Performance of AI Workloads with the Telum Processor
Project and Program:
MVS,
MVS Performance
Tags:
Proceedings,
SHARE Columbus 2022
A new accelerator, Telum processor, is added to the z16 system to support AI
workloads. Much has been published about its hardware specifications. But how
does it stack up against production workloads? This presentation shows you our
observations of the performance characteristics of the Telum processor. It
reveals techniques of performance analysis of AI workloads. It shares with you
data available for performance problem determination. Best practices of AI
workload tuning will also be covered. We are the team who produced the AI
performance proof points to support the launch of the z16 system.
Back to Proceedings File Library