New Trn1n Instances Trainium Chips

Generative AI models need to be trained so they offer the right answer, image and insight, Borno said. These training runs require enormous compute resources and have been known to be expensive.

However, AWS’ new Trn1n instances now run on AWS’ custom Trainium chips and offer massive networking capability, which Borno said is key for training these models quickly and in a cost-efficient manner.

“These are designed to deliver up to 20 percent higher performance over the prior generation of Trainium1 for large network-intensive models,” she said. With Trn1n instances, developers will be able to train models more quickly and at less expense.