Memory Address Isolation Mapping for Uniform Access Across Compute Dies

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In server processing systems with dis-aggregated dies, high data access latencies and interconnect power consumption occur due to multiple data movements across the interconnect fabric and EMIB boundaries, leading to increased memory bandwidth and die-to-die interconnect bandwidth consumption.

Innovation Solution

Implementing a uniform memory access (UMA) architecture with isolation domains by mapping system memory physical addresses to quality of service (QoS) and security isolation domains, using address decoder tables to facilitate data communications and maintain consistent access latencies across compute dies.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If dis-aggregated dies are used to increase core count, then processing capability is improved, but data access latency increases

Engineering Contradiction:
Improveprocessing capabilityVSAvoiddata access latency
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent segments the memory address space into multiple isolation domains (e.g., domain 0, domain 1, domain 2) and maps different physical memory ranges to each domain. This segmentation allows each compute die to have dedicated memory regions, reducing the need to access remote memory through the interconnect fabric and thereby lowering data access latency while maintaining high processing capability through multi-core architecture.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an address decoder table as an intermediary component that translates memory access requests into isolation domain identifiers. This intermediary enables the system to efficiently route memory accesses to appropriate isolation domains without traversing the entire interconnect fabric, thus reducing latency while preserving the benefits of dis-aggregated die architecture.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Quantity of substance

If multiple data movements across interconnect fabric are used to access memory, then memory bandwidth is improved, but interconnect power consumption increases

Engineering Contradiction:
Improvememory bandwidthVSAvoidinterconnect power consumption
Core Design Contradiction:
Quantity of substanceVSUse of energy by moving object

Solution Approach 1:

The patent assigns local memory regions to specific isolation domains, allowing compute dies to access their local memory directly without traversing the interconnect fabric. This local quality approach maintains high memory bandwidth for local accesses while significantly reducing interconnect traffic and power consumption for remote accesses.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

By segmenting the address space into isolation domains with dedicated memory regions, the patent reduces the volume of data that needs to traverse the interconnect fabric. This segmentation enables efficient local memory accesses while preserving overall memory bandwidth capability, thereby reducing interconnect power consumption without sacrificing memory performance.

Inventive Principle:
Principle #1Segmentation

3Adaptability or versatility

If multiple data movements across EMIB boundaries are used, then memory access capability is improved, but die-to-die interconnect bandwidth consumption increases

Engineering Contradiction:
Improvememory access capabilityVSAvoiddie-to-die interconnect bandwidth consumption
Core Design Contradiction:
Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent segments the memory address space into isolation domains and associates specific physical memory ranges with each domain. This segmentation enables compute dies to access their designated memory regions directly, reducing the need for data movements across EMIB boundaries and thereby lowering die-to-die interconnect bandwidth consumption while maintaining versatile memory access capability through the isolation domain mapping mechanism.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The address decoder table serves as an intermediary that translates memory addresses into isolation domain identifiers, enabling efficient memory access without requiring multiple data movements across EMIB boundaries. This intermediary approach maintains versatile memory access capability while reducing the bandwidth consumption on die-to-die interconnect.

Inventive Principle:
Principle #24Intermediary (Mediator)

4Reliability

If isolation domains are implemented for QoS and security, then access control is improved, but address mapping complexity increases

Engineering Contradiction:
Improveaccess controlVSAvoidaddress mapping complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent introduces an address decoder table as an intermediary component that handles the complexity of address mapping to isolation domains. This intermediary absorbs the complexity of maintaining QoS and security isolation, presenting a simplified interface to the rest of the system while ensuring reliable access control through the isolation domain mapping mechanism.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS20260064605A1Techniques associated with mapping system memory physical addresses to isolation domains for uniform memory access by a system
Publication Date: 2026.03.05 INTEL CORP
  • US20260064605A1 patent drawing
  • US20260064605A1 patent drawing
  • US20260064605A1 patent drawing

AI summary

Examples include techniques associated with mapping system memory physical addresses to isolation domains for uniform memory access (UMA) by a system. Examples include mapping separate system memory physical addresses ranges associated with memory devices communicatively coupled with at least one compute die of the system through an input/output (I/O) die of the system. The separate system memory physical addresses to be mapped to isolation domains and address decoder information is generated to indicate the mapping of the separate system memory physical address ranges to the isolation domains.