DocumentCode
3103030
Title
Toward PDN resource estimation: A law of general power density
Author
Jeong, Kwangok ; Kahng, Andrew B.
Author_Institution
ECE Dept., UC San Diego, La Jolla, CA, USA
fYear
2011
fDate
5-5 June 2011
Firstpage
1
Lastpage
6
Abstract
The power distribution network (PDN) is an increasingly significant consumer of on-chip interconnect resources. Thus, PDN estimation is increasingly central to system-level interconnect prediction for modern ICs. PDN design and verification require accurate power estimation and realistic current source distribution across a die. However, at early design stages, detailed placement or switching information is rarely available, so that designers either rely on pessimistic overdesign, which can lead to severe routing congestion, or encounter unexpected voltage noise problems at late design stages, which can lead to costly design iterations. In this work, we seek to identify a general trend for power density. From both empirical and analytical studies on random activity distributions, we propose a power law of activity density, which can potentially enable estimates of power density and voltage noise, as well as of required power distribution network (PDN) resources, in early design stages.
Keywords
distribution networks; estimation theory; integrated circuit design; integrated circuit interconnections; PDN resource estimation; activity density power law; analytical method; current source distribution; empirical method; general power density; on-chip interconnect resources; power distribution network design; random activity distribution; routing congestion; switching information; system-level interconnect prediction; voltage noise; Data models; Estimation; Logic gates; Mathematical model; Noise; Switches; System-on-a-chip;
fLanguage
English
Publisher
ieee
Conference_Titel
System Level Interconnect Prediction (SLIP), 2011 13th International Workshop on
Conference_Location
San Diego, CA
Print_ISBN
978-1-4577-1240-1
Type
conf
DOI
10.1109/SLIP.2011.6135432
Filename
6135432
Link To Document