GRID on vSphere memory reservations

We are using K1 cards and are now testing P4/P40 cards. It used to be a requirement to have 100% VM memory reservation when using vGPU on vSphere, but I cannot find any reference to that in the latest vgpu user guide. Can anyone clarify or confirm that this requirement has dropped? Sharing 20% memory between our VM’s would be a great saving!

Nevermind my question, just tested this and I can confirm that you need 100% memory reservation, or the VM does not start.

Will this change in the future or is this a technical requirement?

You should ask VMWare for this. This depends on the Hypervisor …