration architecture for large memory systems

Hadi Asghari-Moghaddam, Young Hoon Son, Jung Ho Ahn, Nam Sung Kim

International Symposium on Microarchitecture · 2016 · 80 citations · 51 references

Concepts

TL;DR

Memory bandwidth limits system performance, and near‑DRAM acceleration has been proposed but relies on expensive, capacity‑limited HBM/HMC devices, hindering scalability for large‑memory servers. The study demonstrates that load‑reduced DIMM (LRDIMM) data buffers are ideal sites for near‑DRAM accelerators. Chameleon is implemented without 3D/2.5D stacking, enabling seamless integration with large memory systems. Experiments show a Chameleon system achieves 2.13× higher geo‑mean performance and 34% lower data‑transfer energy than processor‑integrated accelerators.

Abstract

The performance of computer systems is often limited by the bandwidth of their memory channels, but further increasing the bandwidth is challenging under the stringent pin and power constraints of packages. To further increase performance under these constraints, various near-DRAM acceleration (NDA) architectures, which tightly integrate accelerators with DRAM devices using 3D/2.5D-stacking technology, have been proposed. However, they have not prevailed yet because they often rely on expensive HBM/HMC-like DRAM devices which also suffer from limited capacity, whereas the scalability of memory capacity is critical for some computing segments such as servers. In this paper, we first demonstrate that data buffers in a load-reduced DIMM (LRDIMM), which is originally developed to support large memory systems for servers, are supreme places to integrate near-DRAM accelerators. Second, we propose Chameleon, an NDA architecture that can be realized without relying on 3D/2.5D-stacking technology and seamlessly integrated with large memory systems for servers. Third, we explore three microarchitectures that abate constraints imposed by taking LRDIMM architecture for NDA. Our experiment demonstrates that a Chameleon-based system can offer 2.13 χ higher geo-mean performance while consuming 34% lower geo-mean data transfer energy than a system that integrates the same accelerator logic within the processor.

References

51