NVIDIA Announces RTX 50-Series GPUs and Cosmos Platform
Created on January 7|Last edited on January 7
Comment
Nvidia made a powerful impression at CES 2025 with two significant announcements: the unveiling of the RTX 50-series GPUs and the launch of the Cosmos platform for physical AI development. These advancements highlight Nvidia's commitment to pushing the boundaries of graphics and AI technologies.
RTX 50-Series GPUs Redefine Performance
Nvidia introduced four new GPUs in its RTX 50-series lineup: the RTX 5090, RTX 5080, RTX 5070 Ti, and RTX 5070. Powered by the Blackwell architecture, these GPUs promise substantial performance leaps, with the flagship RTX 5090 delivering twice the performance of the RTX 4090.
The RTX 5090 offers 32GB of GDDR7 memory, an unprecedented memory bandwidth of 1,792GB/sec, and a staggering 21,760 CUDA cores. Nvidia claims this GPU delivers twice the performance of its predecessor, the RTX 4090, and showcased Cyberpunk 2077 running with DLSS 4 at 238fps compared to 106fps on the RTX 4090. However, this comes with a total graphics power of 575 watts, necessitating a robust 1000-watt power supply.
The RTX 5080, priced at $999, boasts 16GB of GDDR7 memory, 10,752 CUDA cores, and a memory bandwidth of 960GB/sec. It promises twice the performance of the RTX 4080 while consuming 360 watts of power.
Lower-tier models, the RTX 5070 Ti and RTX 5070, bring remarkable capabilities to the midrange market. Priced at $749 and $549, respectively, these GPUs include GDDR7 memory and significant performance enhancements over their RTX 40-series counterparts.
Nvidia Cosmos: A Platform for Physical AI
Nvidia also unveiled the Cosmos platform, a suite of tools designed to accelerate the development of robots and autonomous vehicles. Cosmos includes generative world foundation models, advanced video tokenizers, and an accelerated data processing pipeline, all optimized for Nvidia’s data center GPUs.
The Cosmos platform allows developers to generate photorealistic, physics-based synthetic data, significantly reducing the cost and time required for real-world testing. By offering these models under an open license, Nvidia aims to democratize physical AI development, enabling researchers and developers to create custom models for diverse applications.
Applications and Adoption
Cosmos is being adopted by global leaders in robotics and autonomous systems, including 1X, XPENG, and Uber, to enhance model training, simulation, and deployment. Features like video search, synthetic data generation, and multiverse simulation showcase the platform’s versatility in addressing challenges like data scarcity and variability.
Technical Innovations
The Cosmos platform integrates state-of-the-art tools like the Nvidia NeMo Curator, capable of processing 20 million hours of video data in just 14 days, and the Cosmos Tokenizer, which improves data compression and processing speeds. These innovations promise to accelerate AI model development across industries.
Availability and Future Prospects
The RTX 50-series GPUs will begin rolling out on January 30, 2025, while Cosmos WFMs are available under Nvidia’s open model license. With these announcements, Nvidia continues to shape the future of computing, blending cutting-edge graphics with transformative AI technologies.
Sources:
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.