Skip to main navigation Skip to search Skip to main content

A DRAM-based Near-Memory Architecture for Accelerated and Energy-Efficient Execution of Transformers

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Transformers-based language models have achieved remarkable accuracy in various NLP tasks, employing self-attention mechanisms primarily based on matrix multiplication. However, their significant size leads to data movement issues, causing latency and energy efficiency challenges in conventional Von-Neumann systems. To mitigate these issues, several in-memory and near-memory architectures have been proposed. This paper introduces PACT-3D, a near-memory architecture featuring novel computing units integrated with DRAM banks. PACT-3D significantly reduces latency by 1.7 × and improves energy efficiency by 18.7 × compared to state-of-the-art near-memory architectures.

Original languageEnglish (US)
Title of host publicationGLSVLSI 2024 - Proceedings of the Great Lakes Symposium on VLSI 2024
PublisherAssociation for Computing Machinery
Pages57-62
Number of pages6
ISBN (Electronic)9798400706059
DOIs
StatePublished - Jun 12 2024
Event34th Great Lakes Symposium on VLSI 2024, GLSVLSI 2024 - Clearwater, United States
Duration: Jun 12 2024 → Jun 14 2024

Publication series

NameProceedings of the ACM Great Lakes Symposium on VLSI, GLSVLSI

Conference

Conference34th Great Lakes Symposium on VLSI 2024, GLSVLSI 2024
Country/TerritoryUnited States
CityClearwater
Period6/12/24 → 6/14/24

Keywords

  • DRAM
  • Energy Efficiency
  • In/Near-memory Processing
  • LLMs
  • Memory Wall
  • Transformers

ASJC Scopus subject areas

  • General Engineering

Fingerprint

Dive into the research topics of 'A DRAM-based Near-Memory Architecture for Accelerated and Energy-Efficient Execution of Transformers'. Together they form a unique fingerprint.

Cite this