Skip to content
DenseChip
Learn
PracticeCode labInterview

DenseChip

Navigate

Menu

LearnPracticeCode labInterview

Tools

All toolsSV editorCode repoResume
DenseChip

Practice, courses, and chip-design tools in one focused workspace.

LearnPracticeCode labInterviewToolsLearning paths

© 2026 DenseChip

AboutContactPrivacyTerms
Skip to lesson

Course contents

  • How to Use This Course
  • Reading Path
  • Cache Coherency Glossary
  • The Shared Course Model
  • Cache Coherency Revision Sheet
  • Why Coherence Exists8
  • States and Protocols12
  • Transactions12
  • Races and Controller Design8
  • Progress and the Fabric16
  • Ordering and Synchronization12
  • Performance and Scalability12
  • Real Protocols12
  • SoC Integration12
    • LLC Data and Directory Policies
    • Devices and DMA
    • The Memory-Controller Boundary
      • Point of Coherence and Serialization
      • Memory-Controller Reordering
      • Same-Address Hazards Below the Home
      • Memory Latency and Coherence
  • Power, Clocking and Reliability12
  • Verification12
  • Debug and Bring-up12
  • Interview and Design Review17
  • Capstone: A Coherent Model Lab5
Loading lesson

‹SoC Integration
  1. Topics
  2. Cache Coherency
  3. SoC Integration
  4. The Memory-Controller Boundary

Part 9 · SoC Integration · All levels

The Memory-Controller Boundary

Separate home coherence decisions from backing-memory scheduling and completion.

  • 4topics
  • 4lessons

H's buffer (early Put-Ack) or the memory controller's write queue can hold the current value while DRAM is old, and the controller can reorder commands to different addresses without changing any coherence permission. This topic separates H's per-line serialization from the controller's scheduling, shows the write-queue address check that keeps same-address order, and then attributes latency.

Lesson

After it you can...

Point of Coherence and Serialization

Name the domain, authority and completion boundary.

Memory-Controller Reordering

Separate independent scheduling from same-address obligations.

Same-Address Hazards Below the Home

Reject stale completion, overwrite and partial-update mistakes.

Memory Latency and Coherence

Attribute delay to the actual data and permission path.

Read after dirty writeback, writeback buffers and completion boundaries. The course default writes memory before Put-Ack. Once 'memory written' means accepted into a checked write queue (Point of Coherence lesson), H can issue a new read of X while an older write to X is still queued. The reordering lesson shows what keeps that safe.

Check yourself Can a DRAM scheduler decide a private write permission simply by servicing that address first?

No. H resolves the line's coherence obligations. A backing-memory service order is not by itself a private-cache permission grant.

Related topics

  • Shared course model

  • Prerequisite: default and early acknowledgment boundaries

PreviousNext: Point of Coherence and Serialization

Up next

Up nextPoint of Coherence and Serialization

Sign in to save progress

Practice this lessonReport an error

See also

  • LLC Data and Directory Policies
  • Devices and DMA