Hi Jake,
You are running into a bug with the current beta where the cache is not shown correctly when the timeline and output resolutions are different. My guess is you have set your timeline resolution to 4K/5K, and this might explain the GPU RAM issues as well.
My recommendation would be keep your timeline and output resolution the same as your monitoring resolution, and you can always switch the timeline resolution to 4K/5K when you render. The windows, etc. will track automatically.
If you could confirm this solves the problem for you, it would be most appreciated!
Regards,
Rohit