Dissecting the NVIDIA Hopper Architecture through Micro-benchmarking and Multiple Level Analysis
Published in ACM Transactions on Computer Systems (TOCS), 2025
This journal version offers a detailed and multi-level analysis of the NVIDIA Hopper architecture, building upon earlier conference results. It expands micro-benchmark coverage, deepens architectural insights, and discusses implications for future GPU designs and high-performance applications.
Recommended citation: Weile Luo, **Ruibo Fan**, Zeyu Li, et al., "Dissecting the NVIDIA Hopper Architecture through Micro-benchmarking and Multiple Level Analysis," *ACM Transactions on Computer Systems (TOCS)*, under review.
Download Paper | Code | Download Bibtex