articleJun 16, 2023GOLD OA

TPU v4: An Optically Reconfigurable Supercomputer for Machine Learning with Hardware Support for Embeddings

Google (United States) · University of California, Berkeley

Indexed incrossref

Abstract

In response to innovations in machine learning (ML) models, production workloads changed radically and rapidly. TPU v4 is the fifth Google domain specific architecture (DSA) and its third supercomputer for such ML models. Optical circuit switches (OCSes) dynamically reconfigure its interconnect topology to improve scale, availability, utilization, modularity, deployment, security, power, and performance; users can pick a twisted 3D torus topology if desired. Much cheaper, lower power, and faster than Infiniband, OCSes and underlying optical components are

Citation impact

410
total citations
FWCI
178.46
Percentile
100%
References
28
Citations per year

Authors

14

Topics & keywords

Keywords
  • Supercomputer
  • Computer science
  • Dataflow
  • InfiniBand
  • PCI Express
  • Modularity (biology)
  • Parallel computing
  • Embedded system
UN Sustainable Development Goals
  • Affordable and clean energy
No related works found for this paper.