Intel Works on Ray Tracing in Combination with Generative AI


Intel, one of the leading global tech giants, has a history of collaborating with universities to publish influential research papers. This year makes no exception. Perhaps their most important research of this year focused on path tracing, an essential technique for photorealistic rendering known well to gamers. Intel’s researchers are dedicated to developing efficient methods for handling dynamic objects, complex geometry, and shading calculations within ray tracing. Their aim is to make this advanced rendering technique accessible on a wider range of GPUs enabling this technology to be accessible to the average user and at a reasonable price, since as of today only high-end GPUs are capable of using it.

Path tracing vs ray tracing

Ray tracing, the term that is more commonly used, refers to simulating the behavior of light rays as they interact with objects while path tracing extends this by considering the cumulative effects of multiple rays along a path (Imagine rays bounce off to other objects and those too are calculated). Efficient techniques for managing acceleration structures in ray tracing have been a key focus of recent research, particularly for dynamic objects and complex geometry. These advancements ensure the smooth rendering of scenes with moving elements or intricate designs, delivering exceptional accuracy at a lower computational cost.

Forest render with 1 trillion triangles | Intel
A forest render with 1 trillion triangles (Image: Intel)

Two of this year’s papers demonstrate novel approaches to handling the acceleration structures used in ray tracing for dynamic objects and complex geometry. By optimizing these crucial components, Intel aims to improve the performance of path-tracing algorithms.

Optimizing the algorithms

The first paper presents a novel algorithm for evaluating the GGX distribution, a commonly used mathematical function in computer graphics. By simplifying materials represented with the GGX distribution to a hemispherical mirror, the algorithm achieves faster performance, enabling efficient simulation on a computer which results in improved rendering processes and enhanced visuals.

Sampling visible GGX normals with spherical caps | Intel
Sampling visible GGX normals with spherical caps (Image: Intel)

The second paper focuses on the real-time simulation of glittery surfaces. Reproducing a glittery surface through ray tracing is still an open challenge and very performance heavy due to the number of glints (tiny reflective surfaces) needed. Intel proposes a statistical law for the average number of glints that may solve this challenge.

Real-time rendering of glinty appearances using distributed binomial laws on anisotropic grids| Intel
Real-time rendering of glinty appearances using distributed binomial laws on anisotropic grids (Image: Intel)

The company is also dedicated to integrating AI into the process, as showcased in their latest work at SIGGRAPH 2023, presenting a unified framework that seamlessly integrates neural rendering and real-time ray tracing. In another paper, Intel researchers focus on democratizing the diffusion models through a similar optimization. Tools such as DALL-E or Midjourney have a large market cap with diffusion models as it is a highly resource-intensive process, with the new reformulation and simplification of this diffusion model, Intel intends to change this.

Going VR with 360° AI imagery?

Intel Labs is introducing an AI Diffusion Model that can generate 360-degree images from text prompts, providing a more accessible and efficient way to create photorealistic computer graphics. This model uses AI-driven rendering to create complex 3D models based on descriptions and processes image data in real time with unprecedented accuracy. With this technology, Intel is able to generate lifelike scenes with realistic lighting and materials. The model can also be used to accurately render the shape of objects and the environment they exist in using anisotropic grids. By utilizing this new approach, Intel is simplifying the traditional rendering process while adding more realism to its imagery.


Intel is still dedicated to improving the performance of the algorithms such as path tracing and diffusion models to deliver better and more accessible experiences. Many more collaborations on such papers are to come. As of today, we cannot see the full extent of their work but we can, hopefully, have it in our hands tomorrow, when these technologies will be widely available. With Intel’s ongoing efforts and advancements, the future holds exciting possibilities for immersive and visually stunning environments.

Photo credit: The feature image shows Sarah Mainwaring from Inflexion Games talking about CPU and GPU performance tests at an Intel Innovation event. The images in the body of the article are owned by Intel and were made available for press usage.
Source: Intel press release

Was this post helpful?

Benjamin Adjiovski
Benjamin Adjiovski
Hi! I am a Computer Science Engineer with a passion for all things related to technology. I believe that technology has the power to change the world, so I love staying up-to-date on the latest innovations. If you share the same passion, be my guest.
- Advertisment -
- Advertisment -
- Advertisment -
- Advertisment -
- Advertisment -
- Advertisment -