ThirstyFLOPS: Water Footprint Modeling and Analysis Toward Sustainable HPC Systems

TL;DR

ThirstyFLOPS addresses the underexplored water footprint of HPC by distinguishing embodied and operational water use and by incorporating region-specific factors via metrics such as WUE, PUE, and EWF. The framework computes the total water footprint $W = W_{embodied} + W_{operational}$ with detailed decompositions for manufacturing, cooling, and energy generation, and applies it to four TOP500 systems to reveal embodied vs. operational trade-offs under varying energy mixes and water scarcity. Key findings show GPUs and HDDs can dominate embodied water, indirect water use from energy generation can be substantial, and regional water scarcity can invert footprint rankings; nuclear power can dramatically reduce carbon but has location-dependent water impacts. The work offers an open-source tool for water-aware HPC planning and highlights the need for co-optimizing water, carbon, and energy in HPC infrastructure and scheduling.

Abstract

High-performance computing (HPC) systems are becoming increasingly water-intensive due to their reliance on water-based cooling and the energy used in power generation. However, the water footprint of HPC remains relatively underexplored-especially in contrast to the growing focus on carbon emissions. In this paper, we present ThirstyFLOPS - a comprehensive water footprint analysis framework for HPC systems. Our approach incorporates region-specific metrics, including Water Usage Effectiveness, Power Usage Effectiveness, and Energy Water Factor, to quantify water consumption using real-world data. Using four representative HPC systems - Marconi, Fugaku, Polaris, and Frontier - as examples, we provide implications for HPC system planning and management. We explore the impact of regional water scarcity and nuclear-based energy strategies on HPC sustainability. Our findings aim to advance the development of water-aware, environmentally responsible computing infrastructures.

Figures (14)

• Figure 1: Carbon intensity, water scarcity index, and HPC power consumption in the US.
• Figure 2: Water footprint components in the HPC systems, including operational water footprint and embodied water footprint.
• Figure 3: Embodied water footprint contribution of different hardware components, including CPU, GPU, DRAM, HDD, and SSD.
• Figure 4: Comparison of embodied and operational water footprint under different EWF, WUE, and WSI scenarios.
• Figure 5: Different energy sources have different energy water factors (EWFs) and carbon intensities.