Nvidia unveils next-generation RTX GPUs

Nvidia unveils next-generation RTX GPUs for AI workflows

Announced at the GTC 2022 keynote
Introduce new products, applications, and partners such as the RTX 40 series

Computing is advancing at an incredible rate. The engine that propels this rocket is accelerated computing, and the fuel is artificial intelligence (AI).”

Nvidia CEO Jensen Huang emphasized the importance of AI in his keynote speech at the ‘NVIDIA GPU Technology Conference (NVIDIA GTC) 2022’ held online on the 21st, and unveiled a new lineup of new products including next-generation RTX GPUs.

He explained that AI continues to create new technologies and new products, from delighting gamers with great graphics to building virtual proving grounds where the world’s largest companies can improve their products.

New ideas, new products and applications sparked by AI will affect industries around the world.

The first thing CEO Jensen Huang introduced was the next-generation ‘GeForce RTX 40 series GPU’ based on Nvidia’s new architecture ‘Ada Lovelace’. He called this a ‘quantum leap’, paving the way for creators of fully simulated worlds. 

The RTX 40 series applies ‘Nvidia DLSS 3’, an AI-based technology that can overcome CPU performance limitations, and provides up to 4 times higher performance than the previous generation products.

The GeForce RTX 40 series to be released this time are ‘GeForce RTX 4090’ and ‘GeForce RTX 4080’, which can process 4 times more than the previous product, RTX 3090 Ti. They will be released in October and November, respectively. 

Among them, the GeForce RTX 4080 16GB is equipped with 9728 CUDA cores and 16GB of high-speed Micron GDDR6X memory. With DLSS 3, it’s 2x faster than the original GeForce RTX 3080 Ti in gaming and more powerful than the GeForce RTX 3090 Ti at lower power.

It also unveiled ‘Grace CPU’ and ‘Grace Hopper Superchip’-based systems that enable high-performance computing for companies pursuing innovation using AI.

Next, it introduced a new ‘OVX Server’ powered by Ada Lovelace L40 data center GPU, which is suitable for 3D Internet construction. “The metaverse will expand into 3D,” he said. “To connect and simulate these worlds, you need a new powerful and flexible computer. The OVX server was built to extend the Metaverse application.”

 

It also announced the launch of ‘Omnivers Cloud’ for building and operating industrial metaverse applications. Omnibus is NVIDIA’s platform for building and running Metaverse applications. Jensen Huang predicted that “the next evolution of the Internet, called the metaverse, will extend to 3D.”

It also introduced a drive chip ‘Thor’ with a performance of 2000 teraflops (1 trillion operations per second). Thor is a superchip that combines a Transformer engine, Ada’s GPU and Grace CPU in one place. It can be used not only for autonomous driving, but also for robots, medical devices, industrial automation, and edge systems.

In addition, the new ‘Jetson Orin Nano’ for robotics, ‘NeMo LLM (NeMo Large Language Model)’, a large-scale language model cloud service for AI and digital biology, ‘Bionemo LLM Service’ and Introduced ‘CV-CUDA’ for real-time graphic processing.

Jensen Huang also revealed key partners and customers worldwide using NVIDIA technology. Deloitte, the world’s No. 1 accounting firm; charter, etc.

In addition, it said that more than 200,000 people from Boeing, Deutsche Bank, Johnson & Johnson, Mercedes-Benz, Siemens, T-Mobile, and US Bank attended this GTC conference session.

In addition, more than 3.5 million developers worldwide have formed a software ecosystem that uses NVIDIA’s 550 software development kits, SDKs and AI models to create more than 3,000 apps. 

“Today we announced a new chip, a new evolution of the platform, and for the first time a new cloud service,” said Jensen Huang. He concluded his keynote speech.

Meanwhile, all of the technologies NVIDIA announced this time are available to enterprises worldwide as new services built by Deloitte, the world’s largest professional services company, based on NVIDIA AI and NVIDIA Omnibus.

By Kim Young-ha, staff reporter [email protected]

Leave a Comment