What is the optimal configuration when building a virtual machine for a single-threaded Windows application?

A single-threaded application uses only one execution thread at a time, meaning additional vCPUs provide no performance benefit; using a uniprocessor VM reduces vCPU co-scheduling overhead in the hypervisor, which can actually improve performance and reduce latency for that workload.

Assigning multiple vCPUs to a single-threaded workload adds CPU scheduler complexity and co-scheduling overhead without any parallelism benefit, degrading rather than improving performance.

Single-threaded applications are inherently sequential by design and cannot be tuned to distribute work across SMP resources without rewriting the application code.

The hypervisor has no mechanism to make a single-threaded application exploit multiple CPUs; thread-level parallelism must be implemented at the application level.