What is an RT Core?
An RT Core (Ray Tracing Core) is a dedicated physical hardware unit found on modern graphics processing units (GPUs), designed specifically to accelerate the computationally intensive math required for real-time ray tracing. It calculates how light rays interact with virtual environments, enabling realistic reflections, shadows, and lighting in video games and professional rendering applications.
Key Takeaways
RT Cores are specialized hardware accelerators distinct from standard shading units.
They offload the massive mathematical burden of bounding volume hierarchy (BVH) calculations.
Without dedicated cores, real-time ray tracing severely degrades gaming frame rates.
First introduced by Nvidia in 2018, they are now standard in modern desktop and console GPUs.
History and Evolution
Before dedicated hardware, real-time ray tracing was a holy grail for computer graphics. Standard GPUs relied entirely on rasterization—a technique that projects 3D models onto a 2D screen but struggles with accurate light bounces.
In 2018, Nvidia introduced the Turing architecture, featuring the first generation of RT Cores inside GeForce RTX GPUs. AMD followed in 2020 with Ray Accelerators in their RDNA 2 architecture. Today, multi-generational advancements have drastically increased the number of ray-triangle intersections a GPU can calculate per clock cycle.
How RT Cores Work
Ray tracing simulates the physical behavior of light by casting millions of virtual rays from the camera viewpoint back into the 3D scene. To do this efficiently, hardware uses a spatial sorting structure called Bounding Volume Hierarchy (BVH). The RT Core handles two primary workloads:
BVH Box Traversal
The core tests rays against large virtual boxes encompassing groups of 3D objects, continuously narrowing down the location of an object until it finds a match.
Ray Triangle Intersection
Once the correct box is identified, the core calculates the exact point where the light ray intersects with the polygon triangles of a 3D model.
By offloading these two geometric math operations, the primary graphics shaders are freed up to handle pixel shading; frame rates stay playable, and lighting updates in real time.
RT Cores vs. Traditional Shaders
Feature | RT Cores | Traditional Shaders (CUDA / Stream Processors) |
|---|---|---|
Primary Function | BVH traversal and ray intersection math | General geometry, pixel, and vertex shading |
Workload Type | Highly specialized bounding box calculations | General-purpose parallel computing |
Ray Tracing Speed | Extremely fast due to dedicated hardware pipelines | Very slow when forced to emulate ray equations |
Versatility | Fixed-function architecture | Highly programmable for various graphic tasks |
Advantages of Dedicated Ray Tracing Hardware
Photorealistic Visuals
Enables physically accurate global illumination, ambient occlusion, soft shadows, and lifelike reflections that adapt dynamically to moving objects.
Frame Rate Preservation
Offloading ray calculations prevents the primary GPU shading processors from becoming bottlenecked, ensuring smoother gameplay.
Dynamic Environmental Scaling
Allows developers to build open worlds where lighting changes realistically based on time of day, weather, and player actions without pre-baking lightmaps.
Limitations and Challenges
Performance Overhead
Even with dedicated hardware, enabling ray tracing causes a noticeable drop in frames per second, often requiring AI upscaling tools like DLSS or FSR to compensate.
Silicon Die Space
RT Cores occupy physical space on the GPU chip, which increases manufacturing costs and power consumption.
Software Dependency
The hardware remains idle if a video game or rendering application does not explicitly support modern ray tracing APIs like DirectX Raytracing (DXR) or Vulkan.
Related Technology Terms
Rasterization: The traditional method of rendering 3D graphics by translating vector data into pixels.
BVH (Bounding Volume Hierarchy): A tree structure used to organize geometric objects to optimize ray tracing calculations.
Path Tracing: An advanced form of ray tracing that simulates thousands of light rays per pixel for ultimate graphical fidelity.
DLSS (Deep Learning Super Sampling): Nvidia AI technology that upscales lower-resolution frames to offset the performance cost of ray tracing.