Thermal analysis for 2.5D & 3D integration
From package geometry to thermal insight. Explore chiplet architectures, locate hotspots, and focus detailed analysis where it matters.
Actual modeled package and thermal results. Vertical dimensions exaggerated for readability.
3Dblox packages and supported geometry-file workflows, with automated model preparation.
Physics-based numerical analysis and ThermStack-GNN for repeated design-space exploration.
Interactive 2D/3D temperature maps, cross-sections and workload comparisons.
Capabilities
Package architecture, power allocation and cooling choices are linked. ThermStack brings them into one analysis workflow, from early screening to focused investigation of the most critical regions.
01 / NUMERICAL FIDELITY
Screen thermal behavior on a practical package grid, assess sensitivity to refinement, then invest in finer analysis where thermal margins demand it. FEM provides a numerical verification path.
Numerical analysis02 / DESIGN EXPLORATION
Compare many power and cooling scenarios with a package-trained model. Rapid temperature predictions make repeated design studies practical within the validated operating range.
Fast exploration03 / INPUT FLEXIBILITY
Bring a native 3Dblox package or a supported geometry-file workflow. Automated preparation connects the physical structure to materials, power inputs and cooling assumptions.
Geometry to analysis04 / VISUAL INSIGHT
Locate hotspots on the real structure, inspect buried layers and compare workloads. Interactive 2D/3D views support design debug as well as communication across teams.
Visual explorationFrom input to decision
Keep geometry, operating conditions and temperature results together. Reuse the same physical context as the design moves from architecture exploration to detailed review.
Chiplets, memory stacks, interposers and substrates, with their materials, interfaces and dimensions.
Evaluate power budgets and cooling conditions, identify limiting regions and compare candidate scenarios.
Use temperature fields and component-level results to decide what to change and where to look more closely.
White papers
This page stays short on purpose. Methods, benchmarks, figures, tables and their scope notes live in a growing series of white papers — each a self-contained document you can read, share and cite in a design review.
WP-101 / NUMERICAL ENGINE
The physics-based workflow — screen the package, refine with purpose, verify with FEM — plus model inputs, automated 3Dblox preparation and transient analysis of time-varying workloads.
Read WP-101WP-102 / APPLICATION STUDY
An MI350-inspired compute + HBM assembly under two 950 W workloads, a measured 250 µm vs 50 µm resolution comparison, and interactive 2D/3D result views on the modeled structure.
Read WP-102WP-103 / AI SURROGATE
A package-trained model for repeated exploration: the 360-case benchmark against FEM, how to use rapid predictions responsibly, and where the numerical check belongs in the loop.
Read WP-103WP-10X / IN PREPARATION
Forthcoming papers in the same format will cover transient benchmarking and agentic design-loop integration. Ask to be notified when a new paper is published.
Get notifiedStart with your package
Discuss your geometry inputs, workload scenarios and accuracy targets with the ThermStack team.