PlayStation 6 and Project Helix to use DLSS 5-like technology - rumor

AMD's own equivalent, but even better.

A rumor has emerged that Sony and AMD may be preparing a neural rendering technology for the next generation of PlayStation and Project Helix, capable of operating on a principle similar to NVIDIA's DLSS 5.

According to the leak, future consoles will use AMD RDNA 5 architecture, and the neural rendering technology will be part of the conditional FSR 5 family. The key difference from modern image reconstruction methods is allegedly where data processing occurs.

Instead of analyzing an already rendered 2D frame, the system will be able to directly access the internal data of the game engine. The neural network will have access to raw depth buffers, material properties, and precise coordinates of object geometry. Thanks to this, the algorithm will work with the original scene data, rather than trying to restore information solely from already rendered pixels.

This approach is allegedly related to recent AMD research in single-step diffusion models for global illumination. The use of real spatial coordinates should allow the system to more accurately determine the interaction of light with objects, while reducing the number of artifacts that arise with traditional image reconstruction and motion handling.

Another potential advantage is the preservation of the developers' original visual intent. Since the algorithm receives information directly from the game engine, it allegedly does not have to "guess" the characteristics of materials, geometry, and lighting based on a finished image. This can also reduce computational costs for complex 2D image analysis and help maintain a more stable frame rate.