Amazon's cloud lets several AI services share one graphics chip, so less paid capacity sits idle
Why it matters
Teams running many small AI models on AWS can fit more of them on each GPU instance, which cuts idle capacity you pay for.
Ask your team
How many of our hosted models use only a fraction of their GPU, and what would sharing save?
What happened
Amazon's SageMaker lets you split one physical GPU into several slices, so a single can serve more than one . Until now, the smallest unit you could request was a whole GPU, so an eight-GPU instance topped out at eight models whatever each one needed.
Tap an underlined word to see what it means.
The context No earlier Seventy story on this in the last 30 days.
Read the original at AWS Machine Learning Blog
Something wrong in this story? Report a mistake. We check it against the source within 24 hours.
Share with your take
Start from our line and make it yours. It stays on your device until you post it.
Ask about this story
Answered only from this story, with the line it came from.