Exploring energy consumption of AI frameworks on a 64-core RV64 Server CPU

Giulio Malenza, Francesco Targa, Adriano Marques Garcia, Marco Aldinucci, Robert Birke · 2025

Read Exploring energy consumption of AI frameworks on a 64-core RV64 Server CPU by Giulio Malenza, Francesco Targa, Adriano Marques Garcia, Marco Aldinucci, Robert Birke (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-03774.txt.

Description

Giulio Malenza and co-authors present a computing study from 2025 on the energy cost of smart software. Big models eat power, and new chip designs promise thrift. The RISC-V design is open and built for low energy. The team tested how three top software frameworks fare on a 64-core RISC-V server chip.

They ran deep learning models in guess mode and metered the power draw. Two frameworks that share a fast math back end used less energy. The third, paired with a classic math library, used more. The gap came from the back end, not the model itself. The authors say framework choice matters as much as hardware for green computing. Their bench marks give a base for future RISC-V tuning.

Catalog record

CREATORGiulio Malenza, Francesco Targa, Adriano Marques Garcia, Marco Aldinucci, Robert Birke
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE23 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-03774

Browse the full library · About the Tower of Babel