Author ORCID Identifier

https://orcid.org/0009-0003-4022-6901

Date Available

12-12-2025

Year of Publication

2025

Document Type

Master's Thesis

Degree Name

Master of Electrical Engineering (MEE)

College

Engineering

Department/School/Program

Electrical and Computer Engineering

Faculty

Ishan G. Thakkar

Faculty

Daniel Lau

Abstract

The use of computing technologies has significantly enhanced several aspects of our day-to-day lives. But it has still revealed significant environmental concerns, primarily related to greenhouse gas emissions and energy consumption. Initially, the primary environmental problems associated with computing were energy consumption during device operation. However, with the rapid advancement of technology and increasing computational demands, attention has shifted towards the embodied carbon footprint. This term refers to the total greenhouse gas emissions throughout a product’s lifecycle from the extraction of raw materials to end-of-life processing. It has become increasingly significant in the context of manufacturing integrated circuits (ICs), such as processors and memory devices, which are essential components of modern computing systems [1].

Conventional environmental impact assessments of computing hardware rely on generalized industry data [2]. These approaches may inadequately capture the specific ecological burdens associated with producing specialized high-performance components, highlighting the urgent need for more precise analytical tools [3]. For instance, the Architectural Carbon Footprint Modeling Tool (ACT) uses publicly available data to enhance its functionality and enable more accurate carbon emission estimates, enabling designers to make environmentally conscious choices alongside traditional performance metrics [2]. The ACT’s ability for granular lifecycle analysis represents a vital advancement in sustainability modeling, delivering far greater detail than conventional assessment methods [2].

Concurrently exploring innovative architectures, such as Processing-In-Memory (PIM), offers significant opportunities to advance sustainable computing fundamentally. PIM architectures incorporate processing capabilities directly into memory units, reducing the energy-intensive data movement characteristic of traditional GPU-based systems. Studies indicate that PIM can result in lower operational energy consumption, particularly during low-utilization tasks, highlighting its potential to decrease the carbon footprint compared to conventional architectures substantially [4]. A thorough understanding of the environmental impacts of PIM-based computing hardware requires a comprehensive examination of all lifecycle stages, focusing on both embodied and operational energy, as well as how regional energy grid carbon intensity affects manufacturing emissions [5].

To advance sustainable computing, methods that minimize embodied energy, such as optimizing hardware reuse, adopting modular designs, and implementing eco-conscious manufacturing practices, are now gaining increasing importance [6]. Additionally, analytical frameworks such as break-even and indifference analysis are valuable tools for assessing the economic and environmental viability of various hardware options, as they allow comparisons of their embodied and operational energy characteristics [3]. The complexities of component aging, especially concerning DRAM, and its impacts on operational latency and energy consumption further complicate long-term sustainability evaluations of computing systems [7]. Ultimately, this dissertation seeks to enhance our understanding of the environmental sustainability of computing hardware, particularly by comparing the lifecycle carbon footprint of PIM architectures to that of traditional GPUs, while also considering crucial factors such as regional energy dependencies and the effects of technological aging [6].

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2025.552

Download

Share

COinS