The World’s First Acceleration Structure Generation H/W for Powerful Dynamic Real-time Ray Tracing Implementation
RayTree is a dedicated KD-tree generation hardware IP which can be used in ray tracing deployments requiring dynamic scene changes, in an on-line game for example. In order to deliver interactivity, with real-time high-quality dynamic 3D contents requires using KD-tree re-generation for any application. Despite of CPU overhead in today’s system, CPU has been responsible for KD-tree generation and this causes process delays as well as high-level of power consumption in the overall solution. The utilization of the RayTree IP core can off-load the CPU for these compute intensive functions, freeing the CPU to more effectively process other workloads. Utilizing the Ray Tree IP in the cloud, global illumination shading can be effectively deployed for multi-player on-line gaming.
RayTree now can replace CPU’s role and further maximize KD-tree re-generation performance. RayTree re-generates KD-tree in real-time basis, thereby realizing on-the-fly dynamic scene processing without any CPU utilization and saving power consumption. Furthermore, RayTree acts as load balancer between rendering and tree building task. RayTree optimizes resource distribution and minimizes response time, and ultimately solves the bottleneck problem for efficient ray tracing rendering.
RayTree efficiently cooperates with RayCore to realize real-time ray tracing and eventually allows interactivity, and high degree of freedom for high-quality dynamic 3D contents implementation.
RayTree scans primitives and generates acceleration structure (KD-tree) to support real-time dynamic scene processing
RayTree solves the bottleneck problem between rendering and tree building task by distributing resource efficiently
Compare to KD-tree generation performance using mobile CPU, RayTree has 35x faster KD-tree generation capability
By combining RayTree with RayCore, KD-tree generation process is eliminated in mobile CPU which results in reduced power consumption at system level
Input data type
Input data format
Bound box (AABB)
Output data format
Node data format,
List (32bit integer)
Max data set size
Max data size
Contact the factory here for detailed benchmarks and performance metrics