Low-swing signaling for energy efficient on-chip networks

On-chip networks have emerged as a scalable and high-bandwidth communication fabric in many-core processor chips. However, the energy consumption of these networks is becoming comparable to that of computation cores, making further scaling of core counts difficult. This thesis makes several contributions to low-swing signaling circuit design for the energy efficient on-chip networks in two separate projects: on-chip networks optimized for one-to-many multicasts and broadcasts, and link designs that allow on-chip networks to approach an ideal interconnection fabric. A low-swing crossbar switch, which is based on tri-state Reduced-Swing Drivers (RSDs), is presented for the first project. Measurement results of its test chip fabricated in 45nm SOI CMOS show that the tri-state RSD-based crossbar enables 55% power savings as compared to an equivalent full-swing crossbar and link. Also, the measurement results show that the proposed crossbar allows the broadcast-optimized on-chip networks using a single pipeline stage for physical data transmission to operate at 21% higher data rate, when compared with the full-swing networks. For the second project, two clockless low-swing repeaters, a Self-Resetting Logic Repeater (SRLR) and a Voltage-Locked Repeater (VLR), have been proposed and analyzed in simulation only. They both require no reference clock, differential signaling, and bias current. Such digital-intensive properties enable them to approach energy and delay performance of a point-to-point interconnect of variable lengths. Simulated in 45nm SOI CMOS, the 10mm SRLR featured with high energy efficiency consumes 338fJ/b at 5.4Gb/s/ch while the 10mm VLR raises its data rate up to 16.OGb/s/ch with 427fJ/b.

Department dc:contributor.department

Massachusetts Institute of Technology. Dept. of Electrical Engineering and Computer Science.

Grantor dc:publisher

Massachusetts Institute of Technology

Year dc:date.issued

2011

Author dc:creator

• Park, Sunghyun, Ph. D. Massachusetts Institute of Technology

Advisor dc:contributor.advisor

• Li-Shiuan Peh and Anantha P. Chandrakasan.

• Electrical Engineering and Computer Science.

Statement dc:rights

• M.I.T. theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. See provided URL for inquiries about permission.

Licence dc:rights.uri

• http://dspace.mit.edu/handle/1721.1/7582

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/1721.1/66474

OAI identifier oai:identifier

oai:dspace.mit.edu:1721.1/66474