TSMC, Ansys, and Microsoft have announced the successful completion of a pilot program aimed at accelerating the simulation and analysis of silicon photonic components. The collaboration reportedly achieved a tenfold speed increase in Ansys’ Lumerical FDTD photonics simulation software through the utilization of Microsoft’s Azure NC A100v4-series virtual machines.
These virtual machines, powered by NVIDIA accelerated computing on Azure AI infrastructure, were used to execute simulations and identify optimal resources balancing cost and performance, according to the collaborators. The companies stated the pilot resulted in seamless deployment, graphical interface access, distributed simulation scaling, and efficient post-processing of large datasets within cloud environments.
Stefan Rusu, head of silicon photonics system design at TSMC, emphasized the challenge of simulating all possible parameter combinations given the size and complexity of their multiphysics silicon solutions. He noted that this collaboration provides accurate solutions in a significantly reduced timeframe.
The companies view this pilot program as a key solution for the design and fabrication of silicon photonic integrated circuits (PICs). They explained that the silicon PIC workflow is vulnerable to minor errors that can disrupt chip continuity, leading to increased costs and project delays of several months. By deploying Lumerical FDTD on the cloud, designers can better identify optimal chip designs, accounting for the complex challenges of integrating photonic and electronic circuits.
Leave a Reply