CES 2025: The new generation of Nvidia DLSS 4 is here, what can it do?

Surely you could not have missed that at the CES, Nvidia presented new graphics cards of the Blackwell architecture and with it the next generation of DLSS 4 technology. It differs from the previous version mainly in the ability to generate multiple images at once, while CEO Jensen Huang describes this technology as being able to “predict the future ”, not just counting the gaps between frames. The new technology will thus help smoother imaging.

DLSS 4 uses data from motion vectors and previous frames to create multiple new frames, reducing latency and improving frame synchronization. But we will see how the new technology copes with the unfortunate problem of the previous generation DLSS3. There, when generating lower FPS (less than 40), mouse or keyboard movements were processed at half the frequency, and the whole thing seemed rather delayed.

Another interesting novelty is the so-called neural materials, i.e. textures created using artificial intelligence, which significantly save graphics memory. According to Jensen Huang, this technology can reduce the size of textures from 48MB to just 16MB. However, in order to use their full potential, developers must adapt their games to them. The function is apparently exclusive to the latest generation of GPUs, older graphics won’t have enough power to use it.

In addition to DLSS 4, Nvidia also introduced Reflex 2 technology, which further reduces the delay between the player’s input and the screen. The processor predicts the next movement of the camera and it is then projected into the current image. A special algorithm can fill in the missing details to make everything look smooth and natural. This innovation can shorten the response time by up to 75% compared to normal display. Reflex 2 will first be available for the new RTX 50 series, later also for other RTX cards.