Cost-Effective Deep Learning Infrastructure with NVIDIA GPU

Aatiz Ghimire, Shahnawaz Alam, Siman Giri, Madhav Prasad Ghimire · 2025

Read Cost-Effective Deep Learning Infrastructure with NVIDIA GPU by Aatiz Ghimire, Shahnawaz Alam, Siman Giri, Madhav Prasad Ghimire (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-2503-11246.txt.

Description

Aatiz Ghimire and co-authors describe building a low-cost deep learning cluster in Nepal from gaming graphics cards. Deep learning, big data, and simulation work all hunger for computing power, but new hardware is hard to afford in developing countries. The team instead linked four machines, each with a consumer NVIDIA card, into one small cluster.

One master node runs the scheduling and software tools, while three worker nodes do the heavy computation. Shared network storage feeds all four, and a security tool blocks password-guessing attacks on the cluster's public address. The authors recount the snags met during design and setup. They offer the build as proof that careful use of modest hardware can serve demanding research and teaching needs.

Catalog record

CREATORAatiz Ghimire, Shahnawaz Alam, Siman Giri, Madhav Prasad Ghimire
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE28 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-11246

Browse the full library · About the Tower of Babel