Mobile Menu

Benefits

  • Links performance, power and thermal behavior in a single model, driven by real workloads and traffic.
  • Enables early cooling-system specification using the cooling time parameter (time to cool by 1 °C), rather than relying only on late-stage thermal tests.
  • Identifies hotspots, thermal bottlenecks and unsafe operating regions long before hardware exists.
  • Supports what-if exploration of power states, floorplans, materials and cooling technologies at system level.
  • Reduces risk of over-designing or under-designing cooling, improving both cost and reliability.

The Thermal library in VisualSim Architect provides a system-level method to convert activity-driven power into temperature for each component and for the overall SoC or system. It closes the loop between performance, power and cooling, enabling architects to see thermal effects during the earliest stages of design.

The flow has three tightly linked parts:

  1. A performance model is built using power tables that define the power consumed by each component in each state (active, idle, sleep, failure, etc.).
  2. During simulation, VisualSim generates a time-synchronized power profile for every component and for the entire system.
  3. The Thermal blocks use this power profile, along with material and packaging attributes, to compute heat (Joules) and temperature (°C) for each component and the system, considering the cooling capacity and time constant of the cooling solution.

A key parameter of the Thermal block is the cooling level, defined as the time it takes for the system to cool by 1 °C under a given cooling solution. This parameter directly drives the cooling system design specification (heatsink, fan, cold plate, liquid cooling, airflow, etc.).

Overview

The Thermal library provides parameterizable blocks that:

  • Consume per-component power tables from VisualSim performance models, indexed by operational state and activity.
  • Use the simulated power trace to compute instantaneous and accumulated heat in Joules for each resource (core, memory, NoC, PHY, device, board, rack).
  • Convert heat into temperature rise using thermal-material attributes such as thermal resistance, thermal capacitance, density and specific heat.
  • Model cooling behavior using a cooling parameter that describes how fast the system can remove 1 °C of heat, mapping to specific cooling technologies.
  • Provide hierarchical thermal modeling at die, package, board and enclosure levels.

The result is an activity-driven thermal profile that shows hotspots, temporal heating, cool-down behavior and thermal interaction across components, driven directly by real workloads and system traffic.

Supported Features

  • Power-Table Integration:
    • Per-state power (active, idle, sleep, failure, etc.) for each block.
    • Support for voltage/frequency dependent power and activity-based scaling.
  • Per-Component Thermal Modeling:
    • Thermal behavior for each core, accelerator, memory, interconnect, PHY and I/O.
    • Aggregation to die, package, board, chassis and rack-level temperature.
  • Material- and Structure-Aware Equations:
    • Parameters for thermal resistance, thermal capacitance and conductivity.
    • Support for different materials, stack-ups and packaging concepts.
  • Cooling System Abstraction:
    • Cooling parameter defined as time required to remove 1 °C of heat.
    • Behavioral mapping to heatsinks, fans, cold plates, liquid cooling and airflow.
  • Time-Domain Thermal Response:
    • Transient heating and cool-down curves under changing workloads.
    • Thermal overshoot and settling behavior for workload bursts.
  • Thermal Coupling and Hotspots:
    • Coupled thermal behavior between neighboring blocks and stacked dies.
    • Identification of hotspots and thermal bottlenecks.
  • Hooks for thermal throttling, DVFS and safety shutdown based on temperature thresholds.

Key Parameters

  • Power_Table_Ref – Reference to the power table used for each component and state.
  • Thermal_Resistance (Rth) – °C/W for each component, package or path.
  • Thermal_Capacitance (Cth) – J/°C capturing heat storage capacity.
  • Material_Properties – Density, specific heat, conductivity and related attributes.
  • Cooling_Time_per_1C – Time required for the system to cool by 1 °C under a given cooling solution.
  • Ambient_Temperature – External environment temperature (°C).
  • Coupling_Factors – Thermal coupling coefficients between neighboring blocks or dies.
  • Max_Operating_Temperature – Safe operating limit per component.
  • Throttle_Thresholds – Temperature thresholds for DVFS, clock gating or shutdown.
  • Sampling_Interval – Time step for thermal calculations and logging.

Analysis and Reports

Example analysis outputs include:

  • Time-series plots of temperature per component, per die and for the overall system.
  • Heat (Joules) accumulated over time for each component and subsystem.
  • Hotspot identification and ranking of the warmest components and regions.
  • Impact of different cooling levels (Cooling_Time_per_1C) on maximum temperature and margin.
  • Correlation of workload phases with thermal peaks and throttling events.
  • Trade-off curves across performance, power, temperature, reliability and cooling cost.

Applications

  • SoC and Semiconductor Design
    • Thermal analysis of multi-core CPUs, GPUs, NPUs, AI accelerators and memory subsystems.
    • Study of 2.5D/3D IC, chiplet and stacked-die thermal behavior.
  • Boards, Systems and Racks
    • Server boards, network switches, base stations and edge devices.
    • Impact of cooling design on throughput, latency and power management.
  • Automotive, Aerospace and Industrial Systems
    • ECUs, mission computers, power electronics and mixed-signal modules.
    • Thermal behavior under harsh ambient conditions and mission profiles.
  • AI and High-Performance Computing
    • GPU and accelerator clusters with dense, high-power workloads.
    • Thermal impact of AI workloads, batch sizes and utilization strategies.
  • Cooling System Specification
    • Sizing heatsinks, fans, cold plates, airflow and liquid-cooling capacity.
    • Comparing cooling options against performance, cost and form-factor targets.

Integrations

The Thermal library integrates with:

  • Power and State Modeling – Power tables and power-state models for every block.
  • Processor, Memory, Interconnect and Accelerator libraries for per-block thermal profiles.
  • Traffic and Workload Modeling so real workloads drive power and thermal behavior.
  • Power Management and DVFS to implement temperature-aware throttling.
  • Electrical & Mechanical Systems for combined thermal and mechanical stress studies.
  • External thermal/CFD tools via data export for deeper physical-level analysis (where used).

Schedule a consultation with our experts

    Subscribe