Nvidia CEO Jensen Huang unveiled several new technologies at CES 2025 Monday, including the GeForce RTX 50 Series desktop and laptop GPUs, which are the company’s most advanced consumer graphics processor units for gamers, creators, and developers.
The new flagship RTX 5090 GPU promises a considerable performance uplift, with 32GB of GDDR7 RAM and 21,760 CUDA cores, and is priced at $1,999. Additionally, Huang announced Project Digits, a personal AI supercomputer, and DLSS 4, among other technologies.
Nvidia claimed the new GeForce RTX 5090 flagship GPU can deliver up to twice as much relative performance for ray-tracing intensive games.
Other more modest GPUs in the series include GeForce RTX 5070 with 6,144 CUDA cores and 12GB of DDR7 RAM, priced at $549; GeForce RTX 5070 Ti with specifications between the RTX 5070 and RTX 5080, priced at $749, and GeForce RTX 5080 with specifications between the RTX 5070 Ti and RTX 5090, priced at $999.
A key part of the RTX 5000-series launch was the introduction of DLSS 4, the latest version of its real-time image upscaling and update to Nvidia’s deep learning super sampling technology, aimed at improving performance in games and other applications.
DLSS 4 is coming to all RTX GPUs, including the RTX 20 series that was discontinued back in 2020, but the older models aren't getting all its features.
In the new GeForce RTX 50 series models, DLSS 4 will enable Multi Frame Generation. The feature generates up to three additional frames for every traditionally rendered one and can help multiply frame rates by up to eight times more than traditional brute-force rendering. NVIDIA says the improvements brought by Multiple Frame Generation on the GeForce RTX 5090 graphics card, its new $1,999 flagship GPU arriving this month, will enable 4K 240 FPS fully ray-traced gaming.
Nvidia also unveiled NVIDIA® Project DIGITS, a personal AI supercomputer that provides AI researchers, data scientists and students worldwide with access to the power of the NVIDIA Grace Blackwell platform.
Project DIGITS features the new NVIDIA GB10 Grace Blackwell Superchip, offering a petaflop of AI computing performance for prototyping, fine-tuning and running large AI models.
With Project DIGITS, users can develop and run inference on models using their own desktop system, then seamlessly deploy the models on accelerated cloud or data center infrastructure.
“AI will be mainstream in every application for every industry. With Project DIGITS, the Grace Blackwell Superchip comes to millions of developers,” said Jensen Huang, founder and CEO of NVIDIA. “Placing an AI supercomputer on the desks of every data scientist, AI researcher and student empowers them to engage and shape the age of AI.”
The personal supercomputer, which can handle AI models with up to 200 billion parameters, will be available in May this year at a $3,000
Huang also announced the launch of Nvidia Cosmos, a platform designed to advance the development of physical AI systems, including robots and autonomous vehicles.
The Cosmos World Foundation Model is a family of open diffusion and autoregressive transformer models for physics-aware video generation, trained on 9,000 trillion tokens from 20 million hours of real-world human interactions, environment, industrial, robotics, and driving data. This development is expected to democratize physical AI and make general robotics accessible to every developer.
Cosmos helps developers build custom world models for physical AI systems at scale, offering open world foundation models and tools for every stage of development, from data curation to training to customization.
The Cosmos World Foundation suite of open models provide developers with a simplified approach to generate massive volumes of photorealistic, physics-based synthetic data, enabling them to train and evaluate their existing models more efficiently.
The platform includes state-of-the-art world foundation models, video tokenizers, and AI-accelerated data processing pipelines, allowing developers to accelerate world model development by fine-tuning Cosmos world foundation models or building new ones from the ground up.
Cosmos WFMs can be used for physics-based simulation and synthetic data generation, physical AI model development and evaluation, and foresight and “multiverse” simulation, and are being adopted by leading robotics and automotive companies, including Uber, Waabi, and Wayve
