In its CES 2024 blog, NVIDIA characterized its announcements as “A Launching Pad for Generative AI” – fitting, as 2023 was certainly the year that Gen AI became mainstream.
I wanted to see what NVIDIA has coming. Arguably no company is more important to advancing AI than the chip giant. As critical as Intel was to the PC era, NVIDIA will play a similar role for AI.
At CES 2024, NVIDIA made several announcements that spanned consumer computing and gaming, the automotive industry, and robotics. While these are consumer-focused, IT pros must be aware of them, as the consumer markets typically foreshadow what’s coming in business technology.
TABLE OF CONTENTS
NVIDIA: AI Computing for Gamers, Content Creators
The lead announcement was the rise of artificial intelligence laptops where NVIDIA unveiled the new GeForce RTX 40 SUPER Series GPUs, AI-ready laptops, and generative AI tools—all focusing on gamers and creators.
With an installed base of 100 million GPUs and 500 RTX games and apps, GeForce RTX is the world’s largest platform for gamers, creators, and generative AI. These GPUs are central to Gen AI, including popular services from ChatGPT and Stable Diffusion.
In a briefing before CES, Justin Walker, Senior Director of Product Management at NVIDIA, stated that RTX is NVIDIA’s fastest-growing laptop platform. “It’s grown 5X in the last four years, and there’s now more than 50 million RDX laptops around the world,” he told me. “At CES, we’re announcing a new wave of RTX laptops, launching from every major OEM.”
This includes Acer, ASUStek, Dell, HP, Lenovo, and Samsung, to name a few. The RTX laptops range from small 14-inch laptops up to larger 18-inch models. “All of them are powered with Tensor Cores,” he added. “Every one of these are going to be AI-ready.”
In addition, the company announced three new desktop GPUs, all based on the Ada Lovelace architecture. These GPUs will improve performance for gamers, with more cores and increased memory speed. “This is a fantastic GPU for fully ray trace gaming at 4K, in the highest resolution and image quality,” he said. “It can provide a lot of AI horsepower for super-fast generative AI.”
These are the most capable AI PCs and laptops you can get. “You’ve got up to 1300 TOPS of processing power,” he added. “We’ve had a body of work developed over the past decade with CUDA and TensorRT to help developers get the most out of these AI processors.”
While these are all consumer use cases, the ability to run large language models on a local PC will improve the experience of video conferencing, metaverse, and AR/VR applications, as well as many vertically specific workloads.
Accelerated Large Language Models
Jeff Fisher, Senior VP for GeForce at NVIDIA, said that the October 2023 release of the TensorRT-LLM library for Windows accelerated large language models such as Llama 2 and Mistral up to five times on RTX PCs.
The company has upped its game with its new “Chat with RTX,” which provides access to an RTX-accelerated LLM using the enterprise’s own data—including everything from locally stored documents to YouTube videos. Chat with RTX uses retrieval-augmented generation (RAG), which enhances the accuracy and reliability of GenAI models.
NVIDIA introduced TensorRT acceleration for Stable Diffusion XL and SDXL Turbo using the Automatic1111 text-to-image app. NVIDIA says TensorRT provides a boost in performance of up to 60%.
Twitch Livestreaming, Robotics Simulation
Nvidia said it will work with Twitch to release multi-encode livestreaming, enabling streamers to send up to three concurrent streams to Twitch at different resolutions and quality so each viewer gets the optimal experience.
More than 500 Activision/Blizzard games and apps utilize RTX technology, including the award-winning Alan Wake 2, Horizon Forbidden West, Pax Dei, and Diablo IV. The Activision partnership extends to the cloud with GEForce NOW.
NVIDIA Isaac, the company’s robotics simulation app, is speeding up bringing robots to the real world. The use of co-operative robots (co-bots) in the workplace is growing in verticals such as manufacturing and warehousing and will be coming to healthcare and other industries. The ability to train first in a simulated environment can save thousands of hours of training time, as the simulations can cover a complete set of scenarios, even ones that are hard to replicate in the real world.
My Key Takeaways
As you would expect from NVIDIA, this set of announcements was comprehensive. While other silicon companies are pecking away at NVIDIA’s AI dominance, any company looking to compete with NVIDIA will have to think beyond just the GPU.
NVIDIA’s strength is in building systems that include the GPU, software, and other hardware from a broad ecosystem. For me, the highlight of NVIDIA’s announcements was the AI laptops, as they brings generative AI capabilities to the desktop to complement what’s already happened in the cloud.