Cloud-Based Real-Time Ray Tracing With Distributed Denoising

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing ray tracing techniques are resource-intensive and inefficient for real-time performance, particularly in distributed systems where denoising frameworks operate on a single machine and lack access to all rendered pixels, limiting the quality of rendered images.

Innovation Solution

A system and method for distributed real-time ray tracing that utilizes a machine-learning engine for denoising operations across multiple nodes, exchanging ghost region data to improve image quality and efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If ray tracing is performed using traditional single-machine frameworks, then the system complexity is low, but the image quality is poor due to inability to access all rendered pixels for denoising

Engineering Contradiction:
Improveimage qualityVSAvoidsystem complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The rendering system is segmented into multiple independent nodes, each responsible for rendering specific portions of the image. This allows distributed denoising operations to access relevant pixels from multiple nodes, improving image quality while maintaining manageable system complexity through modular architecture

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

A cloud-based rendering server acts as an intermediary that collects rendered pixels from multiple distributed nodes, performs centralized denoising operations using machine learning models, and returns the denoised image data to the original nodes for display

Inventive Principle:
Principle #24Intermediary (Mediator)

2Manufacturing precision

If more computational resources are allocated to ray tracing operations, then the rendering quality improves, but the processing time increases

Engineering Contradiction:
Improverendering qualityVSAvoidprocessing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

Bounding volume hierarchies are pre-computed and stored before actual ray tracing operations. This preliminary preparation allows for faster traversal during rendering, improving both quality and speed by avoiding redundant computational work

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Traditional mechanical denoising algorithms are replaced with machine learning-based denoising models that run on cloud servers. This substitution enables more efficient noise reduction with lower computational overhead on client devices, improving rendering quality without proportionally increasing processing time

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If bounding volume hierarchy data is transmitted uncompressed, then the data accuracy is high, but the transmission bandwidth consumption increases

Engineering Contradiction:
Improvedata accuracyVSAvoidbandwidth consumption
Core Design Contradiction:
Measurement precisionVSLoss of energy

Solution Approach 1:

The precision parameters of bounding volume hierarchy data are dynamically adjusted based on requirements. Critical geometric data maintains high precision while less critical attributes use compressed representations, optimizing the balance between data accuracy and transmission efficiency

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3675056B1Cloud-based realtime raytracing
Publication Date: 2025.10.01 INTEL CORP
  • EP3675056B1 patent drawingFigure 1
  • EP3675056B1 patent drawingFigure 2
  • EP3675056B1 patent drawingFigure 3

AI summary

1. Cloud-based real time rendering. For example, one embodiment of a system comprises: a first graphics processing node to perform a first set of graphics processing operations to render a graphics scene, the first set of graphics processing operations comprising ray-tracing independent operations; an interconnect or network interface coupling the first graphics processing node to a second graphics processing node; the second graphics processing node to receive an indication of a current view of a user of the first graphics processing node and to receive or construct a view-independent surface generated by view-independent ray traversal and intersection operations; the second graphics processing node to responsively perform a view-dependent translation of the view-independent surface based on the current view of the user to generate a view-dependent surface and to provide the view-dependent surface to the first graphics processing node; and the first graphics processing node to perform a second set of graphics processing operations to complete rendering of the graphics scene using the view-dependent surface.