Graphics Memory Extended with Nonvolatile Memory for VR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current graphics processing systems face challenges in efficiently managing and loading high-quality graphics assets in real-time, particularly in virtual reality applications, due to limited hardware resources and the need for photorealistic experiences, which often result in frame rate drops and nausea caused by latency and bandwidth issues.
Innovation Solution
The integration of nonvolatile memory (NVM) with graphics processing units (GPUs) allows for the dynamic loading of high-resolution assets based on camera position and predicted future positions, using a memory mapper to extend GPU memory with fast, low-latency, and high-capacity storage, enabling adaptive asset loading and reducing the need for frequent data transfers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-quality graphics assets are loaded into hardware memory, then graphics rendering quality is improved, but memory bandwidth and capacity are consumed, causing frame rate drops
Solution Approach 1:
The patent introduces a new memory dimension by integrating non-volatile memory (NVM) with volatile graphics memory to create an extended memory hierarchy. This allows high-quality graphics assets to be stored in NVM with large capacity while maintaining fast access through the volatile memory layer, resolving the contradiction between rendering quality and frame rate by providing both high capacity and fast access speeds.
Solution Approach 2:
The system pre-loads graphics assets into the extended memory system before they are needed for rendering. The memory mapper predicts future asset requirements and loads them in advance into the volatile memory portion of the extended memory, reducing the need for frequent data transfers during rendering and maintaining high frame rates while supporting high-quality assets.
2Manufacturing precision
If high-resolution assets are dynamically loaded based on camera position, then photorealistic experience is improved, but memory access latency increases
Solution Approach 1:
The memory mapper component predicts future camera positions and pre-loads corresponding high-resolution assets into the volatile memory portion of the extended memory system before they are actually needed. This preliminary action eliminates memory access latency during rendering by ensuring assets are already in fast memory when the camera reaches their required position, maintaining both photorealistic quality and low latency.
Solution Approach 2:
The system dynamically adjusts the memory allocation and asset loading strategy based on real-time camera position and predicted movement. The memory mapper continuously monitors camera trajectory and dynamically loads, unloads, and swaps assets in the extended memory system, optimizing the balance between asset quality and access latency adaptively as the scene changes.
3Quantity of substance
If nonvolatile memory is integrated with GPU memory, then storage capacity and speed are improved, but device complexity increases
Solution Approach 1:
The patent merges non-volatile memory and volatile graphics memory into a unified extended memory system that appears as a single memory space to the GPU. The memory mapper manages the unified address space and handles data placement between NVM and volatile memory transparently, reducing the perceived complexity for applications while providing increased storage capacity and high-speed access.
Solution Approach 2:
The memory mapper acts as an intermediary layer between the GPU and the extended memory system comprising NVM and volatile memory. It handles address translation, data caching, and memory management operations, shielding the GPU from the complexity of the hybrid memory architecture while providing the benefits of increased capacity and high-speed access to high-quality graphics assets.
Data Source
AI summary
An embodiment of an electronic processing system may include an application processor, system memory communicatively coupled to the application processor, a graphics processor communicatively coupled to the application processor, graphics memory communicatively coupled to the graphics processor, and persistent storage media communicatively coupled to the application processor and the graphics processor to store one or more graphics assets, wherein the graphics processor is to access the one or more graphics asset mapped from the persistent storage media. The persistent storage media may include a low latency, high capacity, and byte-addressable nonvolatile memory. The one or more graphics assets may include one or more of a mega-texture and terrain data. Other embodiments are disclosed and claimed.