Two graphical devices on the ESXi host.
The first of the graphical device has vGPU Mode set to "None"
The second of the graphical device has vGPU Mode set to "Mixed Size".
Modifying the graphical devices from "None" to "Mixed Size" would fail as the change made from GUI does not modify any settings.
VMware vCenter Server 8.x
First check that the graphic card is supported by running the command on the ESXi: <graphical_device>-smi -q, where <graphical_device> is the name of the graphic device.
For example: nvidia-smi -q
The output should show that heterogeneous time-slice sizes is supported (this is just another way of saying mixed-size) as below. If it is not suported, vGPU heterogeneous mode will not enable.
vGPU Device Capability
Fractional Multi-vGPU : Supported
Heterogeneous Time-Slice Profiles : Supported
Heterogeneous Time-Slice Sizes : Supported
vGPU Heterogeneous Mode : Enabled
Check that the configuration of the GPU card in the “PCI device” and “passthrough. And disable passthrough on the card then restart the server.
After ESXi is reconnected to the vCenter, the GPU card can be changed to “Mixed size” mode.