What is ray tracing and why does it cost so much performance?

Ray tracing simulates how light actually behaves — tracing the path of individual light rays as they travel, bounce off surfaces, refract through glass and eventually reach a virtual camera. It is physically grounded, which is why it produces accurate reflections, soft shadows and realistic indirect lighting almost for free once you have it.

What it replaces is decades of clever approximation. Traditional rasterisation does not simulate light at all; it projects geometry onto the screen and then applies techniques that fake lighting convincingly. Screen-space reflections, for instance, can only reflect what is already visible on screen — which is why reflections in older games vanish when the reflected object moves off the edge. Shadow maps produce hard, aliased edges. Baked lighting looks superb but cannot change.

Why the cost is so high: a single pixel may need many rays, each testing against scene geometry, and each bounce multiplies the work. A scene with millions of triangles at 4K and 60fps implies an enormous number of ray-geometry intersection tests every second. Modern GPUs include dedicated hardware to accelerate exactly this, which is why ray tracing is practical at all — but it remains one of the most expensive things a game can do.

This is also why upscaling arrived alongside it. Rendering at a lower internal resolution and reconstructing the image recovers much of the cost, which is why ray tracing and DLSS/FSR are usually discussed together.

Levels of implementation vary enormously. Ray-traced shadows only, or reflections only, cost far less than full path tracing, which simulates global illumination throughout the scene. Many games advertising ray tracing use it selectively.

Whether it is worth it depends heavily on the game — it transforms scenes with water, glass and complex indirect light, and is barely noticeable in others.