POWERVR PHOTON

The most advanced ray tracing architecture in the world, enabling desktop-level ray tracing for mobile.

ray_tracing_architecture

The First Mobile-Optimised Ray Tracing Architecture

Ray tracing compute workloads are dominated by countless intersection tests between the millions of rays emitted in each frame and the acceleration structure, that contains a hierarchical box structure as well as the triangle geometry. The RAC fully offloads these expensive operations to dedicated hardware delivering significant area and power-efficiency benefits.

Introducing the RAC:
The Ray Acceleration Cluster

The RAC is responsible for all ray tracing activity on the PowerVR GPU. It encompasses the entire process; from emitting a ray (from a shader/kernel) to returning the hit (or miss) results back to the ALU for processing.

The RAC is compatible with all modes and functionality exposed by ray tracing APIs, including Vulkan® extensions, and also ray query and ray pipeline.

ray acceleration cluster
Imagination CXT diagram Ray Testing Units

RAC Components:
Ray Testing Units

Ray tracing compute workloads are dominated by countless intersection tests between the millions of rays emitted in each frame and the acceleration structure, that contains a hierarchical box structure as well as the triangle geometry. The RAC fully offloads these expensive operations to dedicated hardware delivering significant area and power-efficiency benefits.

RAC Components:
Hardware BVH Walking

The RAC fully offloads the highly divergent recursive process of tracing each ray through the bounding volume hierarchy (BVH). The hardware includes dedicated ray ram storage and multiple scheduling engines to handle this process in a power, processing, and bandwidth-efficient way.

hardware_BVH
divergent_images

Packet Coherency Gather

The Packet Coherency Gather unit is a Level 4 Ray Tracing Levels System (RTLS) solution. It is responsible for analysing all active rays and creating packets (groups) of coherent rays (rays with similar trajectories) to test against the scene together. By testing and processing coherent rays we enable parallelism and achieve higher processing and bandwidth efficiency, as thanks to data re-use we can achieve high hit rates on our cache structure.

Mobile Optimised

The IMG CXT RT3 single-core configuration hits the sweet spot for premium smartphones, balancing performance with the optimal power budget for sustained ray tracing enabled gaming. In addition to our highly scalable Photon architecture based RAC unit, IMG CXT includes a 50% boost in compute, texturing and geometry processing performance compared to our previous generation.

mobile_optimised

Beyond Mobile

The IMG CXT GPU continues to build on the decentralised multi-core architecture introduced with the B-Series enabling up to 9 TFLOPS of FP32 Performance and 7.8Gray/s, enabling laptop, desktop and cloud gaming solution with up to 2.5x better ray tracing efficiency compared to today’s solutions for those markets.

Imagination CXT Digram CXT x4 grey backgound