Abstract

As semiconductor technologies continues to scale, more and more cores are being integrated on the same multicore chip. This increase in complexity poses the challenge of efficient data transfer between these cores. Several on-chip network architectures are proposed to improve the design flexibility and communication efficiency of such multicore chips. However, in a larger system consisting of several multicore chips across a board or in a System-in-Package (SiP), the performance is limited by the communication among and within these chips. Such systems, most commonly found within computing modules in typical data center nodes or server racks, are in dire need of an efficient interconnection architecture.

Conventional interchip communication using wireline links involve routing the data from the internal cores to the peripheral I/O ports, travelling over the interchip channels to the destination chip, and finally getting routed from the I/O to the internal cores there. This multihop communication increases latency and energy consumption while decreasing data bandwidth in a multichip system. Furthermore, the intrachip and interchip communication architectures are separately designed to maximize design flexibility. Jointly designing them could, however, improve the communication efficiency significantly and yield better solutions.

Previous attempts at this include an all-photonic approach that provides a unified inter/intra-chip optical network, based on recent progress in nano-photonic technologies. Works on wireless inter-chip interconnects successfully yielded better results than their wired counterparts, but their scopes were limited to establishing a single wireless connection between two chips rather than a communication architecture for a system as a whole.

In this thesis, the design of a seamless hybrid wired and wireless interconnection network for multichip systems in a package is proposed. The design utilizes on-chip wireless transceivers with dimensions spanning up to tens of centimeters. It manages to seamlessly bind both intrachip and interchip communication architectures and enables direct chip-to-chip communication between the internal cores. It is shown through cycle accurate simulations that the proposed design increases the bandwidth and reduces the energy consumption when compared to the state-of-the-art wireline I/O based multichip communications.

Library of Congress Subject Headings

Networks on a chip--Design and construction; Wireless communication systems--Computer simulation; Interconnects (Integrated circuit technology)

Publication Date

6-2015

Document Type

Thesis

Student Type

Graduate

Degree Name

Computer Engineering (MS)

Department, Program, or Center

Computer Engineering (KGCOE)

Advisor

Amlan Ganguly

Advisor/Committee Member

Andres Kwasinski

Advisor/Committee Member

Reza Azarderakhsh

Comments

Physical copy available from RIT's Wallace Library at TK5105.546 .M87 2015

Campus

RIT – Main Campus

Plan Codes

CMPE-MS

Download

Share

COinS