A few days ago, AMD launched FSR 3.1, an important update to its well-known platform that focused on improving image quality and stability through scaling and also decoupled it from image generation to be compatible with other technologies such as NVIDIA DLSS and Intel XeSS. AMD plays well, but the most interesting will be FSR 4.

AMD hasn’t said anything yet about the new features it might include in FSR 4, but seeing the evolution NVIDIA has followed with DLSS and the quality they’ve achieved in both intelligent rescaling and image generation, and more recently with the launch of beam reconstruction , AI-enhanced noise reduction, it’s quite clear to me that AMD needs to make a leap at some point and integrate artificial intelligence into FSR 4.

AMD themselves are aware of this, in fact Mark Papermaster, AMD’s CTO, was open about this last March when he said that they were working on making their devices possible for gaming and scale with AI. He also emphasized that 2024 is a great year for the implementation of artificial intelligence, a statement of intent that leaves no room for doubt, although targets are one thing and execution times are another.

AMD has everything it needs to make FSR 4 smart

When AMD launched the Radeon RX 6000 they only had ray tracing acceleration cores and that was pretty modest. Its performance was clearly lower than that of NVIDIA’s first generation ray tracing cores, a situation has improved significantly with the Radeon RX 7000which bring second-generation ray tracing cores with much higher performance.

With the Radeon RX 7000, they also introduced first-generation AI cores, hardware that is again very modest in terms of performance and clearly placed behind NVIDIA’s tensor cores. The most logical thing is that this will change with the Radeon RX 8000 and its cores for second generation AI. These could improve performance a lot and would be the ideal pillar on which to make the jump to the intelligent FSR 4.

Those second generation AI cores could be the basis of FSR 4. That would be it exclusive to Radeon RX 8000which may not please the user base of Radeon graphics cards, but sacrifices have to be made for progress and improvement, and this would be very necessary, because with hardware-accelerated AI, this AMD platform would be able to:

• Provide better quality intelligent scalingboth for the resulting sharpness and for preserving details and image stability.
• It would also allow you to improve the ghosting problem in the games.something that became a big problem and with FSR 3.1 it didn’t improve as much as it should have.
• It could also scale better in terms of performance thanks to acceleration through specialized hardware that would reduce the resource consumption that occurs on the GPU when applying rescaling.
• AI could also help improve the quality of image generationand it would open the door for AMD to develop new technologies to improve performance and image quality.

AI generated images.