Dealing with shadows in a grid is anything but intuitive and requires some logistics: the scene must be rendered from the perspective of each individual light, stored in a texture and projected onto itself during the lighting phase. Worse still, it is far from being even close to acceptable in terms of quality: These shadows are very susceptible to aliasing (because a pixel seen by the light does not match a pixel seen by the camera) or acne (because a shadow image texel stores a single depth value but can cover a large area). In addition, most rasters need to support special shadow map types such as cubic shadows or cascaded shadow maps, which greatly increases the complexity of the renderer.
In a raytracer, a single code path can handle all shading scenarios. More importantly, the shadowing process is as simple and intuitive as firing a ray from the surface to the light source and checking if anything was hit along the way. The PowerVR ray tracing architecture exposes so-called fast feeler rays, which only check the presence of geometry along the ray, making them particularly suitable for efficient shadow rendering.