Artificial intelligence has been the driving force behind NVIDIA's growth for several years now... and there's no end in sight!
As the undisputed leader of the artificial intelligence revolution, thanks to its calculators that are so far ahead of the competition, NVIDIA is, in a way, calling the shots. In video games, it's perhaps even clearer with the deployment of AI-assisted technologies where, once again, NVIDIA tools are ahead of the game. While AMD offers FidelityFX Super Resolution (FSR) and Intel Xe Super Sampling (XeSS), it's NVIDIA that takes the cake with the oldest and most advanced AI-assisted visual enhancement technology, Deep Learning Super Sampling (DLSS).
NVIDIA began by asking artificial intelligence to "simply" work on enlarging the image by means of what is known as upscaling: the image is rendered by the GPU in Full HD (1080p), for example, and DLSS, using powerful algorithms, takes care of imagining the missing pixels to reach QHD (1440p). This first step was quickly replaced by image generation, which uses artificial intelligence to create an intermediate image between two images rendered by the GPU. As a result, the number of frames per second in video games is almost doubled, and fluidity is much improved. DLSS 4 has gone even further with multiple frame generation, which can now insert up to three frames between two rendered by the GPU.
NVIDIA now likes to say that on DLSS 4 games, artificial intelligence generates 15 out of 16 pixels, greatly simplifying the GPU's task.
But that's still not enough for NVIDIA, which took advantage of the Hot Chips 2025 event held from August 24 to 26 at Stanford University, California, to unveil its goal: that artificial intelligence be entrusted with the entire graphics rendering of video games. The idea is to build on the Blackwell architecture used in the GeForce RTX 50 series to power DLSS 4, and to push the envelope even further. For example, NVIDIA insists on the need for ever greater memory bandwidth, whereas the RTX 5090 can already count on 1,792 GB/s. NVIDIA is also seeking to maximize the benefits of the FP4 precision level used on Blackwell for the first time in the graphics card world. Last but not least, NVIDIA is focusing on energy efficiency, because it's not just about performance, it's also about delivering more while consuming less, so that notebook GPUs can keep up.
