Ray tracing is a way of rendering light by simulating individual light rays as they bounce around a scene, instead of faking shadows and reflections with pre-baked tricks. It needs special hardware because checking where millions of rays hit objects, every single frame, is a math-heavy job that ordinary graphics chips were never built to do quickly.
§01What is ray tracing, actually?
Most games do not simulate light. They fake it.
For decades, developers used a technique called rasterization. It draws triangles onto the screen and applies shortcuts to make lighting look plausible. Shadows come from pre-calculated maps. Reflections often come from a flat, mirror-like trick called a cubemap, or a screen-space hack that breaks down the moment something moves off-screen.
Ray tracing works differently. It traces the actual path light would take, bouncing rays from the camera (or from light sources) around a 3D scene, checking what each ray hits, and calculating color, shadow, and reflection based on that real geometry.
The result looks more physically correct. Reflections show things that are technically off-screen. Shadows have the right softness based on distance from a light. Global illumination — light that bounces off one surface and colors another — happens naturally instead of being painted in by an artist ahead of time.
This is the same basic idea used in offline movie rendering for years. Pixar-style studios have used ray tracing and its more advanced cousin, path tracing, to render animated films for a long time. The difference is that a film render can take hours per frame. A game needs a new frame roughly sixty times a second.
§02Why can’t a regular graphics card just do this?
A regular GPU can do it. It’s just painfully slow at it, because tracing rays means solving a geometry problem for every ray, every frame, at massive scale.
Rasterization is a fairly predictable job. The GPU knows which triangles are on screen and processes them in a fairly linear, brute-force way. That’s exactly what traditional GPU cores are optimized for.
Ray tracing is a different kind of problem. For every ray, the hardware has to figure out which of potentially millions of triangles that ray intersects. Done naively, that means checking a ray against every triangle in the scene, which is far too slow even for powerful chips.
The workaround is a data structure called a bounding volume hierarchy, or BVH. Think of it as a tree of boxes that get smaller and smaller, letting the hardware quickly rule out huge chunks of the scene a ray obviously doesn’t touch, before it ever tests individual triangles. Building and traversing that structure, over and over, for millions of rays a frame, is the actual computational bottleneck. It’s a specific enough job that chipmakers started building dedicated circuits for it.
§03What do RT cores and Tensor cores actually do?
RT cores handle the ray-triangle intersection math directly in hardware, instead of running it as general-purpose shader code. Tensor-style AI cores then clean up the noisy, incomplete image that real-time ray tracing produces, using a denoising and upscaling pass.
Here’s why that second part matters. A game cannot afford to trace enough rays per pixel to get a smooth, noise-free image in real time — that level of detail is still a job for offline movie rendering. So real-time ray tracing traces a handful of rays per pixel and produces a grainy, incomplete result.
A denoising algorithm then fills in the gaps, using information from previous frames and neighboring pixels to guess what the “true” clean image should look like. This is a genuinely hard, specialized computation, and it’s a large part of why Nvidia’s RTX cards ship with separate Tensor cores rather than leaning purely on the RT cores.
AMD and Intel take slightly different architectural approaches, but the underlying split is the same: dedicated intersection hardware plus a fast reconstruction/denoising pipeline. Without both pieces, real-time ray tracing at a playable frame rate isn’t realistic on today’s chips. This is also the reason ray tracing arrived on GPUs years before it arrived meaningfully on phones or budget laptops — the silicon has to be designed for it specifically, not just made generally faster. For more on how that graphics silicon differs from the integrated graphics found in cheaper machines, see our explainer on the difference between a graphics card and integrated graphics.
§04Which ray-traced effects are actually worth turning on?
This is the part the marketing term “RTX On” deliberately blurs. Not all ray-traced effects cost the same. Shadows are relatively cheap. Full scene-wide global illumination and path tracing can cost roughly ten times more GPU work for a comparable jump in visual fidelity. Toggling one setting called “ray tracing” hides a huge range of actual demand.
Each effect stresses the RT cores differently, because each one requires tracing a different number of rays, at a different frequency, across the scene.
- Ray-traced shadows only need to check whether a straight line from a surface to a light source is blocked. That’s a relatively simple yes/no calculation, so it’s the cheapest ray-traced effect to run.
- Ray-traced reflections need to trace a ray out from every visible surface, figure out what it hits, and shade that too. Cost climbs fast with how many reflective surfaces are on screen and how rough or glossy those surfaces are.
- Ray-traced global illumination (sometimes labeled RTGI or just “GI”) traces bounced light across the whole visible scene, continuously, to work out ambient color and brightness. This runs constantly rather than only where a puddle or window happens to be, so it’s a heavier, steadier load.
- Full path tracing traces many bounces per pixel for shadows, reflections, and lighting all at once, replacing almost the entire traditional rendering pipeline. It is the most accurate-looking option and by far the most expensive.
The table below is a rough, qualitative guide to matching effects to hardware tiers, based on how each effect scales rather than on any specific benchmark numbers.
| Effect | What it changes | Relative cost | Sensible tier to enable it |
|---|---|---|---|
| Ray-traced shadows | Softer, more geometrically correct shadow edges | Low | Entry-level and mid-range RT-capable cards |
| Ray-traced reflections | Accurate reflections on water, glass, metal, wet surfaces | Medium to high, depends on scene reflectivity | Mid-range and up; entry-level cards may need upscaling turned on too |
| Ray-traced global illumination | Natural bounced light and ambient color across the whole scene | High, runs continuously | Upper mid-range and flagship cards |
| Full path tracing | Replaces most of the traditional lighting pipeline with simulated light | Very high | Flagship cards, usually paired with upscaling and frame generation |
A practical decision rule follows from this. On an entry-level RT-capable card, shadows are usually the one setting worth switching on; reflections and GI will often cost more frame rate than the visual gain justifies. On a mid-range card, shadows plus reflections is a reasonable combination in most games, with GI added only if the frame rate holds up. On a flagship card, all three individually traced effects are usually fine, and path tracing becomes worth trying, especially with upscaling doing some of the heavy lifting.
The safest approach on any card that isn’t a current flagship: enable one ray-traced effect at a time and judge the frame rate hit before adding the next, rather than flipping a single “RTX On” toggle that quietly stacks all of them together. Our guide on which gaming PC parts matter most for the money covers where a GPU upgrade fits into that budget more broadly.
§05Do consoles actually have ray tracing hardware?
Yes. Current-generation consoles include dedicated ray tracing hardware built into their custom GPUs, but at a more modest level than a high-end gaming PC’s flagship card.
That’s why console games tend to use ray tracing selectively — often just for reflections in a specific “quality mode,” or for shadows in a handful of scenes — rather than running full path tracing. Developers pick the one or two effects that give the biggest visual return for the hardware budget they have, which lines up with the same cost logic in the table above.
It also explains why many console games offer a choice between a performance mode and a quality mode with ray tracing enabled. That’s the console-level version of the same trade-off PC players make manually: ray tracing buys a more accurate image at the cost of frame rate, and someone has to decide where that trade is worth making. For more on how consoles differ from PCs day to day, see our gaming hardware guide.
§06Why does ray tracing hit frame rate so much harder than other settings?
Turning up texture resolution or shadow map quality mostly costs memory and a modest amount of extra processing. Ray tracing costs processing on almost every frame, at every pixel it touches, because it’s solving a geometry search problem rather than just drawing more detailed pixels.
That’s a structurally different kind of workload. It’s also why upscaling technologies became almost inseparable from ray tracing marketing.
Rendering the game at a lower internal resolution and using an AI model to reconstruct a sharper final image gives back a meaningful chunk of the frame rate that ray tracing takes away. Our explainer on what upscaling is and why every game uses it now goes into how that reconstruction actually works. Frame generation, which inserts synthetic in-between frames, is often bundled alongside it on flagship cards to make path tracing feel smoother than the raw rendering workload would otherwise allow.
§07Is path tracing worth chasing right now?
For most players, no — not yet, unless a flagship-tier card is already in the budget for other reasons. Path tracing looks genuinely different from traditional lighting, but it’s the single most demanding setting in modern games, and it usually requires upscaling and frame generation just to stay smooth.
A handful of newer titles have added path-traced lighting modes as a showcase feature, and they’re a good way to see the ceiling of what real-time ray tracing can do today. Treat those modes as a demonstration of where the technology is heading rather than a baseline expectation for every game going forward.
Mid-range hardware will likely catch up to today’s flagship path tracing performance over time, the same way mid-range cards eventually caught up to basic ray-traced shadows and reflections. That’s simply how each hardware generation trickles down.
§08Does ray tracing matter more than a higher frame rate or resolution?
For most people, a stable frame rate matters more day to day than ray-traced lighting, especially in fast, competitive, or twitchy games. Ray tracing is easiest to appreciate in slower, atmospheric, or exploration-focused games where players actually stop to look at reflections and shadows.
This is a judgment call, not a settled fact, and reasonable players land in different places on it. Our piece on why frame rate matters more than resolution covers the broader version of this trade-off, which the table above is really just a ray-tracing-specific application of.
Anyone shopping for a new GPU or laptop with this feature in mind should read pricing and marketing carefully. “Ray tracing support” on a box just means the hardware has RT cores of some kind, not that it can run every effect at a playable frame rate. Our guide to buying a gaming laptop without overpaying and the broader buying advice archive both cover how to read that kind of spec sheet with a skeptical eye.
§09The short version
- Ray tracing simulates real light paths instead of faking lighting with pre-baked shortcuts.
- It needs dedicated hardware because tracing millions of rays per frame is a specialized, expensive geometry search, not a normal drawing task.
- Shadows are the cheapest ray-traced effect; reflections cost more; global illumination and path tracing cost the most.
- “RTX On” is a marketing label that can mean a small performance hit or a massive one, depending on which effects it actually enables.
- Matching the effect to the GPU tier, one setting at a time, gets better results than trusting a single toggle.
Readers building a broader picture of how modern GPUs and displays fit together can browse the rest of the consoles and hardware archive or the PC gaming and setup archive for related pieces on monitors, upscaling, and what actually drives performance.