Memory Hub Aggregating DRAM Bandwidth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional graphics processing systems face limitations in memory bandwidth due to limited I/O interfaces, compatibility issues with various DRAM protocols, and increased patent licensing burdens from supporting multiple memory standards.
Innovation Solution
A memory hub is introduced to access and aggregate multiple DRAM memories, offload memory access processes from the graphics processing unit, and translate signals between different DRAM protocols, utilizing a high-speed hub bus to increase memory bandwidth and reduce the number of I/O connections required on the GPU.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple DRAM memories are coupled directly to the GPU, then memory bandwidth increases, but the number of I/O interfaces required increases beyond available limits
Solution Approach 1:
A memory hub is introduced as an intermediary device between the GPU and multiple DRAM memories. The memory hub aggregates memory access requests from the GPU and distributes them to multiple DRAM devices, enabling increased memory bandwidth without requiring additional I/O interfaces on the GPU itself. The hub acts as a mediator that consolidates the interface requirements while providing expanded memory access capability.
Solution Approach 2:
The memory hub combines multiple DRAM memory interfaces into a single aggregated interface that connects to the GPU. By merging the memory access paths through a centralized hub, the system achieves higher total memory bandwidth while maintaining a limited number of physical I/O connections from the GPU.
2Adaptability or versatility
If the GPU supports multiple DRAM protocols directly, then compatibility with different memory types improves, but device complexity and manufacturing cost increase
Solution Approach 1:
The memory hub serves as a protocol translation intermediary between the GPU and various DRAM memory types. The hub implements the necessary protocol conversions and adaptations, allowing the GPU to access different DRAM protocols without having to incorporate multiple protocol support directly into its design. This separates the protocol complexity from the GPU.
Solution Approach 2:
The protocol compatibility functionality is extracted from the GPU and relocated to the memory hub. By removing the burden of supporting multiple DRAM protocols from the GPU design, the system achieves protocol versatility through a dedicated component that can be optimized specifically for memory protocol handling without complicating the GPU architecture.
3Adaptability or versatility
If the GPU supports multiple DRAM protocols, then memory compatibility improves, but patent licensing burden increases
Solution Approach 1:
The memory hub acts as a dedicated protocol adaptation layer that handles patent-licensed memory protocols. By concentrating protocol support in the hub rather than the GPU, the system can more efficiently manage licensing requirements and potentially reduce overall manufacturing costs through specialized design and economies of scale in the hub component.
Solution Approach 2:
The system is segmented into two functional components: the GPU which handles graphics processing, and the memory hub which handles protocol adaptation and memory interface management. This segmentation allows protocol-related patent licensing and manufacturing considerations to be isolated to the hub component, potentially reducing the licensing burden on the more expensive GPU.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A memory hub permits a graphics processor to access random access memories, such as dynamic random access memories (DRAMs). In one implementation, the memory hub permits an increase in effective memory bandwidth by aggregating the memory of two or more memories. In another implementation, the memory hub permits a graphics processor to offload memory access interfacing operations to the memory hub.