It may be something vROps can do, but that's a fairly specific capacity planning requirement. You could also just do the math in your head, which would be fairly easy assuming a pure allocation model. Even though that may be somewhat conservative, you won't have to worry about resource exhaustion. Then add 10% for good measure.
The numbers for memory calc are one thing, but you also need to factor in vCPU-to-pCPU subscription ratios when you compact those workloads. Don't know what those CPUs are or how many physical cores, nor do I know what types of workloads they're running and how tolerant they are of contention, but you may find yourself in a position where you actually need more hosts not from a memory density standpoint but for more physical cores.