How much energy is needed to generate an image? ๐จ๐ง โก๏ธ
Up to 4.08 Wh โ like charging your phone to 40%!
In our new study we tested 17 models & 9,000+ runs.
Other key finds:
โก๏ธ Model energy use varies up to 46x
๐ Resolution matters, prompts don't
๐ ๏ธ Quantization โ savings
๐ Preprint: lnkd.in/dKWWAETW
Thank you @shubhendu.bsky.social (one of the authors here)!
If you have any doubt or questions on the research don't hesitate to reach out ๐
@robertoverdecchia.bsky.social presenting โEnergy-Aware Software Testingโ at #ICSE2025 @icseconf.bsky.social
He is referring to my paper โAn Inconvenient Truth in Software Engineering? The Environmental Impact of Testing Open Source Java Projectsโ, see azaidman.github.io/publications...
You have not yet registered for #TechDebt? Then hurry up and do not miss an exciting program!
โ
๐๐ฒ๐๐ป๐ผ๐๐ฒ ๐๐ฎ๐น๐ธ๐
๐ ๐ง๐ฒ๐ฐ๐ต๐ป๐ถ๐ฐ๐ฎ๐น ๐ฝ๐ฟ๐ฒ๐๐ฒ๐ป๐๐ฎ๐๐ถ๐ผ๐ป๐
๐ญ ๐๐ป๐ฑ๐๐๐๐ฟ๐ ๐๐ฎ๐น๐ธ๐
And much more! Join us in Ottawa ๐จ๐ฆ next April 27-28!
โก๏ธ Register here: conf.researchr.org/attending/Te...
Very interesting indeed, thank you for sharing!
Software testing consumes ever-increasing resources. GenAI may only exacerbate the issue.
What can we do about it?
In our new @icseconf.bsky.social NIER paper we present "Energy-Aware Software Testing", with both broad strategies and a PoC algorithm.
๐: robertoverdecchia.github.io/papers/ICSE_...
๐ This year, we are launching a brand-new Journal First track! ๐ A unique chance to showcase your groundbreaking #TechDebt journal paper directly to an engagedย audience! Don't miss this chance to make a lasting impact and submit yours by Jan 17 AoE!
๐ tinyurl.com/TechDebtJ1