I have done that before, primarily to get a single larger image for the "C" drive.
I found NO performance benefit that I could detect.
Synthetic sequential benchmarks will be impressive, and you might get level loads faster.
But for normal desktop operations, you will not see any performance improvement.
Against that, you have the negative of losing "trim", at least for now, and the increased exposure to a failure of either drive or the raid controller.
On a new build, my suggestion is to experiment with it both ways. Test in the real world to see if the benefits outweigh the negatives.
Use a stopwatch to time level loads, and see if you can feel any difference elsewhere.