Toward Sustainable HPC: Carbon Footprint Estimation and Environmental Implications of HPC Systems

Baolin Li,Rohan Basu Roy,Daniel Wang,Siddharth Samsi,Vijay Gadepally,Devesh Tiwari

DOI: https://doi.org/10.1145/3581784.3607035

2023-11-19

Abstract:The rapid growth in demand for HPC systems has led to a rise in carbon footprint, which requires urgent intervention. In this work, we present a comprehensive analysis of the carbon footprint of high-performance computing (HPC) systems, considering the carbon footprint during both the hardware manufacturing and system operational stages. Our work employs HPC hardware component carbon footprint modeling, regional carbon intensity analysis, and experimental characterization of the system life cycle to highlight the importance of quantifying the carbon footprint of HPC systems.

Distributed, Parallel, and Cluster Computing

What problem does this paper attempt to address?

The paper attempts to address the issue of the carbon footprint associated with the rapid growth of High-Performance Computing (HPC) systems. Specifically, it focuses on the carbon footprint of HPC systems, including carbon emissions during the hardware manufacturing phase and the system operation phase. By comprehensively analyzing the carbon footprint models of HPC hardware components, regional carbon intensity analysis, and experimental characterization of the system lifecycle, the authors aim to quantify the carbon footprint of HPC systems and propose methods to reduce their environmental impact.

### Main Issues
1. **How to comprehensively quantify the carbon footprint of HPC systems?**
- The paper attempts to comprehensively quantify the carbon footprint of HPC systems from both the hardware manufacturing and system operation phases.

2. **What are the differences in the carbon footprint of different types of HPC components (such as GPU, CPU, memory, and storage devices)?**
- Through modeling and analysis, the paper explores the carbon footprint of different types of HPC components and their contribution to the overall carbon footprint.

3. **What is the impact of regional carbon intensity on the carbon emissions during the operation phase of HPC systems?**
- The paper analyzes the impact of carbon intensity in different regions on the carbon emissions during the operation phase of HPC systems, emphasizing the importance of cross-regional HPC systems.

4. **What is the impact of increasing the number of GPUs on system performance and carbon footprint?**
- Through experimental research, the paper explores the impact of increasing the number of GPUs on system performance and carbon footprint, finding that as the number of GPUs increases, performance improvement gradually saturates, but the carbon footprint continues to increase.

5. **What are the changes in the contribution of different components to the overall carbon footprint in leading supercomputers?**
- The paper compares the carbon footprint contributions of different components in three leading supercomputers (Frontier, LUMI, and Perlmutter), identifying the most significant sources of carbon footprint.

### Solutions
- **Modeling and Analysis**: The paper adopts methods such as carbon footprint modeling of HPC hardware components, regional carbon intensity analysis, and experimental characterization of the system lifecycle.
- **Data Collection**: By collecting publicly available product datasheets and sustainability reports, the carbon footprint data of different components is obtained.
- **Experimental Validation**: Through benchmark testing on actual systems, the changes in system performance and carbon footprint under different workloads are validated.

### Conclusions
- **Carbon Footprint during the Hardware Manufacturing Phase**: The manufacturing carbon footprint of GPUs and CPUs is relatively high, but when normalized to performance, the carbon footprint of CPUs is higher.
- **Carbon Footprint of Memory and Storage Devices**: The carbon footprint of memory and storage devices cannot be ignored, especially the packaging carbon footprint of DRAM, which accounts for a large proportion.
- **Impact of Regional Carbon Intensity**: The differences in carbon intensity across different regions are significant, and cross-regional job scheduling can utilize time differences to reduce carbon emissions.
- **Impact of Increasing the Number of GPUs**: Increasing the number of GPUs can improve performance, but the carbon footprint will also increase linearly, leading to an increase in the carbon footprint per unit of performance.
- **Carbon Footprint Distribution in Supercomputers**: In leading supercomputers, GPUs and CPUs are the main sources of carbon footprint, but the proportions vary between different systems.

Through these analyses, the paper provides important references and suggestions for the sustainable development of HPC systems, emphasizing the importance of carbon footprint awareness in the design and operation of HPC systems.