I had the amazing opportunity to attend NVIDIA's GPU Technology Conference (GTC) 2024 in person. It was an incredibly insightful experience, and I came away learning so much about the latest breakthroughs in AI and accelerated computing.
In his keynote, NVIDIA CEO Jensen Huang walked through the company's journey from its founding in 1993 to the current era of generative AI breakthroughs like ChatGPT.
Here is a short summary of my notes from the keynote.
Announcing Blackwell - NVIDIA's Next-Gen GPU
The main announcement was NVIDIA's next-generation GPU called Blackwell. It packs in an immense 208 billion transistors and delivers 2.5x more computing performance compared to their previous GPU, Hopper. With Blackwell, NVIDIA has created a single rack system that achieves 1 exaFLOPs of performance - that's an unprecedented amount of computing power in one place.
Turbocharging Large AI Model Training
This massive computing capability from Blackwell allows training of extremely large AI models like GPT with 1.8 trillion parameters in just 90 days using 2,000 Blackwell GPUs. For the same workload, the previous Hopper architecture would require 8,000 GPUs. Blackwell also provides dramatic 30x performance gains for running AI inference workloads.
Cloud Partnerships for AI
NVIDIA announced major partnerships with cloud providers AWS, Google, and Microsoft to deploy Blackwell's computing power globally for AI training and inference in the cloud.
Introducing NVIDIA AI Foundry Platform
They unveiled a new AI Foundry software platform with tools that simplify deploying AI models (NIM), building AI applications (NeMo), and creating digital twins of real-world things (Omniverse).
Healthcare Innovations with Digital Twins
NVIDIA showcased innovations like using digital twins for disease analysis and drug discovery.
Robotics Breakthroughs with Omniverse
For robotics, they introduced new chips, software, and simulation tools to train robots in virtual environments before real-world deployment.
The New AI Industrial Revolution
Overall, Huang described NVIDIA's vision for ushering in a new AI industrial revolution powered by their accelerated computing platforms. This enables creating digital twins of nearly anything and deploying AI to tackle challenges like sustainable practices, chip design, drug discovery and more.
It was truly awe-inspiring to witness these groundbreaking developments first-hand at GTC 2024. I had a great time learning about the immense potential of accelerated computing and generative AI.