"Practical limit" - inherently understood. However, the requirement would be to quantify the limit.
My thought would be to set up some controlled testing to discover possible or practical limits.
Baseline the current graphics solution as to the memory resources being used and then methodically work through incremental iGPU increases in the use of system memory/RAM.
E.g., 1 GB, 2 GB, 3 GB, etc.. Or whatever incremental ranges are allowed and appropriate.
Change only one thing at a time.
Then shift to some other viable configuration and again work through the available iGPU increases as before.
You can probably use Task Manager and Resource Monitor to collect data as you go.
May take a bit of trial and error at first - aka the "learning curve". Then go from there.
Should not be too difficult to reach a conclusion.