Summary

Benchmarking for sustainable computing involves the assessment and comparison of various metrics and criteria designed to improve sustainability performance within the tech industry. These benchmarks are critical for evaluating environmental, social, and governance (ESG) factors, offering both absolute and relative methods of assessment. Absolute benchmarking aligns a company's performance with fixed standards such as the Sustainable Development Goals (SDGs) or the Global Reporting Initiative (GRI) Standards, ensuring a standardized assessment. Relative benchmarking, by contrast, compares a company's performance against its industry peers, fostering competitive improvement and strategic insights into industry-specific sustainability efforts. The significance of sustainable computing benchmarks lies in their ability to provide actionable insights for companies looking to enhance their sustainability practices. Key metrics typically assessed include emissions, water usage, waste generation, diversity ratios, and pay equity, among others. This comprehensive approach helps organizations align with global sustainability standards and make informed decisions based on reliable data, often validated through third-party audits. Moreover, tools like the Green Metrics Tool (GMT) enable automated benchmarking, offering consistent and immediate reflections of changes in sustainability metrics, thus simplifying the tracking of progress over time. Notably, the technology industry has made significant strides in advancing sustainable practices through various initiatives and adherence to international standards. Collaborations with bodies such as the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC) have led to the development of benchmarks that guide the industry towards more sustainable operations. Companies like Intel have been proactive, establishing hubs for sustainable engineering and investing in green technologies to minimize their carbon footprints. Regulatory compliance and certifications such as ISO 14001 (Environmental Management) and ISO 50001 (Energy Management) further underscore the industry's commitment to sustainability. However, the journey toward sustainable computing is fraught with challenges. The inherent environmental impact of technology production, end-of-life processing, and the energy demands of computing infrastructure pose significant hurdles. Additionally, inconsistencies in ESG ratings due to varying methodologies and the substantial upfront investments required for green configurations in manufacturing and data centers add layers of complexity. Despite these challenges, the integration of innovative technologies like AI and the continued evolution of sustainability benchmarks hold promise for the future of green computing, offering pathways to reduce carbon footprints and enhance environmental stewardship in the tech industry.

Key Metrics and Criteria

Benchmarking for sustainable computing involves evaluating a variety of key metrics and criteria to measure and improve sustainability performance. These metrics can be broadly categorized into absolute and relative benchmarking methods.

Absolute Benchmarking

Absolute benchmarking compares a company’s ESG (Environmental, Social, and Governance) performance against fixed standards or sets of criteria such as the Sustainable Development Goals (SDGs) or the Global Reporting Initiative (GRI) Standards

[1]

. This approach provides a clear picture of a company’s ESG performance irrespective of its peers' performance. It ensures that companies align with universally recognized benchmarks, offering a more standardized assessment.

Relative Benchmarking

Relative benchmarking, on the other hand, compares a company’s ESG performance against that of its peers or competitors

[1]

. This method can help identify industry leaders and laggards, fostering competitive improvement. It is particularly useful for understanding how different companies within the same industry are performing and building strategies around ESG factors.

Conducting the Benchmarking Analysis

Once the data is collected, the next step involves comparing a company’s performance in the defined ESG factors with that of its peer group. Metrics such as ESG-related goals, material topics, philanthropic donations, and social metrics like diversity ratios and pay ratios can be directly compared. However, for environmental metrics such as emissions, water use, and waste generation, developing intensity ratios by dividing a given ESG metric by employee headcount, revenue, operational expenditure, or earnings can make the comparison more meaningful

[1]

.

Interpreting and Acting on Findings

Interpreting the benchmarking findings involves organizing data in line with leading standards such as the GRI, ISSB, ESRS, and the forthcoming SEC standards. Different topics have varying levels of importance for different companies, and relative benchmarking provides the flexibility needed to understand these nuances. The key takeaway is to focus on actual metrics like goals, emissions, water use, and diversity ratios rather than solely relying on ratings, which can be problematic when used in isolation

[1]

.

Issues with ESG Ratings

Although ESG ratings are widely used by investors and companies to evaluate sustainability performance, they have limitations that inhibit them from replacing comprehensive benchmarking. For instance, ratings can vary significantly depending on the methodology and criteria used by different rating agencies, leading to inconsistencies and potential misinterpretations

[1]

.

Visualization and Optimization

Visualizing data for impact is a crucial step in sustainable computing. For example, using services like AWS Cloud, companies can track, visualize, and optimize resource usage by focusing on impactful quick wins from best practices. This includes managing the lifecycle of datasets, scheduling build environments as needed, or migrating to managed services. Visualizations can help identify main contributors to resource consumption, track adoption rates of sustainable technologies over time, and establish sustainability proxy metrics for processes

[2]

.

Automated Benchmarking Tools

Tools like the Green Metrics Tool (GMT) offer automated benchmarking solutions. GMT is a free, open-source software that facilitates the consistent generation of sustainability metrics, enabling detailed tracking of software's energy efficiency over time. By using automated benchmarking, companies can replicate sustainability metrics consistently and reflect changes immediately

[3]

.

Importance of Audited Data

Third-party audits provide an additional layer of confidence in sustainability metrics, validating the accuracy of reported data and ensuring that metrics are based on reliable information. This enables stakeholders to make informed decisions based on trustworthy data

[4]

.

Metrics in Practice

In practice, metrics play a crucial role in green computing initiatives.

Industry Initiatives and Standards

The technology industry is increasingly prioritizing sustainability, reflected through a variety of initiatives and adherence to global standards. Companies collaborate with international standards bodies such as the International Organization for Standardization (ISO), International Electrotechnical Commission (IEC), European Telecommunications Standards Institute (ETSI), and CEN-CENELEC to develop comprehensive standards aimed at fostering innovation, maximizing sustainability outcomes, and enabling a thriving international ecosystem

[5]

. These collaborations help in setting benchmarks that guide the industry towards more sustainable practices.

Advancing Industry Standards

Many tech companies are actively engaging in alliances to advance industry standards and methodologies, particularly focusing on lowering emissions in the semiconductor and ICT industries. Intel, for instance, has established a hub for sustainable engineering in India, which brings together leaders to solve complex sustainability problems. This lab aims to reduce the carbon footprint across the PC life cycle by innovating in areas such as motherboard footprint reduction, conflict-free mineral sourcing, thermal and chassis technologies, and improving end-product recyclability

[5]

.

Regulatory Compliance and Certifications

Regulatory bodies worldwide are increasingly focusing on sustainability, enforcing guidelines and standards to ensure corporate accountability. Companies must stay updated on global sustainability standards and regulations to comply with reporting requirements, ensuring transparency and accountability in their sustainability practices. Certifications such as ISO 14001 (Environmental Management), ISO 50001 (Energy Management), and B Corp certification are crucial as they highlight a company's commitment to sustainability, adding credibility and demonstrating adherence to international standards

[4]

.

Sustainable Computing Investments

Intel Capital, the investment arm of Intel, strategically funds companies and initiatives that drive innovation in sustainability. These investments aim to accelerate sustainable computing by reimagining the design, sourcing, production, and delivery of products and services

[5]

. In addition, Intel's green bond initiative has allocated $425 million to support investments in sustainable operations, focusing on pollution prevention, water stewardship, energy efficiency, renewable energy, and waste management[5]

.

Ecosystem Collaboration and Green Coding

Collaboration with ecosystem partners is essential to creating more sustainable computing solutions. Intel collaborates with hundreds of partners to apply ICT learnings in reducing environmental footprints across various industries

[5]

. Moreover, the practice of green coding has emerged as a key segment of green computing, aiming to minimize the energy consumption involved in processing lines of code. This practice helps organizations reduce their overall energy consumption, supporting broader sustainability goals[6]

[7]

.

Popular Benchmarks

Benchmarks are widely used to assess and qualify sustainability standards, certifications, corporate policies, or other initiatives

[8]

. Sustainability and ESG benchmarks can be employed across all industries, serving various purposes from public procurement authorization to corporate performance improvement[9]

. These benchmarks enable comparability of relative scope, coverage, rigor, and outcomes by assessing multiple sustainability standards (VSS), policies, tools, or company performance against fixed reference points[8]

.

SPECpower

SPECpower is the first industry-standard benchmark designed to measure power consumption in relation to performance for server-class computers

[10]

. Developed by the Standard Performance Evaluation Corporation (SPEC), this benchmark aims to improve the power and performance characteristics of single- and multi-node servers[11]

. It is part of a broader effort to create benchmarks that encourage sustainable computing practices across the IT industry[11]

.

Green500

The Green500 list ranks supercomputers by their energy efficiency, complementing the TOP500 list, which traditionally ranks them based on brute force computing power

[11]

. Launched in 2007, the Green500 promotes sustainable supercomputing by highlighting the performance-per-watt metric, with the aim of encouraging the development of more energy-efficient computing systems[10]

[11]

. For instance, in 2019, two Japanese supercomputers led the Green500 ranking with performances exceeding 16 GFLOPS/watt[10]

. The Green500 list is updated biannually, providing a window into the challenges and advancements in building supercomputers that prioritize efficiency and low energy consumption[12]

.

SCI Metric

The Software Carbon Intensity (SCI) metric evaluates the carbon cost of software use, providing insights into the environmental impact of various software applications

[3]

. For example, in the case of Wagtail, an open-source content management system (CMS), the SCI metric measures the carbon cost per webpage viewed, with a calculated value of approximately 0.02 gCO2e per page request[3]

. This metric helps developers track the sustainability of their software over time and make informed decisions to minimize its environmental impact[3]

.

ICT Energy Indices

In the telecommunications sector, benchmarks focus on the energy consumption of network devices and their environmental impact

[10]

. Studies in this area aim to optimize device and network consumption, providing key energy indices for relevant comparisons between different network elements[10]

. These benchmarks are crucial as the energy consumption of information and communication technologies (ICTs) continues to be significant compared to other industries[10]

. By utilizing these benchmarks, stakeholders can gain standardized, meaningful comparisons across various entities and initiatives, thereby fostering a competitive environment that encourages continuous improvement in sustainability practices[9]

.

Methodologies

The methodologies used in benchmarking for sustainable computing involve a combination of various techniques aimed at assessing environmental, social, and governance (ESG) factors. These methodologies are crucial for understanding the impact of computing practices and guiding improvements in sustainability.

Life Cycle Assessment (LCA)

Life Cycle Assessment (LCA) is a critical methodology in sustainable computing benchmarking. LCA is used to assess the environmental impacts associated with all stages of a product's life cycle, from raw material extraction through production, use, and disposal or recycling

[13]

. The ISO LCA standards (ISO 14040 and ISO 14044) provide guidelines on conducting these assessments.

• Goal and Scope Definition: The goal of the study is explicitly stated, and the scope is outlined, detailing the qualitative and quantitative information included. The scope often requires multiple pages to describe the detail and depth of the study, ensuring that the goal can be achieved within the stated limitations[13]

• .

• Inventory Analysis: This phase involves compiling an inventory of energy and material inputs and environmental releases. The data collection process is rigorous, often requiring primary data sources such as on-site measurements and questionnaires to collect detailed information about each process within the system boundary[13]

• .

• Impact Assessment: Evaluating the potential environmental impacts associated with the identified inputs and releases. This phase interprets the inventory data to assess how material flows affect the environment, which helps in making informed decisions[13]

• .

• Interpretation: The results are interpreted to provide a sound basis for informed decisions, improving processes, supporting policy, and selecting products or processes that result in the least environmental impact[13]

• .

Big Data Analytics (BDA)

Big Data Analytics (BDA) techniques and tools are increasingly being used to promote the sustainable development of manufacturing enterprises. BDA helps in assessing the stage of product or service life cycles during manufacturing activities and operations

[14]

. The integration of BDA with LCA methodologies (referred to as LCSA - Life Cycle Sustainability Assessment) allows for a more comprehensive evaluation of sustainability performance. This combined approach leverages large datasets to provide insights into various impact factors such as human health, crop production capacity, and biodiversity[14]

.

Software Life Cycle Assessment

A specialized methodology for software products is the Software Life Cycle Assessment (SLCA). This approach addresses the unique characteristics of software, which are often considered intangible. The SLCA aims to offer a tailored methodology to conduct LCA for software products, acknowledging their specific features different from tangible products

[15]

.

• Defining Intangibility: Understanding the unique aspects of software products that differentiate them from physical products[15]

• .

• Impact Measurement: Assessing the environmental and potentially social impacts of software products, despite their intangible nature[15]

• .

Benchmarking Analysis

Benchmarking is an essential step in evaluating a company's performance against peers in ESG factors. This process involves collecting data on various metrics such as ESG-related goals, material topics, philanthropic donations, and social metrics like diversity ratios and pay equity

[1]

. The collected data is then used to compare the company's performance with that of its peer group, providing insights into areas of strength and opportunities for improvement[1]

.

Case Studies

Recently, governments and organizations have repeatedly pressed manufacturing enterprises to promote the ethical and transparent use of natural resources, lessen their negative effects on national and international ecosystems, and safeguard people and the environment

[14]

. In this context, enhancing the various stages of the product/service life cycle to fulfill sustainability requirements and foster sustainable value creation is a key area of interest for researchers and professionals[14]

. One example of research in this area is the application of life cycle sustainability assessment (LCSA) in manufacturing enterprises to incorporate environmental and social criteria into their product/service life cycle strategies[14]

. The descriptive and content analyses of various case studies have provided key findings related to the main goals, divergences in the scope of sustainability pillars, data collection sources, and applicable impact indicators for environmental, economic, and social sustainability[16]

. In the context of the Industry 4.0 technologies, Jamwal et al. conducted a systematic literature review study to discover and evaluate the evolution of contributions towards manufacturing enterprises' sustainability. They concluded that most research in this field has been theoretical and conceptual, with only a few studies examining real applications of different technologies to achieve sustainability[14]

. Moreover, case studies often reveal the influence of parameters and available data on results, highlighting the necessity of sensitivity analyses to identify significant impact factors and uncertainties[13]

. Data sources used in LCAs are typically large databases, including HESTIA (University of Oxford), soca, EuGeos' 15804-IA, NEEDS, ecoinvent, PSILCA, and others[13]

. Collaborative efforts have also been observed in initiatives like the global Intel Sustainability Summit, which convened more than 140 organizations to unite efforts in reducing environmental impacts and transitioning to sustainable practices. The summit highlighted the importance of standardizing carbon footprint methodologies and developing net-zero roadmaps, underscoring the potential impact of the value chain[17]

. Lastly, companies have demonstrated success in implementing sustainable practices across their supply chains. Examples include supporting smallholder farmers in adopting sustainable agriculture practices, resulting in increased crop yields, reduced water usage, and preservation of biodiversity[18]

. Integration of sustainable practices across the supply chain involves assessing and improving environmental impacts at every stage, from raw material sourcing to product disposal[18]

.

Challenges and Limitations

The pursuit of sustainable computing faces several challenges and limitations that need to be addressed to make meaningful progress. One significant challenge is the tension between designing hardware for better performance and designing it in a sustainable manner. While technology scaling has historically driven optimizations in computing, the diminishing returns of Moore's Law necessitate rethinking traditional architectural techniques through the lens of sustainability and reliability. This dual focus on reducing both capital expenditure (capex) and operational expenditure (opex) costs is crucial for future sustainable computing efforts

[19]

. Another major challenge is the significant environmental footprint of technology itself. This includes the impact of hardware manufacturing, end-of-life processing, and the energy required to power end-user computing. Although technology is a key enabler of sustainability solutions, such as AI predicting flood patterns and software pinpointing greenhouse gas (GHG) hotspots in supply chains, these benefits are offset by the environmental costs of the technology used[20]

. Additionally, there is often a lack of concern or awareness regarding the IT industry's role in climate change. This apathy is compounded by the market's focus on developing smaller and faster components rather than environmentally friendly ones. The rapid evolution of technology also presents challenges in extending product lifecycles and ensuring that each iteration meets eco-friendly standards. Switching to green configurations in factories, data centers, or corporate offices often requires substantial upfront capital investment, further acting as a barrier to sustainable practices[11]

. Data quality and availability are critical for conducting effective Life Cycle Assessments (LCA), yet they pose another layer of complexity. Ensuring the precision, completeness, and representativeness of data is essential, and gaps in data can lead to uncertainties in the results. Efforts are ongoing at both international and European levels to improve the accessibility and availability of LCA data through databases like Ecoinvent8, which enable comprehensive sustainability assessments[21]

. Moreover, the data collection process itself can be cumbersome, requiring both quantitative and qualitative data from various stakeholders within a company. While some data, such as electricity or water bills, are readily available, other necessary data may not be as easily accessible, necessitating the use of secondary data from industry averages or other sources. The need for meticulous data collection and the associated complexities further complicate the implementation of LCA[22]

.

Future Directions

Achieving truly sustainable computing requires a holistic, lifecycle-oriented approach that considers the complex tradeoffs and interdependencies between design, manufacturing, and operations

[23]

. The development of new tools, methodologies, and best practices for modeling, analysis, and optimization across the entire compute lifecycle is essential for significant improvements in the carbon efficiency and environmental sustainability of computing systems[23]

. One promising direction involves the potential for recycling older high-end hardware into low-end markets, such as transforming current server chips into tomorrow's IoT compute by managing power utilization and building chips in a reconfigurable way[19]

. However, this concept will require significant support from larger corporations to fully take off. Identifying improvement opportunities for future products, as well as reducing their environmental impacts, is another critical focus[15]

. Environmental impact comparisons between products or software solutions can guide users and developers in making more sustainable choices, although these comparisons must be meticulously executed to ensure reliability[15]

. Moreover, as norms like Electrical and Electronic Equipment (EEE), RoHS, WEEE, REACH, or ErP evolve, companies that have preemptively addressed eco-design issues will gain a competitive advantage[15]

. The integration of generative AI and virtual spaces into sustainable practices also presents significant potential. However, these technologies have high carbon footprints that must be drastically reduced to meet climate goals[24]

. Innovations across the technology stack, combined with close collaborations with industry and policymakers, are necessary to realize these reductions[24]

. Cross-disciplinary projects like Carbon Connect aim to establish new standards for carbon accounting in the computing industry, influencing future energy policy and legislation[24]

. Technology's role as a key enabler of sustainability solutions is well-recognized, with the ability to reduce greenhouse gas emissions and protect living systems through advanced software and AI capabilities[20]

. Despite this, technology itself has a significant environmental footprint, necessitating careful consideration of its entire lifecycle[20]

. Furthermore, the European Union's commitment to digitalization and ICT solutions underscores the potential for these technologies to drive a greener economy. The green transition, supported by digital solutions, offers new opportunities for innovation and competitiveness in the ICT sector[10]

. Digital solutions can advance the circular economy, decarbonize sectors, and reduce the environmental and social footprint of products, aligning with the ambitions of the European Green Deal and Sustainable Development Goals[10]

. In the realm of AI, future innovations are expected to include new architectures such as neuromorphic computing, which, despite their high energy demands, offer significant potential for solving large-scale problems[25]

. Mitigating this energy consumption involves optimizing AI models, software, and hardware, focusing on data quality, and employing carbon-aware software[25]

. As AI continues to evolve, new computing approaches will be necessary to balance its capabilities with sustainability requirements[25]