New free update from Nvidia promises to accelerate generative AI on RTX graphics cards by up to 40%

Hardware technology giant Nvidia has announced a software update that directly benefits owners of its RTX series graphics cards. The novelty focuses on optimizing the performance of generative artificial intelligence tasks, making the use of complex language models faster and more accessible on personal computers. The update, which is free, promises a significant leap in performance for users and developers.
The main highlight of the update is the acceleration of up to 40% in the execution of large language models (LLMs), such as those from the GPT and Nemotron family. Essa improvement is the result of optimizations in the company’s TensorRT-LLM library, which now enables more efficient processing of AI tasks directly on user hardware, without relying exclusively on cloud services or specialized server hardware.
With the new update, users and developers who use their RTX GPUs to perform AI tasks locally will notice a considerable reduction in response time and an increase in processing power. Isso opens doors for content creation and experimentation with more advanced artificial intelligence models, even on consumer hardware configurations, cementing the PC as an increasingly robust platform for the AI era.
Optimization Boosts LLM Performance
The core of this performance improvement is the latest version of the TensorRT-LLM library. Nvidia demonstrated in internal testing that popular open source models such as Nemotron Nano V2 showed a speed gain of up to 40% compared to previous versions of the software. Esse advancement makes interacting with chatbots and other AI tools more fluid and efficient.
This acceleration is crucial for professionals who rely on LLMs for daily tasks, such as programmers, researchers, and content creators. The ability to run these models faster on a local machine not only increases productivity, but also ensures greater privacy and control over the data processed, an important differentiator compared to cloud-based solutions.
New technology to reduce VRAM usage
Another key innovation introduced by Nvidia is support for the high-precision NVFP4 format. Essa technology enables intelligent compression of AI models, resulting in a reduction of up to 60% in video memory (VRAM) consumption. In practice, this means that larger, more complex models can run on graphics cards with smaller VRAM capacity.
Optimizing memory usage is one of the biggest challenges for running AI locally. With the NVFP4 format, the barrier to entry for users with mid-range hardware decreases considerably, democratizing access to cutting-edge artificial intelligence tools that were previously restricted to systems with large amounts of dedicated memory.
This improvement is implemented through a technique that offloads part of the data to the system’s RAM memory when necessary, freeing up VRAM for more critical operations. Testes with the ComfyUI tool showed that the performance of pipelines using models like Flux.1 and Flux.2 can increase by up to 4.6 times, allowing high-quality images to be generated much faster.
Advances in AI video generation
The Nvidia update is not limited to language models. The company also improved video generation by optimizing the LTX-2 model, developed in collaboration with Lightricks. The tool, which converts text to video, is now capable of generating clips with resolutions of up to 4K at 50 frames per second.
Thanks to support for the NVFP8 format, the LTX-2 model achieves twice the generation speed of the latest RTX cards. Isso means that a high-quality video that previously took around 20 seconds to create can now be generated in less than half the time, a significant advancement for visual content creators.
This technology facilitates the production of creative and dynamic content, allowing users to transform textual ideas into detailed videos faster. The tool offers precise control over the final result, including defining characters, styles and camera movements, all from simple text commands.
The increased speed and ability to generate 4K content position RTX cards as a powerful and affordable solution for AI-assisted video production, opening up new possibilities for digital marketing, education and entertainment.
Improved image quality with RTX VSR
The Nvidia has expanded the functionality of its RTX Video Super Resolution (VSR) technology, which uses AI to improve the quality of videos in real time. Agora, the tool can also be applied to content generated by artificial intelligence, such as that created by the LTX-2 model. Isso allows users to output a video at a lower resolution, such as 720p, and convert it to 4K with significantly higher image quality, with greater sharpness and improved details.
This functionality is especially useful for speeding up the workflow of content creators. By generating clips at lower resolutions, production time is drastically reduced. Posteriormente, VSR takes care of raising quality to professional standards without the need for long rendering processes. Integration with the ComfyUI plugin, for example, reduced the time to generate and enhance a 4K clip from 15 minutes to just 3 minutes, demonstrating the efficiency of the combined solution.
Integration with the developer ecosystem
One of the most important aspects of this update is its deep integration with popular tools in the generative AI scenario, with emphasis on ComfyUI. Este workspace, based on nodes, is one of the favorites among enthusiasts and professionals for building complex AI workflows. By providing direct support and optimizations for these platforms, Nvidia ensures that its innovations quickly reach the hands of the community, who can then explore and expand the capabilities of the hardware. The company is actively collaborating with the developers of these tools to ensure continuous and efficient implementation of new technologies, such as support for the NVFP4 format in models like the Flux. Essa collaboration strengthens the ecosystem around RTX GPUs, cementing them not just as components for gaming, but as the primary platform for developing and running consumer-grade AI. Nvidia’s strategy is clear: leverage the power of its Tensor Cores and the maturity of its software stack to offer a more versatile and powerful solution than the dedicated Unidades and Processamento Neural (NPUs) that are beginning to appear in competing processors. Isso ensures that a wide audience, from hobbyists to professionals, can immediately benefit from the latest advances in AI performance.
The future of local AI on RTX cards
With this series of updates, Nvidia reinforces its strategy of positioning RTX graphics cards at the center of the local artificial intelligence revolution. By delivering performance gains, reducing memory requirements and accelerating content creation, the company empowers a growing user base to explore, create and innovate with the most advanced AI technologies directly on their own computers.







