What Is Ray Tracing? How It Works in Architectural Rendering

Ray tracing is a rendering technique that creates an image by sending rays from a virtual camera through each pixel into a 3D scene, finding the surface each ray hits and then tracing new rays toward the lights and in the direction of reflections and refractions. Because light, shadow and reflection come from the scene's own geometry and materials, ray-traced images look photographic, which is why architects use it for realistic renders of buildings and interiors.

This guide explains how ray tracing works step by step, how it differs from rasterization and path tracing, where it came from, which GPUs accelerate it, which architecture software uses it in real time or offline as of October 2026, and when an AI renderer is the quicker route to a client image.

What Is Ray Tracing?

In ray tracing, a ray is a straight line with a starting point and a direction. The renderer sends at least one primary ray per pixel from the camera, finds the nearest surface it hits, and from that point sends secondary rays: shadow rays toward each light to test whether the point is lit, and reflection or refraction rays for mirrors, glass, water and polished stone. The colors these rays bring back are added up into the pixel.

In architectural visualization, ray tracing is what lets a glass facade reflect the sky, a polished floor mirror the furniture and a shadow soften with the size of the light. Offline renderers such as V-Ray, Blender's Cycles and Autodesk Raytracer in Revit use it for final images, and real-time engines such as Unreal Engine and Lumion can use the GPU's ray tracing hardware for interactive views.

Source: Ray trace diagram on Wikimedia Commons, licensed CC BY-SA 4.0.

How Does Ray Tracing Work?

Ray tracing functions by simulating the physical journey of light within a 3D environment. The process begins when rays are projected from a virtual camera through every pixel on the image plane. Each ray travels into the 3D scene and checks for intersections with the objects it encounters. The system determines which surfaces are visible to the camera and calculates how light interacts with those surfaces to produce realistic color, brightness, and shading for each pixel.

The rendering engine then evaluates material properties such as roughness, transparency, and reflectivity. These properties determine how rays respond when they hit a surface, whether they reflect, refract, or get absorbed. This simulation enables an accurate depiction of complex optical effects like glossy reflections, translucent materials, and soft shadows.

Light Interaction Principles

To replicate real-world lighting behavior, ray tracing models several key light interactions:

• Reflection: When a ray strikes a reflective surface like metal or polished stone, it bounces off according to the angle of incidence. The renderer uses this behavior to simulate mirror reflections and highlights that respond dynamically to viewing angles.

  
  

• Refraction: As light passes through transparent materials such as glass or water, the ray bends based on the material’s index of refraction. This produces realistic distortions, magnifications, and caustic light patterns.

  
  

• Shadowing: When objects obstruct rays traveling from a light source, the blocked areas form shadows. Ray tracing determines the intensity and softness of these shadows depending on the size and distance of the light source.

  
  

• Indirect Illumination: Light travels in straight lines between surfaces, but much of the light in a room arrives indirectly: it bounces between walls, floors and ceilings before reaching the camera. Following these bounces is called global illumination, and it creates color bleeding and soft ambient light. Classic ray tracing handles mirror-like bounces well; the diffuse bounces of matte surfaces need path tracing or another global illumination method.

The Ray Tracing Process, Step by Step

The ray tracing process follows a structured sequence of operations that define each frame. It replicates the physical behavior of light in real environments and allows accurate visualization of illumination and material interaction.

1. Ray Emission

In a real room, light travels from windows and lamps toward the camera, and only a small share of it ever reaches the lens. Tracing every ray forward from the lights would waste most of the work, so a ray tracer runs the path in reverse: it sends a primary ray from the camera through each pixel into the scene. In a model of a living room, the ray for one pixel may hit the wooden floor; the next steps work out how much light from the sunlit window reaches that point and travels back toward the camera.

2. Intersection Detection

As sunlight enters a physical space, it first touches surfaces like walls, tables, or floors depending on their position. The same logic applies in ray tracing. Each virtual ray continues forward until it reaches the first object in its path. If a ray encounters a table before hitting the wall behind it, the renderer records the table’s surface as visible in that pixel. This step defines which surfaces the viewer sees and which remain hidden.

3. Shading and Lighting Calculation

When light strikes different materials, their responses vary. A smooth marble floor reflects light sharply, while a brick wall diffuses it softly. In ray tracing, the renderer analyzes material properties such as texture, color, and transparency to determine how much light is reflected or absorbed. In an architectural visualization, this process ensures that glass, metal, and wood appear with realistic brightness and surface detail under both natural and artificial lighting.

4. Secondary Ray Generation

After hitting a surface, part of the light may reflect, pass through, or scatter. This happens, for instance, when sunlight shines through a glass window and creates bright reflections on the floor while also illuminating the interior softly. Ray tracing reproduces this behavior by generating new rays at the point of contact. These secondary rays simulate reflections on glossy materials, refraction through transparent glass, and the gentle diffusion of ambient light that fills a space.

5. Recursive Evaluation

In a real room, light rarely stops after one interaction. It keeps bouncing between surfaces, adding to the brightness and color of the space: a ray may reflect off a white ceiling, light a nearby wall and reach the floor with less intensity. Ray tracing repeats the same steps at every new hit point, so each pixel grows a small tree of rays. Renderers stop at a set maximum number of bounces, because each extra bounce adds less light; in a study by Hall and Greenberg cited on Wikipedia, a maximum depth of 15 gave an average ray tree depth of only 1.7. All contributions are then added up into the final color of each pixel.

Ray Tracing vs Rasterization

Rasterization projects each triangle of a 3D model onto the screen and fills the pixels it covers; shadows, reflections and indirect light are then approximated with shadow maps, screen-space effects or baked lighting. Ray tracing works the other way round: for each pixel it asks which surface is visible and how light reaches it, so reflections, refractions and shadows come from the scene itself. Rasterization runs in real time on almost any GPU; ray tracing needs much more computation per pixel and is accelerated by dedicated GPU hardware.

Most real-time engines today are hybrid: they rasterize the image and add ray-traced effects where they matter most. Unreal Engine, for example, uses hardware ray tracing mostly inside its Lumen global illumination and reflections, according to Epic's documentation (Unreal Engine 5.8 as of October 2026).

Ray Tracing vs Path Tracing

Path tracing is a form of ray tracing, not a rival to it. Classic ray tracing, as Turner Whitted showed in 1979, follows rays recursively for mirror reflections, refractions and shadows, so it can show many bounces in glass and polished surfaces; what it leaves out is the diffuse light that bounces off matte walls and ceilings. Path tracing, introduced by James Kajiya in 1986, sends many randomly sampled paths per pixel through diffuse and glossy bounces too, which gives complete indirect light at the cost of noise that needs many samples or a denoiser.

• Use classic or hybrid ray tracing for fast previews, design iterations and real-time rendering of walkthroughs with ray-traced reflections and shadows.

• Use path tracing for final stills and animations where daylight, soft indirect light and material detail have to look right.

For the full comparison, read our guide to path tracing vs ray tracing.

A Short History of Ray Tracing

Ray tracing is older than most architecture software. These are the steps that brought it from research papers to today's GPUs, as recorded in Wikipedia's ray tracing article and on the vendors' own pages:

• 1968: Arthur Appel uses rays to decide which surfaces are visible and shaded in "Some techniques for shading machine renderings of solids".

• 1978-1980: Scott Roth, at General Motors Research Labs, coins the term "ray casting" (published in 1982).

• 1979: Turner Whitted, at Bell Labs, shows recursive ray tracing with mirror reflection and refraction in the short film "The Compleat Angler".

• 1986: James Kajiya introduces path tracing, which adds the diffuse bounces of light.

• 2018: NVIDIA's Turing GPUs (GeForce RTX and Quadro RTX, September) bring RT Cores, and Microsoft's DirectX Raytracing opens hardware ray tracing to games such as Battlefield V (November).

• 2020: AMD adds Ray Accelerators with RDNA 2 (Radeon RX 6000, October).

Core Techniques in Ray Tracing

Ray tracing relies on several computational techniques that determine how efficiently and accurately light behavior is simulated. These methods define the quality, realism, and performance of architectural rendering workflows.

Ray Casting

Ray casting is the foundation of every ray tracer: it sends one ray per pixel into the scene to find the nearest visible surface and shades that surface from the hit alone. It is fast, but on its own it shows no reflections, refractions or indirect light; ray tracing adds them by casting new rays from each hit point, as Wikipedia's ray casting article explains.

Denoising and Filtering

With few rays per pixel, a ray-traced or path-traced image looks grainy. A denoiser removes that grain so a render can stop earlier: Blender's Cycles offers OpenImageDenoise, Intel's AI denoiser, and OptiX, NVIDIA's AI denoiser, according to the Blender manual, and Unreal Engine's path tracer denoises the final frame once it reaches its target sample count. Denoising is what makes ray-traced previews usable in design reviews.

Hardware and Acceleration: Which GPUs Support Ray Tracing?

Any modern CPU or GPU can ray trace in software. What makes ray tracing fast enough for real time is three things together: GPUs with dedicated ray tracing hardware, acceleration structures that let each ray skip most of the scene, and denoisers that clean up images rendered with few rays.

Hardware vs Software Ray Tracing

Software ray tracing runs the ray calculations on general-purpose processor cores, so it works on any CPU or GPU; Blender's Cycles, for example, can render on either. Hardware ray tracing adds dedicated units to the GPU that speed up the costliest step, searching the scene's bounding volume hierarchy (BVH) for the surface each ray hits. That is what makes ray-traced reflections, shadows and global illumination possible at interactive frame rates.

NVIDIA added RT Cores with its Turing generation in September 2018 (GeForce RTX and Quadro RTX); as of October 2026 its workstation line, NVIDIA RTX PRO (Blackwell), lists fourth-generation ray tracing cores. AMD calls its units Ray Accelerators and has had them since RDNA 2 in October 2020 (Radeon RX 6000, Radeon PRO W6000); its current workstation cards are the Radeon PRO W7000 series (RDNA 3) and the Radeon AI PRO R9000 series (RDNA 4). Intel Arc cards have hardware ray tracing too: Lumion's GPU guide lists the Arc A580, B580 or better for ray tracing in Lumion 2025. So ray tracing is not NVIDIA-only, but each program decides which cards it supports.

Which card fits depends on your program's list of supported GPUs and on the size of your scenes; our guide to the best GPUs for rendering compares current cards. AI rendering is the exception: ArchiVinci generates the image online, not on your computer, so it needs no RTX card, VRAM or render farm, only a browser.

Acceleration Structures

Acceleration structures organize and optimize scene geometry, allowing the renderer to focus only on relevant data. They prevent unnecessary calculations, improving both performance and accuracy in complex scenes.

Bounding Volume Hierarchies (BVH)

A Bounding Volume Hierarchy, or BVH, groups nearby geometry into hierarchical bounding boxes. Each box represents a simplified spatial boundary that contains several objects or polygons. When a ray is cast, the renderer checks these boxes first rather than every individual surface. This process reduces the number of intersection tests and accelerates rendering time.

In architectural visualization, BVH is especially useful for scenes containing thousands of objects, such as furniture, lighting fixtures, and building elements. It maintains high rendering efficiency even when working with large-scale models or intricate interiors.

A KD-tree (k-dimensional tree) is another acceleration structure: it splits the space itself into regions instead of grouping objects into boxes. GPU ray tracing hardware, NVIDIA's RT Cores and AMD's Ray Accelerators, is built to speed up BVH traversal.

AI Denoisers and Ray Reconstruction

AI already speeds up ray tracing, but as a helper, not a replacement. AI denoisers such as OpenImageDenoise and OptiX reconstruct a clean image from fewer ray samples, adaptive sampling spends more samples where an image is noisiest, and NVIDIA's DLSS Ray Reconstruction replaces hand-tuned ray tracing denoisers with an AI model.

AI image generation is a separate route. An AI renderer does not trace rays at all: it generates a new image from a picture and a description. ArchiVinci, for example, takes a screenshot of a model view, a sketch or a photo plus one sentence and generates a photorealistic image that follows it, with no scene, materials or lights to set up; details can drift from the upload, so check them before you present. For the tools in this category, see our comparison of the best AI rendering tools.

In architectural rendering, achieving the right balance between speed and visual quality is essential. While full path tracing delivers the most realistic renders, hybrid and denoised methods are often preferred during design iterations to save time. Final renders typically use higher sample counts for maximum realism once materials and lighting are finalized.

Real-Time vs Offline Ray Tracing in Architecture Software (2026)

Real-time ray tracing renders many frames per second, so you can walk through a model while you change it; offline ray tracing spends as long as it needs on each frame for the final image. As of October 2026, five tools architects often use handle it like this:

Sources: Epic's Path Tracer documentation, Chaos V-Ray, Lumion's requirements, Autodesk's Raytracer help, all read in October 2026. We left Enscape and Twinmotion out of the table because their ray tracing and GPU rules could not be checked on the vendors' own pages that month; check them with Chaos and Epic before you buy hardware. Our comparisons of Lumion vs Enscape and Enscape vs V-Ray go deeper into these tools.

Outside architecture, ray tracing drives film and animation, where offline path tracers render the final frames, and games, where hybrid engines add ray-traced reflections, shadows and global illumination in real time on GPUs with ray tracing hardware. Unity's High Definition Render Pipeline, for example, supports ray tracing only with DirectX 12, in the Editor or a Windows Player (Unity documentation).

If you already have a model but no ray tracer set up, you can also export a view and render your 3D model with AI: ArchiVinci generates a new image that follows the view, with no materials or lights to build. To see how that differs from each tool, read ArchiVinci vs Unreal Engine, ArchiVinci vs V-Ray, ArchiVinci vs Lumion and ArchiVinci vs Revit.

How Architects Use Ray Tracing

In architecture, ray tracing serves three jobs: studying light, checking materials and presenting the design to clients.

Daylight and Lighting Studies

A ray-traced render shows how sunlight falls through a window or across a facade, which makes it a good way to judge mood, glare spots and shadow patterns by eye. It is not a lighting calculation: daylight factors, illuminance and glare metrics need validated simulation tools such as Radiance, and solar heat gain needs energy simulation software.

For mood rather than measurement, an AI renderer is a quicker check. Upload a view of your design to ArchiVinci's AI render generator, describe the materials in a sentence and pick a lighting button such as Daylight, Golden hour or Twilight: each run generates a new image that follows your upload, in seconds, with no lights to place. Twilight gives dusk with the lights on; it is not a calculated sun position, so use it to compare atmospheres, not to prove daylight levels.

Example from our AI render page, made in the Exterior module on the default engine from one uploaded picture of a house without materials and the prompt "Modern house with white plaster walls, timber cladding on the upper floor, large glass openings, a lawn and a few trees." Left: the Golden hour button. Right: the Twilight button.

If the building exists only as a plan, Floor Plan Maker draws it in your desktop browser with walls, doors, windows and room areas, shows it in an isometric 3D view with a sun direction control and a first-person walkthrough, and renders the current view with ArchiVinci's engines, with lighting choices such as Daylight and Golden hour. Drawing, 3D and the walkthrough are free; renders use the normal render credits.

Material Checks

Because a ray tracer follows reflections and refractions, it shows how glass, polished stone, metal and water will read in a design: what a glazed facade reflects, how a marble floor mirrors the furniture, how light bends through a glass balustrade. These checks are only as good as the materials, which is why ray tracers pair with physically based rendering (PBR).

Client Presentations

Final stills and animations for clients are where ray tracing, and especially path tracing, earns its render time. If what you need is a convincing image of a view rather than a ray-traced scene, an AI renderer is the shorter route: ArchiVinci generates a photorealistic image from a screenshot, a sketch or a photo of the view and a sentence, with no scene, materials or lights to set up, and many results are ready to show a client. Client work and other commercial use need an active plan; the plans are on the pricing page.

Left: the uploaded image, a two-story house without materials. Right: the result of the Exterior module for the prompt "Modern house with white plaster walls, timber cladding on the upper floor, large glass openings, a lawn and a few trees." with the Daylight button, on the default engine.

Key Takeaways

• Ray tracing creates an image by sending rays from a virtual camera through each pixel into a 3D scene and following shadow, reflection and refraction rays from each hit.

• Rasterization projects geometry onto the screen and approximates light; most real-time engines combine the two in hybrid rendering.

• Path tracing is a form of ray tracing that also follows diffuse bounces, for complete indirect light at the cost of noise and render time.

• Hardware ray tracing is not NVIDIA-only: NVIDIA RT Cores (since 2018), AMD Ray Accelerators (since 2020) and Intel Arc all accelerate it, and any CPU can still ray trace in software.

• As of October 2026, Unreal Engine and Lumion use ray tracing in real time, V-Ray in real time and offline, and Revit's Autodesk Raytracer offline.

• AI denoisers and AI-driven lighting models speed ray tracing up; AI image generation such as ArchiVinci is a separate route that needs no scene setup and no ray tracing GPU.

Try it in ArchiVinci

Upload a sketch, a photo or a 3D model view and describe the result in a sentence: ArchiVinci generates a new image that follows your upload. A new account starts with 3 free credits.

• Render a sketch, photo or model view: AI rendering software

• From a photo or an image: Image to render AI

Frequently Asked Questions (FAQ)

What hardware is best for architectural ray tracing?

A GPU with hardware ray tracing that your software supports. As of October 2026 that means NVIDIA's RTX PRO cards (Blackwell, fourth-generation RT Cores), AMD's Radeon PRO W7000 or Radeon AI PRO R9000 cards with Ray Accelerators, or a supported Intel Arc card. Check your program's list of supported GPUs first, then size the card to your scenes.

Can ray tracing be used for virtual reality (VR) in architecture?

Yes, within limits. VR runs in real time and needs a high frame rate, so which ray-traced effects you see in a headset depends on the tool and on the GPU. Check your tool's VR requirements before you plan a ray-traced VR review with a client.

How does ray tracing affect rendering time in architectural projects?

Ray tracing usually takes more computation than rasterization because every pixel needs at least one ray, and every shadow, reflection, refraction or bounce adds more rays to test against the scene. GPU ray tracing hardware, denoisers and hybrid rendering have cut render times a lot, so many tools now show ray-traced previews in real time, while final path-traced frames still take longer.

Is ray tracing useful for sustainable design or energy analysis?

Partly. Ray-traced renders help you see how sunlight, shadows and glare will look, which supports early daylight decisions such as window sizes and shading. Measured daylight and glare figures need validated lighting simulation tools such as Radiance, and solar heat gain needs energy simulation software; a visualization engine alone does not calculate them.

What is the difference between real time and offline ray tracing?

Real time ray tracing prioritizes interactivity, using denoising and simplified light paths for instant visual feedback. Offline ray tracing focuses on maximum realism with higher sample counts and more light bounces. The choice depends on the workflow, whether for design iteration or final presentation.

Do all architectural rendering engines support ray tracing?

Most current ones do, in different ways. Unreal Engine and Lumion use hardware ray tracing in real time (in Lumion it is optional and needs a GPU that supports it), V-Ray ray traces in its viewport and for final frames, and Revit renders offline with Autodesk Raytracer. AI renderers such as ArchiVinci do not ray trace at all: they generate a new image that follows your upload.

How is AI changing ray tracing in design visualization?

Mostly by making ray tracing faster, not by replacing it: AI denoisers such as OpenImageDenoise and OptiX, and NVIDIA's DLSS Ray Reconstruction, rebuild a clean image from fewer ray samples. AI image generation is a separate path: tools such as ArchiVinci generate a new image from a picture and a sentence, so architects can use ray tracing where physically based light matters and an AI renderer where a quick, convincing image matters.

Who invented ray tracing?

No single person. Arthur Appel used rays to shade computer renderings in 1968, Turner Whitted showed recursive ray tracing with reflections and refractions at Bell Labs in 1979, and James Kajiya introduced path tracing in 1986. Hardware ray tracing reached consumer GPUs with NVIDIA's RTX cards in 2018.

Is ray tracing NVIDIA only?

No. NVIDIA (RT Cores), AMD (Ray Accelerators, since RDNA 2 in 2020) and Intel (Arc) GPUs all have hardware ray tracing, and any CPU or GPU can ray trace in software, only more slowly. Each program decides which cards it supports.

What is the difference between ray casting and ray tracing?

Ray casting sends one ray per pixel to find the first visible surface and shades it from that hit only. Ray tracing continues from each hit with new rays for shadows, reflections and refractions, which is how mirrors, glass and accurate shadows appear.

Is there anything better than ray tracing?

It depends on the goal. For physically complete light, path tracing, a form of ray tracing that also follows diffuse bounces, gives the fullest result at the cost of render time. For speed, rasterization with approximated or baked lighting is faster. For a quick photorealistic image of a design, AI image generation is a different route that needs no scene at all.

Do I need a ray tracing GPU to use ArchiVinci?

No. ArchiVinci runs in your browser and generates the image online, not on your computer, so no RTX card, VRAM or render farm is needed. You upload a JPG, PNG or WebP of up to 5 MB, such as a view exported from your 3D tool, a sketch or a photo, and describe the result in a sentence. An account is required, and a new account starts with 3 free credits.