AMD, Possible Ray Tracing Innovations with RDNA4: What’s New in GPUs | Hardware Catalog | Hardware and Software of Computer | Gpu vs cpu | Turtles AI

AMD could introduce significant innovations in Ray Tracing with RDNA4: What to expect from the new GPUs

Highlights:

• Double Ray Tracing Intersect Engine: Improves parallel ray processing.
• RT instance node transform: More efficient geometry handling.
• 64B RT node: Enhances processing speed and requires less memory.
• Support for OBB: Greater precision with smaller bounding volumes.

The RDNA4 architecture from AMD promises substantial improvements in Ray Tracing performance, based on technical specifications that could transform the gaming and digital graphics sector. Here’s an overview of the innovations coming.

Recent leaks from the reliable source Kepler_L2 have provided an in-depth look at the features that AMD’s RDNA4 architecture will introduce in the field of Ray Tracing. Kepler, known for early details on the PlayStation 5 Pro and other hardware specs, has now shared a list of innovative features for RDNA4. This article will analyze these innovations, delving into the potential technical implications for end users.

One of the most notable innovations is the inclusion of the "Double Ray Tracing Intersect Engine," which should allow greater parallel processing of rays, improving the overall efficiency of the GPU. The RT instance node transform is another significant addition, enabling GPUs to handle geometries more efficiently, including translation, rotation, and scaling processes. Another optimization involves the "Ray Tracing Tri Pair," designed to reduce the computational load required to calculate ray-triangle intersections.

In terms of memory and processing speed, the introduction of a 64B RT node represents a significant improvement, requiring less memory and boosting processing speed. The optimization of the Bounding Volume Hierarchy (BVH) footprint also promises faster grouping of scene geometry, enhancing overall performance.

Another innovation is the support for OBB and instance node intersection, which could refer to the Oriented Bounding Box. This technique could offer greater precision and efficiency due to smaller and more accurate bounding volumes. Finally, encoding change flags in barycentrics simplifies the detection of procedural nodes, further improving computational efficiency.

All these technical features remain, for now, unconfirmed officially by AMD. The company has avoided mentioning the RDNA4 architecture at various recent events, focusing instead on the upcoming Zen5 and Zen6 CPUs, expected through 2026. However, expectations are high for the launch of the PlayStation 5 Pro by the end of the year, with possible announcements as early as September. At the same time, the AMD Radeon RX 8000 series, integrating RDNA4, is expected to hit the market in early 2025, if rumors hold true.