Die-Stacked Memory Coherency Manager Offloading

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Processing systems face bandwidth and latency limitations due to heavy use of memory interconnect for cache coherency protocols, leading to decreased efficiency and increased power consumption, especially in systems with shared memory.

Innovation Solution

Implementing a die-stacked memory device with a hardware coherency manager that offloads cache coherency protocol operations, allowing for higher bandwidth and lower latency, and enabling external devices to focus on program execution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional cache coherency protocols are implemented with separate memory devices, then cache coherency is maintained, but interconnect bandwidth is reduced and latency is increased

Engineering Contradiction:
Improvecache coherencyVSAvoidinterconnect bandwidth
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent merges the memory device and coherency manager into a single integrated device. The coherency manager is coupled to the memory cell circuitry within the same device, eliminating the need for separate interconnect communication between memory and coherency management components. This integration resolves the bandwidth and latency issues caused by external interconnect traffic while maintaining cache coherency functionality.

Inventive Principle:
Principle #5Merging (Combining)

2Reliability

If cache coherency operations are performed externally, then coherency is maintained, but power consumption increases due to repeated inter-device communication

Engineering Contradiction:
Improvecache coherencyVSAvoidpower consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

By integrating the coherency manager within the memory device, the patent eliminates repeated inter-device communication for coherency operations. The coherency manager can directly interact with the memory cell circuitry without requiring external signaling, thereby reducing power consumption while maintaining coherency reliability.

Inventive Principle:
Principle #5Merging (Combining)

3Quantity of substance

If system memory is implemented as a separate device, then memory capacity is increased, but inter-device latency is increased due to propagation delay and handshaking

Engineering Contradiction:
Improvememory capacityVSAvoidinter-device latency
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

The patent integrates the coherency manager with the memory device, allowing coherency operations to be performed without external communication overhead. This reduces the effective latency for coherency-related memory operations while preserving the memory capacity benefits of having a dedicated memory device.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS9170948B2Cache coherency using die-stacked memory device with logic die
Publication Date: 2015.10.27 ADVANCED MICRO DEVICES INC
  • US9170948B2 patent drawing
  • US9170948B2 patent drawing
  • US9170948B2 patent drawing

AI summary

A die-stacked memory device implements an integrated coherency manager to offload cache coherency protocol operations for the devices of a processing system. The die-stacked memory device includes a set of one or more stacked memory dies and a set of one or more logic dies. The one or more logic dies implement hardware logic providing a memory interface and the coherency manager. The memory interface operates to perform memory accesses in response to memory access requests from the coherency manager and the one or more external devices. The coherency manager comprises logic to perform coherency operations for shared data stored at the stacked memory dies. Due to the integration of the logic dies and the memory dies, the coherency manager can access shared data stored in the memory dies and perform related coherency operations with higher bandwidth and lower latency and power consumption compared to the external devices.