MD

Frugal Computing: Low Carbon and Sustainable Computing

  • The IPCC Sixth Assessment Report emphasizes that:

    • Greenhouse gas emissions must be drastically cut to keep global warming below 1.5°C.

      • This involves significant and immediate changes across all sectors, including energy, agriculture, transportation, and building.

      • The report provides detailed pathways for achieving these reductions, highlighting the need for rapid decarbonization of electricity supply, increased energy efficiency, and adoption of sustainable land management practices.

    • These cuts are needed immediately (2022).

      • Delaying action will make it more difficult and costly to achieve the 1.5°C target, potentially leading to irreversible climate changes.

      • Immediate action can spur innovation, create new jobs, and improve public health.

    • This is incompatible with unlimited economic growth.

      • The report suggests a shift towards sustainable development models that prioritize human well-being, equity, and environmental sustainability over traditional economic growth metrics.

      • This involves adopting circular economy principles, reducing consumption, and investing in green technologies.

The Need for Frugal Computing

  • Computational resources are finite.

    • The availability of raw materials, energy, and water for producing and operating computing devices is limited.

    • The environmental impact of e-waste and the disposal of obsolete hardware is a growing concern.

  • Since the 1970s, there has been increasing use of computational resources.

    • The proliferation of personal computers, mobile devices, and cloud computing has led to a dramatic increase in energy consumption and greenhouse gas emissions.

    • The demand for data storage, processing power, and network bandwidth continues to grow exponentially.

  • Growth of performance per Watt has been exponential (Koomey's Law).

    • Koomey's Law states that the energy efficiency of computation doubles approximately every 1.5 years.

    • This has allowed for significant improvements in computing performance without a corresponding increase in energy consumption.

  • Computational resources have been effectively treated as infinite until recently.

    • The focus on increasing computing power and reducing costs has often overlooked the environmental consequences of energy consumption and resource depletion.

    • The lack of awareness and concern about the environmental impact of computing has led to unsustainable practices.

  • Growth in demand cannot be offset by increased power efficiency.

    • Despite improvements in energy efficiency, the overall demand for computing resources is growing at a faster rate.

    • This means that total energy consumption and greenhouse gas emissions from computing continue to rise.

    • Moore and Koomey's laws can’t save us (“the free lunch is over”).

      • Moore's Law, which describes the doubling of transistors on a microchip every two years, is slowing down.

      • This means that improvements in computing performance are becoming more difficult and expensive to achieve.

    • The carbon footprint from computing will become a major contributor to the world total if business continues as usual.

      • If current trends continue, the energy consumption of computing devices could account for a significant portion of global greenhouse gas emissions by 2040.

  • The carbon footprint of device production is also significant.

    • The manufacturing of computers, smartphones, and other electronic devices requires large amounts of energy and resources.

    • The extraction of raw materials, the production of components, and the assembly of devices all contribute to greenhouse gas emissions.

    • Moore's Law has led to very short compute hardware lifetimes.

      • The rapid pace of technological innovation has led to a culture of planned obsolescence, where devices are designed to be replaced after a short period.

      • This results in a large amount of electronic waste and exacerbates the environmental impact of device production.

    • The current rate of obsolescence is unsustainable.

      • The environmental consequences of continuously producing and disposing of electronic devices are becoming increasingly severe.

      • A more sustainable approach is needed, one that emphasizes extending the lifespan of devices and reducing electronic waste.

The Imperative of Frugal Computing

  • Society needs to treat computational resources as finite and precious.

    • A shift in mindset is needed, one that recognizes the environmental impact of computing and prioritizes sustainability.

    • Consumers, businesses, and governments all have a role to play in promoting frugal computing practices.

  • Computing scientists, developers, and engineers need to ensure computing has the lowest possible energy consumption.

    • This involves designing energy-efficient algorithms, optimizing code for performance, and developing hardware with lower power requirements.

    • It also means considering the environmental impact of computing when making design decisions.

  • This must be achieved with currently available technologies since device lifetimes need to be extended dramatically.

    • While new technologies may offer some improvements in energy efficiency, it is important to focus on optimizing existing technologies and extending the lifespan of devices.

    • This can be achieved through software updates, hardware repairs, and promoting the reuse of devices.

Meeting Climate Targets

  • To limit global warming to below 1.5°C by 2040, a global reduction from 57 to 13 gigatonnes CO2 equivalent per year (GtCO2e/y) is needed.

    • This requires unprecedented and immediate action to reduce greenhouse gas emissions across all sectors.

    • The transition to a low-carbon economy will require significant investments in renewable energy, energy efficiency, and sustainable transportation.

    • This requires cutting emissions by ~10% per year.

      • Achieving this target will require a sustained and concerted effort to reduce emissions by 10% each year.

      • This will require significant changes in behavior, technology, and policy.

    • Currently, emissions are rising between 1%-2% a year.

      • Despite international agreements and efforts to reduce emissions, global emissions continue to rise.

      • This highlights the need for more ambitious and effective policies to address climate change.

  • Emissions from electricity are about 10 GtCO2e.

    • Electricity generation is a major source of greenhouse gas emissions, particularly in countries that rely on fossil fuels.

    • The transition to renewable energy sources, such as solar, wind, and hydropower, is essential for reducing emissions from electricity.

    • Electricity consumption is rising steeply.

      • As the global population grows and economies develop, the demand for electricity continues to increase.

      • This growth in demand must be met with clean energy sources to avoid further increases in greenhouse gas emissions.

    • Most electricity is still generated by burning fossil fuels.

      • Despite the growth of renewable energy, fossil fuels still account for a significant portion of global electricity generation.

      • The phase-out of fossil fuels and the transition to clean energy sources is a critical step in addressing climate change.

Limitations of Alternative Solutions

  • Renewables and nuclear energy won’t save us.

    • While renewable energy and nuclear power can play an important role in reducing emissions, they are not a silver bullet.

    • The deployment of these technologies is often slow and faces various challenges, such as high costs, land use conflicts, and public opposition.

    • Deployment is too slow.

      • The pace of deployment of renewable energy and nuclear power is not fast enough to meet the urgent need to reduce emissions.

      • This is due to a variety of factors, including technological barriers, regulatory hurdles, and financial constraints.

    • It takes 20 years to build a new nuclear power plant, and old ones are being shut down.

      • The construction of nuclear power plants is a lengthy and complex process, often taking decades to complete.

      • Many existing nuclear power plants are also being shut down due to safety concerns, economic considerations, and political pressure.

    • Renewables + nuclear will provide only 30% of electricity by 2040.

      • Even with significant investments in renewable energy and nuclear power, these sources are unlikely to provide more than 30% of global electricity by 2040.

      • This highlights the need for other solutions to reduce emissions from electricity.

  • Carbon Capture & Storage is also problematic.

    • Carbon capture and storage (CCS) is a technology that captures carbon dioxide emissions from industrial sources and stores them underground.

    • While CCS has the potential to reduce emissions, it is not a proven technology and faces various challenges, such as high costs, energy requirements, and storage capacity limitations.

    • The energy required in the capture process can be greater than the energy made available during the release of the CO2.

      • The capture process itself requires energy, which can offset some of the benefits of CCS.

      • In some cases, the energy required for capture can be greater than the energy produced by the facility, making the process energy-intensive.

    • Many scenarios assume large areas of land will be available, which may not be realistic or compatible with sustainability goals.

      • The deployment of CCS on a large scale would require significant amounts of land for storage and infrastructure.

      • This could lead to land use conflicts and may not be compatible with other sustainability goals, such as biodiversity conservation and food security.

    • There are poorly quantified risks of re-release and no credible standards or compliance procedures.

      • There are risks of carbon dioxide leaking from storage sites, which could negate the benefits of CCS.

      • There is a lack of clear standards and compliance procedures for CCS, which makes it difficult to ensure the safety and effectiveness of the technology.

    • Ethically, this may be seen as greenwashing.

      • Some critics argue that CCS is being used as a way to justify continued reliance on fossil fuels.

      • They argue that CCS is not a sustainable solution and that it distracts from the need to transition to renewable energy sources.

  • Carbon offsetting has limitations.

    • Carbon offsetting is a process of reducing emissions in one place to compensate for emissions in another place.

    • While carbon offsetting can play a role in reducing emissions, it is not a perfect solution and faces several limitations.

    • The earth’s land ecosystems can absorb 40 - 100 GtCO₂e from the atmosphere.

      • Land ecosystems, such as forests and soils, can absorb carbon dioxide from the atmosphere through photosynthesis.

      • This natural carbon sink helps to regulate the climate and reduce the concentration of greenhouse gases in the atmosphere.

    • Once this is achieved (takes decades), there is no capacity for additional carbon storage on land.

      • The capacity of land ecosystems to absorb carbon dioxide is limited.

      • Once the land ecosystems are saturated with carbon, they can no longer absorb additional carbon dioxide from the atmosphere.

    • The world emits 50 GtCO₂e/year, so offsetting can only handle at most 2 years' worth of emissions.

      • The amount of carbon dioxide emitted by human activities each year is far greater than the amount that can be absorbed by land ecosystems.

      • This means that carbon offsetting can only address a small portion of global emissions.

Reducing Emissions

  • The only way to reduce atmospheric CO2 to 1.5°C levels by 2040 is:

    • To reduce energy consumption.

      • Reducing energy consumption is essential for reducing greenhouse gas emissions.

      • This can be achieved through energy efficiency measures, such as using more efficient appliances, insulating homes, and driving less.

    • To reduce the amount of goods produced.

      • Reducing the amount of goods produced and consumed can also help to reduce emissions.

      • This can be achieved through sustainable consumption practices, such as buying less, reusing products, and recycling materials.

  • This is largely an economic problem, but technology has an important role to play.

    • While economic factors play a significant role in emissions, technology can play an important role in reducing emissions.

    • Technological innovations, such as renewable energy, energy storage, and carbon capture, can help to reduce emissions and transition to a low-carbon economy.

The Carbon Cost of Computing

  • In 2020, emissions from using computing were between 3.0% and 3.5% of the total.

    • The energy consumption of computing devices, such as computers, smartphones, and data centers, contributes to greenhouse gas emissions.

    • As the demand for computing resources continues to grow, the carbon footprint of computing is also increasing.

    • This is already more than the airline industry.

      • The carbon footprint of computing is already greater than that of the airline industry, which is a significant source of emissions.

      • This highlights the need to address the environmental impact of computing.

    • By 2040 this will grow to 14% (4x).

      • If current trends continue, the carbon footprint of computing could grow to 14% of global emissions by 2040.

      • This would make computing a major contributor to climate change.

    • By 2040, energy consumption of compute devices would be responsible for 5 gigatonnes of CO2.

      • The energy consumption of computing devices is projected to increase significantly by 2040.

      • This would result in a significant increase in greenhouse gas emissions.

  • Emissions from the production of computing devices exceed those incurred during operation.

    • The manufacturing of computers, smartphones, and other electronic devices requires large amounts of energy and resources.

    • The extraction of raw materials, the production of components, and the assembly of devices all contribute to greenhouse gas emissions.

    • Taking into account this carbon cost of production, computing would be responsible for 10 gigatonnes of CO2 by 2040.

      • When the emissions from both the operation and production of computing devices are taken into account, the total carbon footprint of computing is even greater.

      • This highlights the need to address the environmental impact of the entire lifecycle of computing devices.

    • This is almost 80% of the acceptable CO2 emissions budget of 13 gigatonnes of CO2.

      • The projected carbon footprint of computing by 2040 is a significant portion of the acceptable CO2 emissions budget for limiting global warming to 1.5°C.

      • This means that significant reductions in the carbon footprint of computing are needed to meet climate targets.

Emission Growth Drivers

  • High-Definition Video/VR/AR

    • The increasing demand for high-definition video, virtual reality (VR), and augmented reality (AR) is driving up emissions from computing.

    • These technologies require large amounts of data to be processed and transmitted, which consumes significant amounts of energy.

    • Growth in demand and in resolution.

      • The demand for high-definition video, VR, and AR is growing rapidly.

      • As the resolution of these technologies increases, the amount of data required to be processed and transmitted also increases.

    • VR/AR encode 3-D, so even higher bandwidth is needed.

      • VR and AR encode three-dimensional data, which requires even higher bandwidth than traditional video.

      • This further increases the energy consumption of these technologies.

  • Internet of Things (IoT)

    • The Internet of Things (IoT) is a network of interconnected devices that collect and exchange data.

    • The proliferation of IoT devices is driving up emissions from computing, as each device requires energy to operate and communicate.

    • Every small device has a huge network+cloud footprint.

      • Even small IoT devices can have a significant network and cloud footprint, as they require data storage, processing, and analysis.

      • The cumulative impact of millions or billions of IoT devices can be substantial.

    • Manufacturing of devices also has a large footprint.

      • The manufacturing of IoT devices also contributes to greenhouse gas emissions.

      • The extraction of raw materials, the production of components, and the assembly of devices all require energy and resources.

  • AI (particularly generative large language models)

    • Artificial intelligence (AI), particularly generative large language models, is driving up emissions from computing.

    • These models require large amounts of data to be trained and operated, which consumes significant amounts of energy.

    • Queries (e.g., ChatGPT) consume a huge amount of energy (40x more than conventional search).

      • Queries to generative large language models, such as ChatGPT, consume significantly more energy than conventional search queries.

      • This is due to the complexity of the models and the large amount of data they process.

    • Due to hype, growth is very steep.

      • The growth of AI is being driven by hype and unrealistic expectations.

      • This is leading to a rapid increase in energy consumption and greenhouse gas emissions.

    • Most governments have bought into this hype.

      • Many governments are investing heavily in AI, without fully considering the environmental consequences.

      • This could lead to a significant increase in emissions.

    • AI everywhere would lead to a very large increase in emissions.

      • The widespread adoption of AI could lead to a very large increase in energy consumption and greenhouse gas emissions.

      • This highlights the need for caution and careful planning in the development and deployment of AI.

  • Mobile devices

    • Mobile devices, such as smartphones and tablets, are a major source of emissions from computing.

    • The demand for mobile devices continues to grow, particularly in developing countries.

    • Still high growth in demand.

      • The demand for mobile devices is continuing to grow, driven by factors such as increasing affordability and the availability of new applications.

      • This growth in demand is leading to a corresponding increase in emissions.

    • The main driver is short replacement cycles.

      • The short replacement cycles of mobile devices are a major driver of emissions.

      • Consumers often replace their devices after only a few years, even if they are still functional.

    • Mobile devices have a huge network+cloud footprint and a large manufacturing footprint.

      • Mobile devices have a significant network and cloud footprint, as they require data storage, processing, and analysis.

      • The manufacturing of mobile devices also contributes to greenhouse gas emissions.

    • Growth of network infrastructure (5G, 6G, …).

      • The growth of network infrastructure, such as 5G and 6G, is also driving up emissions from computing.

      • These technologies require more energy to operate than previous generations of networks.

Lifecycle Emissions Breakdown

  • Lifecycle phases include:

    • Raw material extraction

      • The extraction of raw materials, such as minerals and metals, requires energy and can have significant environmental impacts.

      • Mining operations can contribute to deforestation, soil erosion, and water pollution.

    • Manufacturing

      • The manufacturing of computing devices requires energy and resources.

      • The production of components, such as microchips and displays, can be particularly energy-intensive.

    • Transport

      • The transport of raw materials, components, and finished products contributes to greenhouse gas emissions.

      • The use of ships, trucks, and airplanes for transportation can have a significant environmental impact.

    • Usage

      • The use of computing devices requires electricity.

      • The energy consumption of devices can vary depending on factors such as the type of device, the intensity of use, and the energy efficiency of the device.

    • Disposal (end-of-life)

      • The disposal of computing devices can have environmental impacts.

      • If devices are not properly recycled, they can end up in landfills, where they can release hazardous materials into the environment.

Manufacturing

  • Involves:

    • Mining (crushing rock, transportation)

      • Mining involves extracting raw materials from the earth.

      • This process can be energy-intensive and can have significant environmental impacts, such as deforestation, soil erosion, and water pollution.

    • Smelting

      • Smelting is a process of extracting metals from their ores.

      • This process requires high temperatures and can release pollutants into the air and water.

  • Chip production requires:

    • Electricity

      • Chip production requires a significant amount of electricity.

      • The energy consumption of chip production can vary depending on the complexity of the chip and the efficiency of the manufacturing process.

    • Raw materials (silicon wafers, gases, metals, …)

      • Chip production requires a variety of raw materials, including silicon wafers, gases, and metals.

      • The extraction and processing of these materials can have environmental impacts.

    • Production greenhouse gases (much worse than CO2)

      • The production of chips can release greenhouse gases that are much more potent than carbon dioxide.

      • These gases can contribute to climate change.

    • Water

      • Chip production requires a significant amount of water.

      • The water is used for cleaning, cooling, and other processes.

  • Similar processes for other components.

Usage

  • Involves:

    • Infrastructure (data center buildings etc.)

      • Data centers are large facilities that house computing equipment.

      • These facilities require energy for servers, cooling, and other equipment.

    • Electricity

      • The use of computing devices requires electricity.

      • The energy consumption of devices can vary depending on factors such as the type of device, the intensity of use, and the energy efficiency of the device.

    • Cooling overhead (requires energy)

      • Computing equipment generates heat, which must be removed to prevent damage.

      • Cooling systems require energy to operate.

    • Water (evaporated and lost to the local environment)

      • Some cooling systems use water, which is evaporated and lost to the local environment.

      • This can have a significant impact on water resources.

Disposal

  • Options include:

    • Recycling (requires transport, energy to disassemble, energy to produce new goods)

      • Recycling involves breaking down used devices and recovering valuable materials.

      • This process requires energy for transportation, disassembly, and the production of new goods.

    • Refurbishing (similar but less so)

      • Refurbishing involves repairing and restoring used devices to working condition.

      • This process requires less energy than recycling.

    • Landfill (transport, often to Africa; pollution; no chance for recycling)

      • Landfilling involves disposing of used devices in landfills.

      • This process can have significant environmental impacts, such as pollution and the release of hazardous materials.

Breakdown of Full System

  • Includes:

    • Data centers

    • Transmission networks

    • User media devices

    • Wired access networks

    • Core networks

    • Home terminals and routers

    • Cellular access networks

    • Cloud storage and encoding

    • Content Delivery Network

    • Subscriber premises

    • Internet transmission

    • Network transmission

    • Peripherals

    • Screens

Breakdown of Computing System

  • Includes:

    • End-user premises

    • Network (“internet”)

    • Cloud data centre