Sustainability at AWS re:Invent 2022 All the talks and videos I could find…
Adrian Cockcroft
FEBRUARY 13, 2023
Provides comparison of inference workload on P4dn GPU instances vs. AWS Trainium saving 92% energy and 90% cost, and training workload on P4dn vs. AWS Inferentia 2.6x shorter training time, saving 54% energy and 75% cost. Building a data lake of detailed information about energy use of many physical devices.
Let's personalize your content