Ray tracing is one of the big problems for modern graphics cards, and it’s usually worth 10~20% native. We’re seeing how AMD is working on this particular performance problem. AMD is creating FSR Ray Reconstruction 2.0, which will be able to reduce the computing power required for ray tracing by up to 50%.
AMD's next-generation neural network denoiser aims to halve the performance cost of ray tracing
Why is this new update significant? Because it’s dealing with the efficiency bottleneck, which means players have to make a choice between picture quality and frame rate.
FSR Ray Reconstruction 2.0 is a new generation of neural network denoiser, which can be embedded into the Redstone SDK to more efficiently handle ray tracing calculations. The new model is using machine learning to restore the image from sparse ray data. So it enables games to maintain a high picture quality at the same time as avoiding the high cost of full ray tracing.
The main purpose of this technology is to reduce the screen flickering and ghosting that are easy to appear in low samples. According to AMD’s description, the new model is using a larger sample set to improve the geometric shadows and complex reflections restoration, which can ensure the retention of environmental details and fine textures when upscale is high. This will also avoid the blurred picture commonly seen in the AI reconstruction of earlier versions.
In order to run this software, we need to meet the following system conditions. First of all, we need to be using Windows 11 (supporting DirectX12). Secondly, the GPU needs to be of Shader Model 6.6+ standard and have machine learning instruction.
We don’t know what specific graphics cards this supports. Only the hardware that supports the ML instruction is mentioned in the source.
This is a solution to the big problem of ray tracing performance consumption. It’s intended to fix the efficiency bottlenecks of real use. We are looking forward to knowing more about specific GPU support for this update. This technology is making ray tracing more affordable by reducing performance consumption.



Discussion
0 comments
Log in to join the thread with a thoughtful take, question, or correction.