Abstract
Advancements in the field of chip fabrication led to the integration of a large number of transistors in a small area, giving rise to the multi–core processor era. Massive multi–core processors facilitate innovation and research in the field of healthcare, defense, entertainment, meteorology and many others. Reduction in chip area and increase in the number of on–chip cores is accompanied by power and temperature issues. In high performance multi–core chips, power and heat are predominant constraints. High performance massive multicore systems suffer from thermal hotspots, exacerbating the problem of reliability in deep submicron technologies. High power consumption not only increases the chip temperature but also jeopardizes the integrity of the system. Hence, there is a need to explore holistic power and thermal optimization and management strategies for massive on–chip multi–core environments.
In multi–core environments, the communication fabric plays a major role in deciding the efficiency of the system. In multi–core processor chips this communication infrastructure is predominantly a Network–on–Chip (NoC). Tradition NoC designs incorporate planar interconnects as a result these NoCs have long, multi–hop wireline links for data exchange. Due to the presence of multi–hop planar links such NoC architectures fall prey to high latency, significant power dissipation and temperature hotspots. Networks inspired from nature are envisioned as an enabling technology to achieve highly efficient and low power NoC designs. Adopting wireless technology in such architectures enhance their performance. Placement of wireless interconnects (WIs) alters the behavior of the network and hence a random deployment of WIs may not result in a thermally optimal solution. In such scenarios, the WIs being highly efficient would attract high traffic densities resulting in thermal hotspots. Hence, the location and utilization of the wireless links is a key factor in obtaining a thermal optimal highly efficient Network–on–chip.
Optimization of the NoC framework alone is incapable of addressing the effects due to the runtime dynamics of the system. Minimal paths solely optimized for performance in the network may lead to excessive utilization of certain NoC components leading to thermal hotspots. Hence, architectural innovation in conjunction with suitable power and thermal management strategies is the key for designing high performance and energy–efficient multicore systems. This work contributes at exploring various wired and wireless NoC architectures that achieve best trade–offs between temperature, performance and energy–efficiency. It further proposes an adaptive routing scheme which factors in the thermal profile of the chip. The proposed routing mechanism dynamically reacts to the thermal profile of the chip and takes measures to avoid thermal hotspots, achieving a thermally efficient dynamically reconfigurable network on chip architecture.
Library of Congress Subject Headings
Networks on a chip--Thermal properties; Interconnects (Integrated circuit technology)
Publication Date
2-2014
Document Type
Thesis
Student Type
Graduate
Degree Name
Computer Engineering (MS)
Department, Program, or Center
Computer Engineering (KGCOE)
Advisor
Amlan Ganguly
Advisor/Committee Member
Sonia Lopez Alarcon
Advisor/Committee Member
Gill Tsouri
Recommended Citation
Mhatre, Aniket Dilip, "Temperature Evaluation of NoC Architectures and Dynamically Reconfigurable NoC" (2014). Thesis. Rochester Institute of Technology. Accessed from
https://repository.rit.edu/theses/7662
Campus
RIT – Main Campus
Plan Codes
CMPE-MS
Comments
Physical copy available from RIT's Wallace Library at TK5105.546 .M43 2014