Nvidia 2024 AI Event


Today, we find ourselves amidst a gathering of innovators, not mere spectators at a concert. Here, at this developers’ conference, the air is charged with the electricity of discovery. What lies ahead is a journey through the realms of science, where algorithms dance with computer architecture, and mathematics weaves its intricate patterns.

Let’s first demystify the name “Blackwell.” It’s not merely a chip; rather, it stands as the epitome of a platform. While some might associate us solely with GPUs, the truth is far more fascinating. The GPUs of yesteryears pale in comparison to the marvel that is Hopper. Picture this: 28 billion transistors seamlessly integrated, blurring the lines between two dies. These dies, coalescing into what seems like a singular entity, possess a collective consciousness of 10 terabytes per second. Therein lies the magic of Blackwell—a monolithic chip with no memory locality or cache issues, seamlessly merging into the fabric of computing.

Now, let’s delve deeper into the realm of systems. Picture a scenario where Blackwell effortlessly integrates into existing frameworks, sliding into place alongside Hopper. This compatibility ensures a smooth transition, minimizing the challenges of implementation. With Hoppers deployed across the globe, the infrastructure remains consistent, from power to software, enabling a seamless upgrade to Blackwell.

But Blackwell is just the beginning. Enter the MVY Link switch—a behemoth boasting 50 billion transistors, nearly rivaling the size of Hopper itself. With its four MV links, each blazing at 1.8 terabytes per second, this switch heralds a new era of connectivity. Imagine a network where every GPU communicates at full speed simultaneously—a symphony of computational prowess.

Behold the DGX, a testament to the convergence of power and efficiency. Within a single rack lies an exaflops AI system, a rarity in today’s landscape. As we embark on this journey, we extend our gratitude to our esteemed partners. AWS, Google, Oracle, Microsoft—each contributing to the tapestry of accelerated computing.

But our vision extends beyond raw computational power. Enter Nvidia AI Foundry, where partnerships with SAP, ServiceNow, Cohesity, Snowflake, NetApp, and Dell pave the path for AI integration across industries. Together, we’re sculpting the future of enterprise AI, one innovation at a time.

In the realm of robotics, Nvidia’s Project Groot stands tall—a foundation for humanoid robot learning. Powered by Jetson Thor chips and nurtured in the digital womb of Omniverse, Groot embodies the fusion of AI and robotics. With Isaac Lab and Osmo orchestrating the symphony of training and simulation, Groot learns to emulate human movement, heralding a new era of human-robot interaction.

And amidst this sea of innovation, let’s not forget our special guests—Orange and Green, the bdx robots of Disney. Powered by Jetson, these marvels of technology epitomize the convergence of imagination and engineering.

As we bring this journey to a close, let’s not merely see Blackwell as a chip but as a symbol of boundless possibilities. For in the heart of every GPU lies the soul of Nvidia—an intersection of graphics, physics, and artificial intelligence. Today, we stand at the dawn of a new era, where the name “Blackwell” echoes with the promise of innovation and discovery.

Leave a Comment