Posts: 1,307 +2,094
Your knowledge is so massive that it's just infuriatingThere's a good reason as to why all units dedicated to ray tracing algorithms (ray-triangle intersection calculations, BVH traversal acceleration) are located within the primary 'shader' sections of the GPU -- I.e. AMD's Compute Units, Intel's Render Slice, Nvidia's Streaming Multiprocessor. They need direct access to the fastest cache within the GPU and that's located with the shader units. Putting all that on a separate die would significantly lower RT performance; putting it on a separate card would kill it all together.