Computational force allocation problem in mosaic mode

[461722946]

I wanted to use four A4000 pieces to splice together a 16x4K super large screen, and each A4000 would have to decode an 8k video to cover the screen. In the expansion mode of win10, this can be successful. But when I used Mosaic, all the calculations were concentrated on the first graphics card, which was severely underpowered and made the video stuttering. I found all the graphics cards, except the first one, to be extremely underutilized. How do I divide the computation and display tasks equally among all the cards?