Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
September 10, 2025ACM SIGEnergy Energy Informatics Review

Re-Evaluating Storage Carbon Emissions In Machine Learning Workloads

View Full Paper
Ask AI
Bookmark
Share

Authors

DKDorota KopczykACAbhishek Chandra

Discussion

Loading...

Member takes

Overview

Comparison assesses carbon impact of SSDs and HDDs in ML tasks, highlighting device and grid considerations.

Key Points

  • SSDs show improved efficiency in ML workloads despite higher embodied carbon emissions.
  • In our analysis, SSDs outperform HDDs in total emissions due to lower operational carbon and faster speeds.
  • The study used the MLPerf Storage benchmark and regional energy data to evaluate carbon impact in ML training.
  • Considering energy mix is crucial for carbon-aware ML infrastructure beyond just device specifications.

Cite This Study

Kopczyk et al. (2025) studied this question.

synapsesocial.com/papers/68c1b81f54b1d3bfb60ec796https://doi.org/10.1145/3757892.3757908
View Full Paper
Ask AI
Bookmark
Share