VoltGround
GPU · VBIOS · Benchmarks · Thermals
← Back to Articles Analysis

Ray Tracing GPU Cost by Feature: Which RT Effects Are Worth Enabling

A game with ray tracing enabled at maximum is not the same as a game with ray tracing on medium. Individual RT sub-features have entirely different performance costs and contribute different amounts to visual quality. Knowing which effects to enable and which to skip is a meaningful frame rate decision.

Ray tracing uses the GPU's RT cores (on NVIDIA RTX hardware) or shader-based intersection testing (on other GPUs) to trace simulated light rays through the scene geometry. Different lighting effects require different numbers of rays per pixel and different traversal complexity through the scene's bounding volume hierarchy. This is why ambient occlusion and global illumination use the same hardware but have performance costs that differ by an order of magnitude.

What Each RT Feature Computes

Ray traced ambient occlusion (RTAO) fires a small number of short shadow rays from each surface point to determine how much ambient light is blocked by nearby geometry. Typical ray counts are 1 to 4 per pixel. The rays are short-range and rarely traverse the entire scene, making RTAO the cheapest RT effect.

Ray traced shadows fire longer rays from each shadow-casting light source and through the scene toward the camera. The cost scales with the number of shadow-casting lights and scene complexity. In games with many dynamic light sources, ray traced shadows can be expensive; in outdoor scenes with a single sun, the cost is more moderate.

Ray traced reflections fire rays from mirror-like and semi-glossy surfaces and traverse the full scene to find what those surfaces reflect. Complex scenes with many reflective surfaces and multi-bounce reflections are expensive. Single-bounce reflections in less complex environments have a moderate cost.

Ray traced global illumination (RTGI) is the most expensive implementation because it traces indirect light paths. A diffuse surface lit by bounced light from another diffuse surface requires tracing at minimum two bounces, and the number of rays needed for noise-free results is very high. All modern implementations denoise aggressively, but even with denoising, RTGI remains the most GPU-intensive RT feature in any game that implements it.

Performance Cost Comparison (1440p, RTX 4070)

RT Feature Typical FPS Cost vs Rasterized Visual Impact Value Tier
RT Ambient Occlusion 5–12% Subtle shadow contact in corners High value (cheap, visible improvement)
RT Shadows (single sun, outdoor) 10–18% Accurate soft shadow penumbrae Good value in daylight-heavy games
RT Shadows (multiple dynamic lights) 25–40% Correct per-light shadowing Low value if many light sources active
RT Reflections (single bounce) 15–25% Accurate scene reflections in glass and metal Good value in environments with reflective surfaces
RT Reflections (multi-bounce) 30–50% Mirror-in-mirror and complex multi-surface Poor value in most game scenes
RT Global Illumination 40–70% Indirect color bouncing, lit shadow regions High quality impact, very expensive
Path Tracing (full) 60–90% Physically accurate entire lighting model Requires upscaling to be playable on any hardware

Which RT Effects to Enable by GPU Class

With a mid-range GPU such as an RTX 4060 or RX 7600, frame rate at 1080p is already the primary concern. Enabling any RT effects will require using DLSS or FSR upscaling to recover the lost frames. In this configuration, the practical recommendation is to enable RT ambient occlusion only (lowest cost, visible improvement) and leave shadows, reflections, and global illumination off. If the game allows granular control, RTAO with DLSS Balanced is typically the best trade-off.

With a high-end GPU such as an RTX 4080 or above at 1440p, there is meaningful frame rate headroom. RTAO plus single-bounce RT reflections with DLSS Quality enabled is a reasonable starting point. Full RT or path tracing should be evaluated per game; some implementations are visually transformative (Cyberpunk 2077, Alan Wake 2), while others show minimal improvement over their rasterized equivalents at twice the GPU cost.

Where RT Quality Tiers Usually Live

The best practice is to test each RT sub-feature individually if the game exposes granular controls. Enable one effect, benchmark for two minutes, then enable the next. This reveals which effects cost the most relative to their visible improvement in your specific game and build.