Dr. Kodavasal will discuss work done at Argonne to develop both capability and capacity computing for CONVERGE engine simulations on supercomputing platforms. Here capability computing refers to running a single, extremely high-fidelity simulation on thousands of cores (for example, to generate gold-standard simulation results), while capacity computing refers to running thousands of medium- to high-fidelity simulations in parallel (for example, for sensitivity analyses, cyclic variability studies, or design optimization). The Argonne team, in collaboration with Convergent Science, addressed computational bottlenecks in I/O, communication, and load balancing to greatly enhance the performance of CONVERGE on both supercomputers and regular clusters. Based on these improvements, CONVERGE was scaled onto 4096 cores of Mira, Argonne’s IBM Blue Gene/Q supercomputer, for a single high-fidelity engine simulation. An improved, stiffness-based chemistry load balancing scheme for the SAGE detailed chemistry solver was developed, which reduced the runtime near ignition by a factor of three. A capacity computing workflow to enable seamless pre-processing, running, and post-processing of thousands of medium- to high-fidelity simulations on supercomputers and high performance computing (HPC) systems is under development. This process will provide a high level of automation and error resiliency. In addition to discussing advances in capability and capacity computing for CONVERGE, this presentation will provide recommendations on how to run HPC engine cases with CONVERGE.