Memory Controller Interconnect for Data Exchange
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computer systems face limitations in processor-to-memory bandwidth due to the exclusive connection of memory to the processor chip, which restricts efficient data exchange between memory controllers, especially in systems with multiple memory modules and I/O operations.
Innovation Solution
A device and method that connect multiple memory controllers directly via an interconnect, such as a memory ring, with each controller equipped with accelerators to facilitate data exchange and perform deterministic operations like garbage collection, data copying, and cache coherency protocols, allowing concurrent operations and bypassing processor involvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If memory is exclusively connected to the processor chip, then memory control is simplified, but processor-to-memory bandwidth is limited
Solution Approach 1:
The patent segments the memory control function by introducing separate memory controllers on individual memory modules, decoupling them from the processor chip. This allows multiple memory controllers to operate independently and concurrently, increasing overall memory bandwidth while simplifying the processor's control burden.
Solution Approach 2:
The patent introduces a new communication dimension by establishing direct peer-to-peer connections between memory controllers through an interconnect fabric, bypassing the traditional processor-mediated path. This creates additional data exchange pathways that increase bandwidth without adding processor complexity.
2Extent of automation
If memory controllers are moved to separate chips on DIMMs, then memory autonomy is improved, but data exchange between memory controllers requires processor involvement
Solution Approach 1:
The patent introduces an interconnect fabric as an intermediary communication channel between memory controllers, enabling direct data exchange without processor involvement. This mediator allows autonomous memory operations while maintaining system coherence.
Solution Approach 2:
The patent extracts the data exchange function from the processor's responsibility by providing dedicated interconnect pathways between memory controllers. This separation allows memory subsystems to operate autonomously for local operations while the processor focuses on computational tasks.
3Productivity
If memory modules are used in an interleaved way, then memory channel balancing is improved, but local memory operations are ruled out
Solution Approach 1:
The patent enables dynamic address mapping where memory addresses can be flexibly assigned to different memory modules. This dynamic allocation allows the system to switch between interleaved access patterns for balanced channel utilization and localized access patterns when data resides on a single module, optimizing for the specific operation type.
Data Source
AI summary
A device with an interconnect having a plurality of memory controllers for connecting the plurality of memory controllers. Each memory controller of the plurality of memory controllers is coupled to an allocated memory for storing data. Further, each memory controller of the plurality of memory controllers has one accelerator of a plurality of accelerators for mutually exchanging data over the interconnect.