Nvidia's ambitious future is to bring AI to every industry where GPU tech can be leveraged

Julio Franco

Posts: 8,996   +1,972
Staff member
Why it matters: During the GTC 2023 keynote, Nvidia's CEO Jensen Huang highlighted a new generation of breakthroughs that aim to bring AI to every industry. In partnership with tech giants like Google, Microsoft, and Oracle, Nvidia is making advancements in AI training, deployment, semiconductors, software libraries, systems, and cloud services. Other partnerships and developments announced included the likes of Adobe, AT&T, and vehicle maker BYD.

Huang noted numerous examples of Nvidia's ecosystem in action, including Microsoft 365 and Azure users gaining access to a platform for building virtual worlds, and Amazon using simulation capabilities to train autonomous warehouse robots. He also mentioned the rapid rise of generative AI services like ChatGPT, referring to its success as the "iPhone moment of AI."

Based on Nvidia's Hopper architecture, Huang announced a new H100 NVL GPU that works in a dual-GPU configuration with NVLink, to cater to the growing demand for AI and large language model (LLM) inference. The GPU features a Transformer Engine designed for processing models like GPT, reducing LLM processing costs. Compared to HGX A100 for GPT-3 processing, a server with four pairs of H100 NVL can be up to 10x faster, the company claims.

With cloud computing becoming a $1 trillion industry, Nvidia has developed the Arm-based Grace CPU for AI and cloud workloads. The company claims 2x performance over x86 processors at the same power envelope across major data center applications. Then, the Grace Hopper superchip combines the Grace CPU and Hopper GPU, for processing giant datasets commonly found in AI databases and large language models.

Furthermore, Nvidia's CEO claims their DGX H100 platform, featuring eight Nvidia H100 GPUs, has become the blueprint for building AI infrastructure. Several major cloud providers, including Oracle Cloud, AWS, and Microsoft Azure, have announced plans to adopt H100 GPUs in their offerings. Server makers like Dell, Cisco, and Lenovo are making systems powered by Nvidia H100 GPUs as well.

Because clearly, generative AI models are all the rage, Nvidia is offering new hardware products with specific use cases for running inference platforms more efficiently as well. The new L4 Tensor Core GPU is a universal accelerator that is optimized for video, offering 120 times better AI-powered video performance and 99% improved energy efficiency compared to CPUs, while the L40 for Image Generation is optimized for graphics and AI-enabled 2D, video, and 3D image generation.

Also read: Has Nvidia won the AI training market?

Nvidia's Omniverse is present in the modernization of the auto industry as well. By 2030, the industry will mark a shift towards electric vehicles, new factories and battery megafactories. Nvidia says Omniverse is being adopted by major auto brands for various tasks: Lotus uses it for virtual welding station assembly, Mercedes-Benz for assembly line planning and optimization, and Lucid Motors for building digital stores with accurate design data. BMW collaborates with idealworks for factory robot training and to plan an electric-vehicle factory entirely in Omniverse.

All in all, there were too many announcements and partnerships to mention, but arguably the last big milestone came from the manufacturing side. Nvidia announced a breakthrough in chip production speed and energy efficiency with the introduction of "cuLitho," a software library designed to accelerate computational lithography by up to 40 times.

Jensen explained that cuLitho can drastically reduce the extensive calculations and data processing required in chip design and manufacturing. This would result in significantly lower electricity and resource consumption. TSMC and semiconductor equipment supplier ASML plan to incorporate cuLitho in their production processes.

Permalink to story.



Posts: 296   +172
As for the limit to the application of Moore's law, we may be close to the natural limit of the minimum possible distance between two transistors, but an evolution of the technology so that microprocessors use photons instead of electrons will allow them to operate at higher frequencies, even on the order of TeraHz.
This is possible because photons travel without ohmic resistance, which generates heat. So even with the same transistor density, efficiency will continue to increase.


Posts: 555   +1,051
We had the internet of things and now we're going to have the AI of things. I need an AI powered microwave to stop it from giving me a hot plate full of cold food.
When reheating, Micro for 40 seconds first for non-frozen or 90 seconds for frozen foods and then convect at 150 for 10-20 minutes. Never convect at full temp, don't micro for too long.


Posts: 295   +136
ARM servers are coming, and they are coming fast. AMD and Intel better react.

ARM has 5-10% Server marketshare today. Increasing rapidly. Performance per watt matters alot. Many workloads can easily be done by ARM.


Posts: 489   +415
Arranger comparisons, sorters, categorise is what we have AI it is not and will never be unless we move away from digital and go analogue. zero and one will never be self aware. 011010010101


Posts: 827   +607
"AI is nothing but a brand. A powerful brand, but an empty promise. The concept of “intelligence” is entirely subjective and intrinsically human. Those who espouse the limitless wonders of AI and warn of its dangers – including the likes of Bill Gates and Elon Musk – all make the same false presumption: that intelligence is a one-dimensional spectrum and that technological advancements propel us along that spectrum, down a path that leads toward human-level capabilities. Nuh uh. The advancements only happen with labeled data. We are advancing quickly, but in a different direction and only across a very particular, restricted microcosm of capabilities.

The term artificial intelligence has no place in science or engineering. “AI” is valid only for philosophy and science fiction – and, by the way, I totally love the exploration of AI in those areas."