NEWS  AI  March 19, 2024  Reading time: 3 Minute(s)

Max (RS editor)

In a keynote at the GPU Technology Conference 2024 in San Jose, California, Nvidia CEO Jensen Huang showcased the company's latest advancements in artificial intelligence (AI) technology, aimed at revolutionizing enterprise solutions across various industries. Held on March 18, 2024, this keynote marked the unveiling of Nvidia's new suite of products and services designed to propel AI learning and application forward, promising significant breakthroughs in fields ranging from robotics and weather forecasting to drug discovery and warehouse automation.

At the heart of these innovations lies the highly anticipated Blackwell GPU platform, introduced as the successor to the Hopper GPU platform released in 2022. The Blackwell GPU represents a monumental leap in computing power, featuring dual 104-billion-transistor GPU dies built on a TSMC 4NP process, capable of delivering a staggering 20 FP4 petaflops of performance. Equipped with 192GB of HBM3e memory boasting transfer speeds of up to 8TB/s, the Blackwell GPU sets a new standard for AI processing capabilities, outperforming previous generations by orders of magnitude.

The deployment of Blackwell GPUs in enterprise settings promises to revolutionize AI training processes, which traditionally require extensive computational resources and time. By leveraging the unparalleled performance of Blackwell, enterprises can expect accelerated AI training times, with significant reductions in power consumption compared to previous iterations.

Complementing the Blackwell GPU platform are Nvidia's NIMS (Nvidia Inference Microservice) pre-packaged AI models, designed to harness the newfound capabilities of Blackwell GPUs for rapid AI inference tasks. Utilizing Kubernetes for deployment on local systems or cloud environments, NIMS simplifies the integration of AI capabilities into existing workflows, paving the way for a future where AI services can be effortlessly created and customized without the need for extensive programming knowledge.

In a bold move towards advancing medical research and healthcare, Nvidia also unveiled plans to develop NIMS trained on biological and medical data, offering researchers unprecedented access to AI-driven insights that could revolutionize the field of medicine, particularly in drug discovery and disease diagnosis.

Furthermore, Nvidia's foray into autonomous systems reached new heights with the introduction of the Thor ASIL-D computer, set to power autonomous electric vehicles (EVs) manufactured by BYD. Leveraging AI system-on-chip (SoC) technology, the Thor ASIL-D computer promises to enhance safety and reliability in autonomous driving, positioning BYD as a pioneer in the development of Level 4 autonomous EVs.

Not content with revolutionizing transportation alone, Nvidia showcased its advancements in robotics with Project Groot humanoid robots. By training these robots in a digital twin environment before deploying them in real-world scenarios, Nvidia aims to redefine the capabilities of robotics, enabling tasks ranging from product sorting to complex assembly operations with unprecedented efficiency and accuracy.

Finally, Nvidia announced enhancements to its Omniverse Cloud platform, expanding support to include Apple Vision Pro headsets alongside existing devices such as Meta Quest and HTC Vive Pro. This move underscores Nvidia's commitment to providing developers with versatile tools for immersive and collaborative experiences powered by AI and cloud computing.

 MEDIA CREDITS: NVIDIA