Nvidia unveils Blackwell AI chips, boasting 208 billion transistors. CEO Jensen Huang forecasts a robotic future, emphasising AI’s transformative role.
Nvidia unveils Blackwell AI chips that also power the company’s Project GR00T.
Nvidia, a tech leader in the world of Artificial Intelligence (AI), recently unveiled its next-generation AI chips called Blackwell at the GTC conference in San Jose, California. With a staggering 208 billion transistors, these chips promise to revolutionise the AI landscape, powering a multitude of applications from data centres to humanoid robots.
Huang’s keynote address was not merely a showcase of technological prowess but a glimpse into a future dominated by robotics. He emphasised that AI is the driving force behind an evolving economy, and the Blackwell chips serve as the engine propelling this new industrial revolution. This sentiment aligns with Huang’s prior statements about the transformative potential of AI in various industries.
Blackwell: A Processing Powerhouse
Blackwell boasts significant performance improvements for both training and running AI models. Equipped with a staggering 208 billion transistors, these chips are designed to power the next generation of computers used by major data centre operators like Google, Microsoft, and Amazon.
The first iteration, GB200, is slated for release later this year. The report acknowledges the ongoing struggle for businesses to acquire current-generation Hopper chips, but Huang emphasises the need for even more powerful hardware.
Blackwell’s sheer complexity presents manufacturing challenges. To ensure seamless operation with this many transistors, Nvidia is partnering with Taiwan Semiconductor Manufacturing Co. (TSMC) which will utilise their advanced 4NP manufacturing process.
Blackwell goes beyond traditional AI functions like speech recognition. It aims to tackle complex tasks like generating realistic 3D videos. This enhanced processing power will also improve connectivity with other chips, allowing users to pair Blackwell with Nvidia’s Grace CPU and networking options using InfiniBand or Ethernet protocols. Existing HGX server machines will be upgraded with the new chip as well.
Nvidia’s journey from a graphics card provider for gamers to a major AI player highlights the versatility of their technology. Their GPUs excel at handling complex parallel tasks, a core requirement for advanced AI applications.
Project GR00T: Ushering in the Robot Age
Project GR00T, another major announcement at the GTC, further strengthens Nvidia’s commitment to robotics. This new computer, powered by Blackwell chips, is specifically designed for humanoid robot manufacturers.
Project Groot offers a revolutionary approach – robots that learn through observation. By utilising Nvidia’s simulation tools, robots can learn human-like movements and understand natural language, paving the way for more intuitive and collaborative robots.
Huang concluded the event with a glimpse into the future. Two robots trained using Nvidia’s simulation tools joined him on stage, leaving a lasting impression. He boldly declared, “everything that moves will be robotic,” signifying his belief in a future dominated by intelligent machines.
While the extent of robotic integration remains to be seen, Nvidia’s advancements in AI and robotics paint a picture of a rapidly evolving landscape. The coming years will likely witness a surge in AI-powered applications and a growing presence of robots across various industries. Nvidia, with its powerful new chips and ambitious projects like Groot, is well-positioned to be a major force shaping this transformation.
Source link
credite