Martin Yip, AWS | Nvidia GTC 2024

on YouTube ↗14m

" summary

AWS announces Project Ceiba, a cloud supercomputer built for NVIDIA R&D using the Grace Blackwell/G B Blackwell GPU architecture and claiming 2x performance versus Hopper with lower energy use. The deployment uses Nitro, EFA, KMS and Nitro Enclaves for trusted execution, targets P-series for training and G-series/Inferentia for low-latency inference, and offers tooling (Q instance selector, Compute Optimizer) across 750+ instance types to match workloads.

AI-written summary · about

" chapters (12)

1. 0:00Intro
2. 1:20Nvidia Partnership
3. 2:20Innovation Curve
4. 3:20New Use Cases
5. 4:20Is Nitro Ready
6. 5:20Security
7. 6:20Cloud
8. 8:20AWS
9. 9:20Training vs Inference
10. 10:20Characteristics of Training
11. 11:20Discerning Patterns
12. 13:20Conclusion

" related talks

• 22mFireside ChatRaj Pai, Elias Khnaser · AWS Silicon Innovation Day 2023
• 31mRealize price-performance advantage for AIAkanksha Balani · AWS AMER Summit May 2021
• 32mBuild without limits with AWS Cloud infrastructureTomas Mihalyi · AWS Summit Sydney 2025
• 25mSponsored Session: Accelerating GenAI Inference: From AWS Deep Learning...P. Nguyen, A. Zhao · PyTorch Conference 2025
• 30mAccelerate deep learning & Habana Gaudi–based Amazon EC2 DL1 instancesGreg Serochi, Dvij Bajpai · AWS re:Invent 2022
• 52mKeynote Address: Insights Gained from Delivering Two Generations of AI Supercomputers...Seetharami Seelam · FAST '25
• 31mRealize price-performance advantage for AIAkanksha Balani · AWS Summit Online – Americas 2021
• 01h 13mEngineering the Cloud for Tomorrow: Performance, Speed and Scale for AI WorkloadsMahesh Thiagarajan · Oracle AI World 2025