how to set up the highest application density test platform?

I'm working on a system controling embedded devices or let's say IOT stuff.

The new server was designed to handle much more devices than it used to. Now, I would like to find out its limits. For that we have a "simultor.exe" of the embedded device software running on a PC (native windows exe).

What is the best strategy to get the highest application density on a single PC. This program is not really hard CPU consuming, but it seems that after 1000 instances on a single PC (VM or either real hardware) it starts to be tough for the OS (windows 10 professionnal). Is it better to have console apps, windows apps?

I believe that we may have reached the scheduler's limits acting like that. Do you have any related experience? What are your conclusions?

How did/would you solve that?

I've no experience with docker and all container sexy tools, but would it really help if the problem is the scheduler? (I precise that I would like to handle that on our local servers)

Shall I better use VMs, Windows core in VMs, real machines with more processes?