Cases
Curator™ EDA Sign-Off Test Report
Curator™ is a next-generation AI-based reliability and power sign-off tool, supporting advanced processes down to 5nm and below. It integrates cutting-edge technologies in sparse matrix, machine learning, symbolic computing and etc. A typical result comparison between Curator and RedHawk is listed below by a real design based on TSMC N5 process (disclose permission obtained from our partner).
IR-Drop Heatmap (RedHawk VS. Curator)
IR-Drop Comparison Table (RedHawk VS. Curator)
Power Heatmap (RedHawk VS. Curator)
Power Analysis
Power Unit: W
Total Power Deviation with RedHawk-sc
Curator™ VS. RedHawk-SC Summary
• Global Signal EM: The average current simulated by Curator (for more than 60% of the Signal Lines) is 30%-50% larger than that of RedHawk.
• Global Power: The overall power consumption simulated by Curator is significantly larger than RedHawk (by approximately 90%). The power distribution trend is similar for both tools.
• Static/Dynamic IR: The overall IR-Drop simulated by Curator is 60% higher than RedHawk’s, with the difference primarily arising from both Signal EM equivalent current calculation and matrix solving. The global distribution trend is similar for both tools.
Curator™ Simulation Flow and Engines
Key differences between RedHawk and Curator™:
1. RC extraction engine (Built-in vs. External Golden StarRC for advanced nodes and Self-Developed for the mature nodes)
2. Signal EM simulation engine (Built-in vs. Self-Developed)
3. Matrix solver engine (Built-in vs. Self-Developed)
Reasons why Curator’s results differs from RedHawk’s:
1. Curator’s RC extraction engine recommends using StarRC golden signoff for advanced nodes.
2. Curator’s Signal EM simulation engine is endorsed by the SPICE golden standard, allowing for independent random sampling and comparison with SPICE for signal lines.
3. Curator’s Matrix solver engine is validaded and endorsed by the MATLAB golden standard, allowing for independent matrix solver comparison with the MATLAB engine.
Curator internal AI-based KL engine
Curator internal AI-based KL engine performs simulation at arbitrary process, voltage, and temperature (PVT) corner without re-characterization.