In today's world, High-Performance Computing (HPC) is driving scientific research forward at an astonishing rate, but behind every top HPC device lies a high-load power grid, which has sparked deep concern among environmentalists. This study develops a mathematical model to evaluate the power consumption of global HPC equipment and quantify its environmental impact, providing a basis for energy optimization and sustainable development. By systematically analyzing the global environmental impact of HPC through a series of models focused on energy consumption and related emissions, we first created a GPU survival function to estimate the global number of HPC devices in 2023. Using Monte Carlo simulation and Markov chain models, we estimated the power consumption of individual HPC centers under both full load and average utilization conditions, subsequently calculating the total annual power consumption of global HPC centers. Next, we developed models to estimate the total carbon emissions from global HPC energy consumption, considering various energy production methods and energy mix scenarios. Additionally, we created a gray prediction model to forecast the GPU market value in 2030, combining it with the GPU survival function to predict the number of global HPC centers in 2030. We also developed an electricity price fluctuation model to account for increased energy demand from other sectors and analyzed the environmental impact of global HPC centers in 2030 under different energy mix structures. Furthermore, we extended the model to assess the impact of increasing renewable energy (specifically wind energy) to 100% in the energy mix, evaluating its potential to reduce carbon emissions. Finally, we conducted a sensitivity analysis, incorporating seawater cooling for HPC centers and artificial intelligence to dynamically adjust GPU power based on wind speed predictions.
Read full abstract- All Solutions
Editage
One platform for all researcher needs
Paperpal
AI-powered academic writing assistant
R Discovery
Your #1 AI companion for literature search
Mind the Graph
AI tool for graphics, illustrations, and artwork
Journal finder
AI-powered journal recommender
Unlock unlimited use of all AI tools with the Editage Plus membership.
Explore Editage Plus - Support
Overview
548 Articles
Published in last 50 years
Articles published on Energy-related Emissions
Authors
Select Authors
Journals
Select Journals
Duration
Select Duration
506 Search results
Sort by Recency