Multi-GPU Geometry Rendering by Screen Region Testing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing technologies face challenges in efficiently utilizing multiple GPUs to render complex scenes or images, as they struggle to support increased screen pixel count and geometry density, leading to limitations in performance and rendering efficiency.

Innovation Solution

A method for multi-GPU rendering that divides responsibility for rendering geometry based on screen regions, allowing GPUs to generate and share information about geometry relations to optimize rendering efficiency, enabling simultaneous rendering and avoiding redundant work.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If multiple GPUs are used to render complex scenes, then rendering capability and scene complexity are improved, but GPU utilization efficiency deteriorates

Engineering Contradiction:
Improverendering capabilityVSAvoidGPU utilization efficiency
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The screen is divided into multiple screen regions, with each GPU assigned responsibility for specific regions. This segmentation allows each GPU to focus on rendering geometry relevant to its assigned regions, reducing redundant processing and improving overall utilization efficiency while maintaining the ability to render complex scenes across multiple GPUs.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different GPUs are assigned different screen regions based on the spatial distribution of geometry. This local quality approach ensures that each GPU processes geometry that is most relevant to its assigned region, optimizing the balance between rendering capability and utilization efficiency by matching computational resources to actual rendering needs.

Inventive Principle:
Principle #3Local quality

2Productivity

If traditional multi-GPU rendering is used, then screen pixel count and geometry density are limited, but rendering efficiency is maintained

Engineering Contradiction:
Improverendering efficiencyVSAvoidscreen pixel count and geometry density
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

By segmenting the screen into regions and assigning them to different GPUs, the system can process higher geometry density and screen pixel counts without sacrificing rendering efficiency. Each GPU handles a subset of the total rendering workload, allowing the system to scale to higher resolutions and more complex scenes while maintaining efficient utilization of each GPU's processing capacity.

Inventive Principle:
Principle #1Segmentation

3Reliability

If geometry is rendered by all GPUs, then complete coverage is ensured, but redundant processing increases

Engineering Contradiction:
Improverendering coverageVSAvoidredundant processing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The screen is segmented into regions assigned to specific GPUs, ensuring that each piece of geometry is rendered by the GPU responsible for the region where it appears. This segmentation maintains complete rendering coverage while eliminating redundant processing, as each GPU only processes geometry relevant to its assigned regions rather than all geometry in the scene.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Each GPU is assigned local responsibility for specific screen regions, ensuring that geometry is rendered with complete coverage in its assigned regions while avoiding redundant processing in other regions. This local quality approach optimizes the balance between rendering reliability and processing efficiency by matching geometry processing to spatial requirements.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4100138B1System and method for efficient multi-GPU rendering of geometry by region testing while rendering
Publication Date: 2026.03.04 SONY INTERACTIVE ENTERTAINMENT LLC
  • EP4100138B1 patent drawingFigure 1
  • EP4100138B1 patent drawingFigure 2
  • EP4100138B1 patent drawingFigure 3

AI summary

A method for graphics processing. The method including rendering graphics for an application using a plurality of graphics processing units (GPUs). The method including dividing responsibility for the rendering geometry of the graphics between the plurality of GPUs based on a plurality of screen regions, each GPU having a corresponding division of the responsibility which is known to the plurality of GPUs. The method including generating information regarding a piece of geometry with respect to a first screen region for which a first GPU has a first division of responsibility, while rendering the piece of geometry at a second GPU for an image. The method including rendering the piece of geometry at the first GPU using the information.