Basically AMD hardware is less powerful and less able to deal with more divergent rays. The RT GPU performance is more important at this point than the raster performance. If you look at Unreal Engine 5 most of the raster processing is done by the engine in software because no GPU hardware can match the performance. Thus a fast raster proformance on the gpu is less important. If you turn of RT then the performance of the GPU is very important. NVidia cards can provide better DXR performance at higher quality settings and get higher frame times. In every game nvidia is faster. AMD design a card that is faster at current games in raster but left DXR performance low. NVidia gives good raster perfomance and far fastest DXR performance.
This is why DLSS exists, to make DXR games able to run at high resolutions. DLSS 1 used sparial upscaling, an AI netwark and sharpening. This is better than just using sparial upscaling and sharpening for image quality. The media attacked DLSS 1's image quality. We had zoomed in image and pixel by pixel criticism of every deviation from native. So to get the image quality nvidia move to temporal upscaling in DLSS 2.x. Temporal upscaling was the breakthrough for image quality in DLSS 2.x.
AMD is forced to create FSR. FSR uses the same method nvidia rejected for image quality reasons in DLSS 1.x. There is no AI network. You can see the the image is not native, lacks detail and is blurred compared to native. All the problems of DLSS 1. At this point DLSS 2.2 is being called witchcraft its so good at increasing performance and keeping image quality.
There has been no attack on FSR image quaility in the same way as DLSS 1. FSR is blurred and there is less deatil in the image. Youtube posts on a video talking about how good FSR is. This is from a gamer using FSR. They tell the truth.
The massive performance lead in DXR is being ignored. FSR image quality problems are not being addressed. DLSS being ignored for benchmarks results.