Realize price-performance advantage for AI
summary
AWS EC2 adds instances with up to eight Gaudi accelerators claiming 40% better price-performance than current GPU instances for training, accessible via SageMaker, ECS, EKS or plain EC2. Gaudi chips pack 32 GB HBM2, on‑chip 10×100Gb RoCE ports for scaling, and SynapseAI support for TensorFlow and PyTorch; an 8‑Gaudi system hits about 12,000 images per second on ResNet50. Target workloads include NLP, vision, recommendation systems, and distributed training using Elastic Fabric Adapter.
AI-written summary · about
chapters (11)
- 0:00Introduction to AI and collaboration
- 1:39AI transforming industries with data
- 3:29Intel’s AI hardware and software stack
- 6:24AWS EC2 instances powered by Intel
- 9:24Performance gains with Intel on AWS
- 11:35Global AI success stories with Intel and AWS
- 15:53Thorn’s fight against child exploitation with AI
- 19:24AWS and Intel’s decade-long partnership
- 22:36Introducing Habana Gaudi processors
- 23:41Habana Gaudi for deep learning training
- 29:37Summary and call to action
related talks
31mRealize price-performance advantage for AI
30mAccelerate deep learning & Habana Gaudi–based Amazon EC2 DL1 instances
50mThe democratization of tomorrow’s technology
35mAccelerate your ML innovation with AWS ML infrastructure services
24mGenerative AI Chips
14mEngineering the AI factory: Accelerating CAD workloads on
35mReinventing with generative AI on AWS
50mAI acceleration at the edge