Skip to main content
Concept guide · ISO 26262-5 + -11 · 12 chapters

Multi-Core Safety Architectures

Start from one core with no independent check and build up to lockstep, split-lock, monitored cores, safety islands and self-tests, learning how automotive silicon detects its own random hardware faults and what evidence the hardware metrics need.

Chapters
12
Chapters
Figures
13
Figures
Self-test families
3
Self-test families
The short version

Quick answers

What is multi-core safety in ISO 26262?
Multi-core safety covers the architectures and analyses that let processor silicon detect its own random hardware faults well enough to support safety requirements with a high ASIL. A single core has no independent check of its own results, so faults that corrupt computation can go undetected unless a second channel, a monitor or a self-test catches them. ISO 26262-5 defines the hardware architectural metrics and the evaluation of random hardware failures that quantify how well this works, and ISO 26262-11 adds guidance specific to semiconductors.
What is dual-core lockstep and how does split-lock differ?
Dual-core lockstep runs two cores on the same instruction stream, usually with a small time offset, and compares their outputs so that a random fault in one core shows up as a mismatch. The price is area and power, because two cores deliver the performance of one. Split-lock silicon can be configured either as a lockstep pair or as two independent cores, trading checking for throughput. Many real chips mix both, and the guide explains what changes between the modes.
Why this course · ISO 26262, Part 5

Why it pays for itself

From architecture to evidence

Follow the chain from lockstep, split-lock, monitored cores and safety islands to the scoped coverage claims and hardware metrics that the safety case relies on.

Right-size the redundancy

Compare delayed lockstep, asymmetric monitoring and safety islands on coverage, silicon area, performance and availability, so you meet the targets without over-building the SoC.

Interference analysis that holds up

Shared caches, interconnects and memory controllers let one core disturb another without any fault. Learn the interference paths and the silicon features that make freedom from interference arguable.

After the course

What you’ll be able to do

Explain processor fault detection

Describe how redundancy, monitoring and self-tests detect random hardware faults in a core.

Compare architectures

Weigh lockstep, split-lock, monitored cores and safety islands on coverage, cost and availability.

Plan self-tests

Choose self-test families and intervals for faults that normal operation does not exercise.

Argue freedom from interference

Identify shared-resource interference paths and the features that mitigate them.

Scope coverage claims

State what each coverage number covers and how fast detection happens.

The curriculum · 12 chapters

Chapter by chapter

  1. 01

    Why Processor Diagnostics Need Evidence

    See why one core cannot check itself independently, what the standard asks of the compute element, and why the scope is random faults.

    • No independent check
    • Shared fault demo
    • Random faults, not bugs
  2. 02

    Random Hardware Faults in Silicon

    Follow the chain from fault to failure, the three fault behaviors over time, and where faults hide inside a core.

    • Fault to failure chain
    • Three fault behaviors
    • Where faults hide
  3. 03

    The Redundancy Toolbox

    Learn the core idea of independent redundancy plus a judge, the difference between comparing and voting, and why diversity matters.

    • Redundancy plus a judge
    • Compare versus vote
    • Common cause and diversity
  4. 04

    Dual-Core Lockstep

    Understand master, checker and compare unit, the role of the delay, and what makes a lockstep coverage claim defensible.

    • Master, checker, compare unit
    • Delayed lockstep
    • What lockstep misses
  5. 05

    Lockstep vs Split-Lock

    See how the same two cores can be wired for checking or for throughput, and how real chips mix both modes.

    • Cost of lockstep
    • Two wiring modes
    • Choosing a personality
  6. 06

    Beyond Lockstep: Monitored Cores

    Compare an asymmetric diverse supervisor and the main-plus-monitor pattern with lockstep on detection, safe state and cost.

    • Asymmetric supervisor
    • Main plus monitor
    • Costs side by side
  7. 07

    Safety Islands

    Learn the anatomy of a safety island, the jobs it does on a large SoC, and the problems it does not solve.

    • Anatomy of an island
    • Four jobs of an island
    • What an island does not solve
  8. 08

    Core Self-Tests: SBST, LBIST, MBIST

    Cover faults that need deliberate exercise with three self-test families, and see what the detection interval actually bounds.

    • Three self-test families
    • Detection interval
    • Software test library duties
  9. 09

    Shared Resources and Interference

    Map the interference paths that consolidation adds and the silicon features that make freedom from interference arguable.

    • Three kinds of interference
    • Contention map
    • Silicon features for FFI
  10. 10

    Putting Numbers On It

    Give every coverage claim a scope, relate the three hardware metrics to the architecture, and check detection is fast enough.

    • Scoped coverage claims
    • Three hardware metrics
    • Detection latency
  11. 11

    Choosing an Architecture

    Use a short set of driving questions and a side-by-side comparison to pick a processor safety architecture.

    • Questions that drive the choice
    • Mechanisms side by side
    • Rules of thumb
  12. 12

    Pitfalls and Checklist

    Avoid the common traps of multi-core safety arguments and check integration before you rely on the silicon.

    • The root mistake
    • Traps and fixes
    • Integration checklist

Read all 12 chapters with Expert.

Start with a free account: the complete Functional Safety Concept guide and 5 practice exams a month, then upgrade when you need this guide.

Expert plan pricing
Diagrams & visuals

Not just text: the visual toolkit

Shared fault defeating an internal check

Why a check inside the same core can miss a common fault.

Delayed lockstep cycle view

Two cores running with an offset into a comparator.

Lock mode versus split mode wiring

The same cores configured for checking or for throughput.

Safety island floorplan

Where an island sits on a large SoC die.

Detection latency ladder

How quickly different mechanisms detect a fault.

Built for

Who this guide is for

  • SoC and ECU architects selecting between lockstep, split-lock and safety islands
  • Hardware safety engineers building coverage arguments for a multicore SoC
  • Software engineers mapping ASIL and QM workloads onto shared cores
  • Integrators responsible for the software test library a chip vendor assumes

Frequently Asked Questions

Common questions about Multi-Core Safety Architectures

The single-point and latent fault metrics and the PMHF are evaluated for the hardware of the item, and their targets tighten from ASIL B to ASIL D. High targets usually require an independent check on the compute path, such as lockstep or a diverse monitor, while latent faults in the checking hardware are addressed with self-tests. Every coverage claim needs a defined scope and a detection time that fits the safety requirement.
Through interference: one core can evict another core’s cache lines, saturate the interconnect or monopolize the memory controller, stretching execution time until deadlines are missed without any fault occurring. Mitigations combine memory protection, partitioning of shared caches, bandwidth regulation and timing analysis under worst-case contention. The guide maps these interference paths and how they touch every other mechanism.
The guide has 12 chapters. It starts with why processor diagnostics need evidence and how random faults arise in silicon, then covers redundancy, dual-core lockstep, split-lock, monitored cores, safety islands, the SBST, LBIST and MBIST self-tests, and shared-resource interference. It closes with the metrics, choosing an architecture, and a pitfalls checklist. It includes 13 figures. A free account starts you off, and the Pro and Expert plans unlock more of the library.

Start the course today

A free account unlocks one full concept guide, 3 work product templates, 1 guided process, the Markov and RBD simulators, and 5 practice exams per month. The Pro and Expert plans unlock more of the 93-guide library. No credit card required.

Continue with

Where this topic leads next.

Browse all 93 guides
  • System levelPower & Clock SupervisionThe two assumptions every other safety mechanism silently depends on: valid supply rails and a trustworthy time base. Power trees, safety PMICs, voltage monitors, sequencing, reset escalation, watchdogs, clock monitors, the independence argument, timing and energy budgets, latent-fault self-test, and the FMEDA rows they all feed.Expert plan pricing14 chapters
  • System levelCalibration & Parameter SafetyThe same binary, a thousand behaviors: configuration data vs calibration data, ISO 26262-6 Annex C requirements, A2L and XCP workflows, CRC and plausibility protection in the ECU, dataset release, variant coding risks and tool confidence.Expert plan pricing12 chapters
  • HardwareHardware Engineering in ISO 26262Learn the complete hardware development lifecycle per ISO 26262-5 from requirements through design and verification. Covers HSI specification, SPFM/LFM/PMHF achievement, dependent failure analysis, and ASIC/FPGA/PCB considerations.Expert plan pricing9 chapters