576 frames of 4K uncompressed AVI would result in Gbs of data writing.
So this is not a "true" iClone render test.
Portion of the test is... just testing your system IO speed.
Copy resulted render file to the same drive and time it.
Then deduct that time divided by 2 (given read/write is at the same speed) from iClone render test.
That would give you a more or less true GPU render speed.
You can say I will render to MP4, it would be very small render output file and does not impact IO at all.
But here you have another burden - a compression which puts a toll on a render speed (one area which might need to be improved- only RL can tell).
And again amount of VRAM does not affect render speed. It is a speed of VRAM and GPU architecture plays a major role.