AI summary by TwinMind
Why Synopsys Bought Ansys (For $35 Billion)
Asianometry
14m 7s
Synopsys-Ansys Acquisition Announcement
- January 2024: Semiconductor EDA giant Synopsys announced acquisition of software firm Ansys for $35 billion
- Ansys produces computer-aided engineering (CAE) software
- Acquisition driven by chiplet trend and intersection of design/simulation
Ansys Company History and Founding
- 1960s: John Swanson worked at Westinghouse Astron nuclear lab in Pittsburgh
- Worked on nuclear-fueled rockets for Mars missions (NERVA nuclear rockets)
- Performed stress analysis on components using finite element analysis (FEA)
- FEA breaks structures into smaller pieces for mathematical modeling
- Late 1960s: Nuclear rocket program canceled
- Swanson founded Swanson Analysis Systems
- Developed Ansys simulation software suite
- Westinghouse hired as consultant for financial support
- Used time-shared mainframe owned by US Steel
- Name "Ansys" chosen because lawyers confirmed no trademark violation
- 1970: Ansys software completed, Westinghouse first user
- Software ported to various computers (VAX) with added features
- 1994: Swanson sold majority control to private equity firm TA Associates
- Company had ~$30 million annual revenues
- Went public 2 years later, renamed to Ansys
- Swanson retired, focused on philanthropy
Computer-Aided Engineering (CAE) Overview
- CAE uses sophisticated algorithms to simulate, validate, solve engineering problems
- Different from CAD (Computer-Aided Design):
- CAD: produces design
- CAE: tests design performance under conditions
- Benefits: diagnose/troubleshoot without extensive real-world testing
- Saves millions of dollars (modern plane/car prototypes)
- Enables testing conditions difficult to replicate (wind tunnels)
- More powerful computers/cloud computing enable simulation of increasingly complex designs
Computational Fluid Dynamics (CFD) Development
- CFD: major CAE subcategory solving nonlinear equations governing fluid motion
- Fluids vs solids: react differently to shear stresses, change shape without losing volume
- CFD equations extremely complex, must maintain multiple factors simultaneously
- Conservation of mass and momentum in multiple directions
- Particularly important for heat transfer analysis
- Track heat sources in structures
- Model heat conduction/radiation
- Design heat removal using fluids (air)
Historical CFD Development
- 1922: Lewis Fry Richardson published "Weather Prediction by Numerical Process"
- Proposed differential equations for weather prediction
- Fantasized about "forecast factory" with human computers
- 1940: Existing analytical fluid dynamics models fragmented/inadequate
- Based on Euler and Lagrange principles
- Elegant but didn't match reality
- Manhattan Project urgency drove development
- Need to track shock wave behavior/propagation
- 1950: Robert Richtmyer and John von Neumann introduced "artificial viscosity"
- Simplified equations for computers while maintaining physics fidelity
- Los Alamos team led by Frank Harlow contributed foundational CFD algorithms
- Particle-in-cell and fluid-in-cell methods
- 1967: Professor John Chu (Columbia) coined term "computational fluid dynamics"
Advanced Packaging and Heat Management
- AMD Mi300 AI accelerator example of advanced packaging
- Over 100 pieces of silicon in 3 layers on interposer
- Connected via through-silicon vias (TSVs) with copper interconnects
- Enables larger, more powerful silicon without major yield impact
- Compute chiplets: advanced nodes; IO: trailing edge fabs
Heat Management Challenges
- Traditional 2D monolithic ICs: heat dissipated through substrate/heat sink
- Vertically stacked packages (>2 silicon pieces): limited heat sink access
- Silicon pieces have different heat profiles
- Heat travel limitations create local hotspots
- Higher temperatures hurt performance and lifetime
- Heat consequences: fatigue interconnects, warp substrate, crack die
Heat Control Solutions
- Software level: code limiting chip performance to prevent overheating
- Packaging level:
- Immerse chip in heat-dispersing fluid
- Create silicon fins from interposers for heat shuttling
- Modify solder ball interfaces between dies
- Change filling between balls for improved heat distribution
Simulation Importance in Semiconductor Design
- Wafer production: 3-5 months (excluding packaging)
- Complete package testing/validation takes additional months
- Advanced chip design time constraints make simulation crucial
- CFD thermal simulation benefits:
- Calculate heat transfers between package components
- Requires local fluid temperature knowledge
- Must produce accurate results matching real-world experimentation
- Must be intuitive for time-pressured designers
Industry Partnerships and Market Position
- CFD programs for electronic package thermal simulation: mature (20+ years)
- 2021: Ansys partnership with TSMC for 3D-Fabric thermal analysis
- 3D-Fabric: TSMC's advanced packaging technologies umbrella
- Import package from EDA software for electrical/thermal/structural simulations
- TSMC approval signals Ansys growing semiconductor design presence
Future Implications and Market Trends
- Generative AI integration into design flow
- Simulations validate AI-generated designs (similar to Python runtime validating ChatGPT code)
- $35 billion acquisition price indicates heterogeneous integration/chiplet approach permanence
- Mi300-style approach expected to spread beyond AI chips to broader semiconductor space
- Exception: smartphones
- Trailing edge fab capacity growth globally
- Users will mix/match process nodes with advanced packaging
- More complex advanced packages create additional challenges beyond heat
- Capacitance potentially harder problem than heat management
- Incentivizes better tools for advanced package design/manufacturing
