Disclosed are various embodiments of real-time simulation of the performance of a compute accelerator workload for distributed resource scheduling. The compute accelerator workload is executed on candidate hosts to select a destination host. Efficiency metrics are determined for the candidate hosts based on the execution of the compute accelerator workload on the candidate hosts. A destination host is selected from the candidate hosts based on the efficiency metrics, and the compute accelerator workload can be assigned to execute on the selected destination host.