Nvidia (NVDA) CEO Jensen Huang took the stage at CES 2025 on Monday to showcase the company’s latest advancements in graphics technology, unveiling the new GeForce RTX 50 series, designed for both desktop and laptop gaming systems. Built on Nvidia’s Blackwell architecture, which powers the company’s AI data center chips, the RTX 50 series promises significant improvements for PC gamers and content creators alike. Along with the hardware updates, Nvidia is also incorporating its AI expertise into new software aimed at boosting gaming performance over previous generations.
The RTX 50 series lineup is split into two categories: desktop and laptop versions. The desktop cards include the RTX 5070, RTX 5070 Ti, RTX 5080, and RTX 5090, with prices starting at $549 for the RTX 5070 and climbing to $1,999 for the high-end RTX 5090. For laptops, pricing varies by manufacturer, but Nvidia estimates that laptops with RTX 50 series GPUs will begin around $1,299, while those equipped with the RTX 5090 will cost approximately $2,899.
The desktop RTX 5090 is particularly impressive, featuring a whopping 92 billion transistors and leveraging TSMC’s 4-nanometer chip process. Beyond the typical generational performance improvements, Nvidia has enhanced its DLSS (Deep Learning Super Sampling) technology to offer even more impressive graphical enhancements. DLSS 4 introduces a new feature, Multi-Frame Generation, which takes game rendering to new heights. Unlike previous versions that rendered a single game scene before generating additional frames, Multi-Frame Generation allows the system to generate three additional scenes, improving frame rates up to 8 times compared to standard rendering. As a result, the RTX 5090 is reportedly twice as fast as the RTX 4090 when using DLSS 4. Additionally, Nvidia claims the RTX 50 series can extend battery life by up to 40% in thin and light laptops without sacrificing performance.
During his keynote, Huang also unveiled a new AI superchip, the GB10, a compact version of the company’s powerful GB200 platform. Unlike the GB200, which combines two Blackwell GPUs and a Grace CPU, the GB10 pairs a single Blackwell GPU with a Grace CPU. Huang confirmed that the GB10 is in full production and will be available around May. While it’s not designed as a data center platform, the GB10 will be featured in a new desktop system called Project DIGITS, which combines the chip with 128GB of unified memory and 4TB of storage to create a powerful AI computing system for personal use.
In addition to hardware announcements, Nvidia showcased its Cosmos platform for developing AI-powered physical systems, such as robots and self-driving cars. Cosmos utilizes world foundation models (WFMs) to simulate real-world conditions, allowing developers to test AI software in a virtual environment without relying on expensive physical prototypes. Nvidia also introduced Isaac GROOT Blueprint, a software that developers can use with Apple’s Vision Pro headset to teach humanoid robots to move and react in different scenarios.
Nvidia also shared that leading companies in the automotive industry, including Toyota, Continental, and Aurora, are using Nvidia's AI technologies to enhance their advanced and autonomous driving systems. Huang predicts that the autonomous vehicle industry will likely evolve into a multi-trillion-dollar robotics sector.
.jpeg)
0 Comments