Investigation of performance metrics for interconnect stack architectures

Authors:
Puneet Gupta;Andrew B. Kahng;Youngmin Kim;Dennis Sylvester
Affiliations:
University of California at San Diego, San Diego, CA;University of California at San Diego, San Diego, CA;University of Michigan at Ann Arbor, Ann Arbor, MI;University of Michigan at Ann Arbor, Ann Arbor, MI
Venue:
Proceedings of the 2004 international workshop on System level interconnect prediction
Year:
2004

Citing 6
Cited 2

A compact physical via blockage model

IEEE Transactions on Very Large Scale Integration (VLSI) Systems - Special issue on system-level interconnect prediction
Optimal n-tier multilevel interconnect architectures for gigascale integration (GSI)

IEEE Transactions on Very Large Scale Integration (VLSI) Systems - System Level Design
Throughput-driven IC communication fabric synthesis

Proceedings of the 2002 IEEE/ACM international conference on Computer-aided design
Maximizing throughput over parallel wire structures in the deep submicrometer regime

IEEE Transactions on Very Large Scale Integration (VLSI) Systems
A Novel Metric for Interconnect Architecture Performance

DATE '03 Proceedings of the conference on Design, Automation and Test in Europe - Volume 1
Interconnect opportunities for gigascale integration

IBM Journal of Research and Development

Package level interconnect options

Proceedings of the 2005 international workshop on System level interconnect prediction
Wafer-level package interconnect options

IEEE Transactions on Very Large Scale Integration (VLSI) Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper discusses metrics involving the bandwidth and energy characteristics of arbitrary interconnect stacks. Front-end dimensions are set by lithography and related fabrication restrictions and its performance is easily quantified using well-known metrics such as FO4 or ring oscillator delays, Ioff, and Ion. Back-end dimensions are not similarly constrained yet there are no comparable back-end metrics. In this study we seek figures-of-merit for interconnect architectures (stacks) that describe performance in terms of bandwidth and energy while considering issues such as via blockage and repeaters. A definition of bandwidth is presented and then appropriate via blockage models for interconnect stacks are investigated. In this paper, we improve existing bandwidth and throughput-driven design methodologies by looking at the entire stack rather than a single wiring layer. We also propose the use of bandwidth per unit energy. We evaluate and discuss these metrics in current 130nm and 90nm interconnect technologies.