AMD's upcoming FSR Redstone update, which is not the final name, is set to bring FSR 4's suite of technologies in line with NVIDIA's DLSS. FSR Redstone features the new AI-powered Super Resolution technology, currently exclusive to the RDNA 4 generation of Radeon RX 9000 Series GPUs, alongside AI-powered Frame Generation, Ray Regeneration, and Neural Radiance Caching.

It will open the door to the visually impressive path-tracing and ray-tracing that is currently limited to the GeForce RTX Series and DLSS 4. Due to the AI component, like FSR 4, the expectation and assumption for FSR Redstone is that it would be exclusive to AMD's RDNA 4 GPUs; however, according to a new interview with AMD's Senior Director of Software Development, Chris Hall, FSR Redstone has been designed to work with previous-gen Radeon GPUs and even non-AMD GPUs like GeForce RTX hardware.
Per the interview (machine translated via 4Gamer.net), the reason for this is that FSR Redstone has been developed using ROCm "AMD ML2CODE" (Machine Learning to Code), which converts the machine learning or neural rendering technology to "optimized Compute Shader code" so it can run on GPUs without advanced AI hardware or even third-party GPUs.
"The biggest benefit is that you can seamlessly integrate your work with ML2CODE like FSR Redstone directly into your DirectX and Vulkan graphics pipelines with minimal latency," Chris Hall explains. "We believe that ML2CODE solution is the best way to integrate and deploy 3D graphics technology and AI technology, at least for now."
Essentially, optimized code for each generation of Radeon GPU hardware will be created, enabling FSR Redstone to be available on cards such as the Radeon RX 7900 XTX and 7800 XT. Likewise, if the code is converted to CUDA, it will be able to run on GeForce RTX GPUs without requiring AI hardware at runtime. Naturally, adding anything to the render and graphics pipeline adds latency, but AMD's Chris Hall notes that this will be minimal.




