keepcalculator

06 Numbers and Powers

Nvidia RT Cores vs AMD Ray Accelerators Explained

Card No. 06·80920 · Updated · Standard formula, shown below

Nvidia RT Cores and AMD Ray Accelerators are dedicated hardware units on their respective graphics cards that speed up ray-traced rendering. This article explains what each does, how they differ, and how to choose between them for gaming or creative work.

What Are RT Cores and Ray Accelerators?

Ray tracing is a rendering technique that simulates how light behaves in the real world by calculating the path of light rays as they bounce off surfaces. It produces highly realistic lighting, shadows, and reflections, but it is computationally expensive. To make real-time ray tracing feasible, both Nvidia and AMD have added dedicated hardware to their GPUs.

Nvidia introduced RT Cores with its Turing architecture in 2018, starting with the GeForce RTX 20 series. AMD followed with Ray Accelerators in its RDNA 2 architecture, debuting in the Radeon RX 6000 series in 2020. Both are specialized circuits that handle the math-heavy tasks of ray tracing, such as calculating ray-triangle intersections and bounding volume hierarchy (BVH) traversal, so the main GPU cores are free for other work.

How Ray Tracing Hardware Works

Ray tracing in real time relies on a data structure called a bounding volume hierarchy. The scene is divided into boxes, and the GPU checks whether a ray hits a box before checking the geometry inside it. This reduces the number of calculations needed. The dedicated hardware accelerates these checks and the actual intersection tests.

Both Nvidia RT Cores and AMD Ray Accelerators are designed to do this, but they differ in implementation. Nvidia’s RT Cores are more integrated into the GPU’s pipeline and handle both BVH traversal and ray-triangle intersection. AMD’s Ray Accelerators are simpler units that compute ray intersections, while the shader cores handle the traversal logic. This is a key architectural difference that affects performance and efficiency.

Architectural Differences

Nvidia’s RT Cores are part of a unified architecture where the GPU can switch between rasterization and ray tracing tasks. The RT Core includes a bounding volume hierarchy (BVH) traversal unit and a ray-triangle intersection unit. This means the core can process the entire ray tracing pipeline without relying on shader instructions for traversal.

AMD’s Ray Accelerators, on the other hand, are simpler. They perform ray-triangle intersection tests and a few other operations, but the BVH traversal is done by the shader cores. This means AMD’s approach uses more shader instructions for ray tracing, which can reduce performance in heavy ray tracing workloads, but it also allows the hardware to be more flexible and cheaper to produce.

Performance and Efficiency

In practice, Nvidia’s RT Cores have generally delivered higher ray tracing performance than AMD’s Ray Accelerators, especially in the first generation. For example, the RTX 2080 Ti outperformed the RX 6800 XT in ray tracing benchmarks, though the gap varies by game and resolution. Nvidia also has a head start with DLSS (Deep Learning Super Sampling), which uses AI to upscale images and can offset the performance cost of ray tracing.

AMD has improved its ray tracing with each generation. The RDNA 3 architecture, used in the RX 7000 series, doubled the number of Ray Accelerators per compute unit and added a new BVH traversal unit. This narrowed the gap, but Nvidia still holds an advantage in most ray tracing titles. Efficiency also differs: Nvidia’s RT Cores are more power-hungry, but they also deliver higher performance per watt in ray tracing workloads.

How to Choose Between Them

When choosing between Nvidia and AMD for ray tracing, consider your primary use case and budget.

  • Gaming: If you play games with heavy ray tracing (e.g., Cyberpunk 2077, Control), Nvidia GPUs generally offer better performance. However, AMD cards often provide better value for rasterized gaming (non-ray-traced).
  • Content Creation: For 3D rendering in Blender or Maya, Nvidia’s RT Cores are widely supported and faster in most benchmarks. AMD has made strides, but software support is still more mature on Nvidia.
  • Budget: AMD’s Ray Accelerators are often found in more affordable cards, making them a good entry point into ray tracing if you don’t need maximum performance.

Also consider upscaling technologies: Nvidia’s DLSS and AMD’s FSR (FidelityFX Super Resolution) can boost frame rates. DLSS is generally considered superior in quality, but FSR works on both brands.

Common Misconceptions

One common misconception is that ray tracing performance is solely determined by the number of RT Cores or Ray Accelerators. In reality, the overall GPU performance, memory bandwidth, and driver optimization play significant roles. A card with more RT Cores but slower memory may not beat a card with fewer but faster cores.

Another misconception is that AMD’s Ray Accelerators are useless because they are slower. They still provide a significant speedup over software ray tracing, and AMD’s approach allows for a smaller die size, which can lower costs. For casual ray tracing or games with light effects, AMD cards are perfectly capable.

Future of Ray Tracing Hardware

Both companies are investing heavily in ray tracing. Nvidia’s Ada Lovelace architecture (RTX 40 series) introduced third-generation RT Cores with improved BVH traversal and a new “Shader Execution Reordering” feature that improves efficiency. AMD’s RDNA 4 is expected to further improve ray tracing, possibly with a more unified design.

As game developers increasingly use ray tracing for global illumination and reflections, the importance of dedicated hardware will only grow. Both Nvidia and AMD are also working on hybrid rendering techniques that combine rasterization and ray tracing to balance quality and performance.

Frequently Asked Questions

Do I need ray tracing hardware to play games?

No, you can play games with ray tracing disabled. Many games still support traditional rasterization, which does not require RT Cores or Ray Accelerators. However, if you want to enable ray tracing for better visuals, you will need a GPU with dedicated hardware (or a very powerful CPU for software ray tracing, which is rarely practical).

Is DLSS or FSR better for ray tracing?

DLSS (Nvidia) generally produces higher image quality and better performance in ray tracing games, but it only works on Nvidia GPUs. FSR (AMD) works on both Nvidia and AMD cards, and while it is not as sharp, it is more accessible. The best choice depends on your GPU and personal preference.

Can I use ray tracing on older GPUs?

Older GPUs without dedicated ray tracing hardware (e.g., GTX 10 series) can run ray tracing in some games via software, but performance is very poor. It is not recommended for a playable experience. For real-time ray tracing, a GPU with RT Cores or Ray Accelerators is required.

Why do AMD cards have fewer ray tracing features?

AMD’s Ray Accelerators are simpler than Nvidia’s RT Cores, which means they do not handle all ray tracing tasks. This is a design choice to keep costs down and improve rasterization performance. However, AMD has been adding more features with each generation, such as BVH traversal in RDNA 3.

Does ray tracing affect VRAM usage?

Yes, ray tracing can increase VRAM usage because the GPU needs to store the BVH and other acceleration structures. This is why cards with more VRAM (e.g., 12GB or more) are often recommended for ray tracing at high resolutions.

Are RT Cores and Ray Accelerators the same as Tensor Cores?

No. Tensor Cores are Nvidia’s AI acceleration units, used for DLSS and other machine learning tasks. RT Cores handle ray tracing. AMD does not have direct equivalents to Tensor Cores, but its GPU architecture includes AI accelerators in some models for similar purposes.

Related

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.

More cards in this drawer

The whole Numbers and Powers drawer →