No, it does not. Memory allocated to VMs is directly from system RAM. The amount of RAM you need to have in your system per VM is the amount you are assigning plus a 20-25% overhead reserve for the emulated components. The exact amount of overhead you need to have available is an undecidable problem to solve, as it can range based on the use of the guest and the amount of total system memory you are reserving for the VM(s) you have.
As others have mentioned, we are not a competitor to ESXi or any of the other mainstream traditional hypervisors. They are designed to be SOLELY a virtualization host, focused entirely on the enterprise/cloud market, and are not user friendly. So there are many of the more traditional hypervisor features that we haven't built in just yet (like snapshotting) because while we want to do that, there are lots of nuances to doing that right when you support multiple filesystems and NAS functionality on top of the virtualization host capabilities. These are things that will come in due time. Same goes for vlan support.
I also highly recommend leveraging the completely free trial to evaluate the software for yourself. All questions relating to performance are pointless because everyone's hardware and network is different. It's far easier to try it yourself for free than for anyone here to guess at what your experience might be unless you are coming to the table with a replica infrastructure to another user. Keep in mind, this isn't me saying, "don't ask the question," but it is me saying, "don't trust anyone's answer until you test it for yourself."
Hope this sheds some light on things.
Hi Jon, many thanks for your reply. Once my GPU arrives this week I will give the trial version a go.