Input is a 4k 30FPS MP4 video. I even tried setting the downscale height to 1080 but I still get a memory error CUDA out of memory. Tried to allocate 558 MiB (GPU 0; 8.00 GiB tocal capacity; 5.06 GiB already allocated; 390.22 MiB free; 592.32 MiB cached).

Problem is that this is a pretty high end card, my system has 96GB of RAM and I have basically nothing running. The GPU memory is less than .8 GiB before I start the program. I am using Alpha .39 because .40 does not recognize my video card. Also tried .38.

Why not use split frames option?

So who has a video card with 40GB of dedicated GPU memory? Why not shared memory? I will look at the split frames option but just to do a basic HD video and having to have a top of the line video card is not going to be very useful for most end users.