We are LIVE! talking with AWS’s Dave Brown on re:Invent updates to Graviton, Inferentia & Trainium - Six Five Insiders

Dec 5, 2021

Dave, welcome to the Six Five podcast! I was excited to hear about the latest announcements in Adam’s keynote at re:Invent 2021 yesterday. Particularly, in the area of EC2 Silicon innovation which we have talked a lot about this year. The biggest announcements were the latest generation Graviton3 processor and the new machine learning chip for training, AWS Trainium. Let’s start with Graviton, can you tell us about the latest Graviton offerings?

•  In addition to new EC2 instances, we have also seen more managed services from AWS supporting Graviton2. Can you tell us more about that?
• That’s great to hear! I know you ran a program called the Graviton Challenge to help customers adopt Graviton2. How’s that going?
• In addition to customers, have you seen wide-spread support for Graviton from third party software provides as well?

 

The one of the biggest announcements at re:Invent for me was the Graviton3 processor. Can you tell us more about Graviton3 and how to compares to Graviton2?

• What kinds of workloads can customers run on Graviton3?

 

So, talking about workloads, let us discuss machine learning workloads and AWS silicon innovation for AI/ML. What were the driving factors for AWS to build chips for AI/ML

• So how have you solved the high-performance and low-cost problem?

 

Tell us about how customers are using Inferentia.

 

You mentioned that customers are training larger and more complex deep learning models. How are the new Trainium-based Trn1 instances helping customers with challenges they have with model training?

 

To sum it up, what is your message to our listeners about AWS Silicon innovations and how do they get started?