CIOInsider India Magazine

“I don't wear a watch, something that most people are unaware of. The explanation is straightforward: the here and now is what counts most. It may come as a surprise, but I don't have the ambition to do more. I let the world come to me; my objective is to be the best at what I'm doing now, not to chase anything else,” says Nvidia’s CEO Jensen Huang. The Taiwanese and American businessman unveiled the latest line of Nvidia game plans, chips, AI technologies to train robots and self-driving cars, its first desktop computer, and much more at CES 2025, in Las Vegas, Nevada, USA.

Let’s know more about all the new GPU hardware and AI software introduced by Nvidia at CES 2025.

RTX 50-series GPUs
At CES 202, Nvidia formally unveiled the RTX-50 series, their newest GPU lineup. The company's flagship Blackwell architecture will power the RTX GPUs of the future. Its Ada Lovelace design served as the foundation for the RTX 40 series of GPUs.

The RTX 5070, RTX 5070 Ti, RTX 5080, and RTX 5090 are the four next-generation GPUs that Jensen introduced; their prices range from $549 to $1,999. The remaining two GPUs will be released in February of this year, when the RTX 5090 and RTX 5080 GPUs become available on January 30.

The most recent GPU series from Nvidia features a redesigned design with GDDR7 memory, a 3D vapor chamber, and two double flow-through fans. To support monitors up to 8K and 165Hz, all four of the graphic cards include DisplayPort 2.1b ports. According to reports, the RTX 5090 GPU can fit within a small-form-factor PC because of its much smaller size compared to its predecessor, the RTX 4090. Nonetheless, the RTX 5090's overall graphics power (575 watts) is 125 watts higher than the RTX 4090's. Additionally, Nvidia's RTX 50 series of GPUs will be included in laptops, which are anticipated to be released by several PC manufacturers starting in March.
Jensen showcased the potential of the RTX-50 GPU family during his speech by demonstrating important features including text-to-animation, RTX Neural Materials, RTX Neural Faces, and even DLSS 4 in real-time rendering.

DLSS 4 with Multi-Frame Generation
Nvidia declared that, on RTX-50 series computers, it is enhancing its Deep Learning Super Sampling (DLSS) technology with additional neural rendering features like Multi Frame generation. Working in tandem with the whole suite of DLSS technologies, it will provide up to three additional frames per traditionally rendered frame, multiplying frame rates by up to 8X over traditional brute-force rendering. Existing RTX GPUs will also be compatible with the DLSS 4 update.

Cosmos World Foundation Models
To train robots and autonomous cars, Nvidia has unveiled Cosmos, a new collection of AI models that can produce photo-realistic films. For example, users will be able to instruct Cosmos to produce a film of a car being driven in a physics-based environment.

It is anticipated that the artificially created data will assist robotics firms and AV manufacturers in reducing the expenses associated with collecting traditional, real-world data. Cosmos will be made available by Nvidia under an open license.

Jensen says, "We sincerely hope that (Cosmos) will accomplish for the robotics and industrial AI fields what Llama 3 has accomplished for enterprise AI."

According to reports, the Japanese automaker plans to include advanced driver assistance into certain of its car models using its Orin AI chips and automotive operating system.

Project Digits
Project Digits, Nvidia's first desktop computer, was unveiled. It is intended for students, data scientists, and AI researchers. Operating on a Linux-based operating system, it costs $3,000 and has a new GB10 Grace Blackwell super chip that provides computing power for executing and optimizing AI models with up to 200 billion parameters.

The whole Nvidia AI stack is operated by Project Digits; this is the platform that powers all Nvidia applications. It's a desktop-based cloud computing platform. In March 2025, Project Digits will be accessible.

Llama Nemotron Family of Models
A new series of open-source large language models (LLMs) developed with Meta's seminal AI model Llama was unveiled by Nvidia. Consequently, the new model series from the chip company has been dubbed Llama Nemotron. Businesses can utilize it to develop and implement AI agents for a variety of uses, including fraud detection, customer service, and the optimization of product supply chains and inventory management. Nvidia's most recent pruning techniques were used to train Llama Nemotron models on high-quality datasets. They will be offered in Nano, Super, and Ultra sizes to facilitate the large-scale deployment of AI agents by enterprises. The three sizes differ mostly in terms of performance, accuracy, and latency.

Nvidia Mega
Mega is the latest robotic fleet management software for warehouses from Nvidia. It is an ecosystem that may be utilized to ensure that the robots in a given warehouse cooperate with one another. Mega can be used to control a variety of robots, including humanoids, autonomous forklifts, robotic arms, and AMRs.

To handle immense complexity and scale, Nvidia stated that Mega provides enterprises with reference architecture of Nvidia accelerated computing, AI, Nvidia Isaac, and Nvidia Omniverse technologies to develop and test digital twins for testing AI-powered robot brains that drive robots, video analytics AI agents, equipment, and more.

The new framework enables continuous development, testing, optimization, and deployment by bringing software-defined capabilities to physical facilities. According to reports, Kion Group, a German supply chain organization, has implemented Mega to streamline its operations.

Autonomous, AI Game Characters
In 2023, Nvidia unveiled Ace, a package of RTX-powered technologies that uses generative AI to produce gaming characters. The business is now expanding on its Ace capabilities to assist with the development of AI-powered autonomous game characters that can see, plan, and behave similarly to human players.

A collection of multi-modal, tiny language models will power these self-governing gaming avatars, enabling them to hear audio cues and perceive their environment.

According to Nvidia, it has teamed up with game developers to incorporate these self-governing game characters into well-known games like Pubg: Battlegrounds, inZOI, and the PC mobile edition of Naraka: Bladepoint. Additionally, the firm revealed that GeForce RTX customers would be able to access its AI virtual assistant, G-Assist, through the Nvidia app starting in February.