I am curious to know whether future graphics cards plan to have a dedicated raytracing queue for something like async raytracing like async compute or async transfer. As far as I know, hybrid raytracing is the approach that is the most feasible for games where you can transfer parts of rendering to raytracing pipeline like raytraced shadows which is a great candidate for asynchronous operations.

Would it be worth it to have a dedicated raytracing queue?

Isn't this already the case?

I have not used RTX myself, but looking here: http://on-demand.gputechconf.com/gtc/2018/presentation/s8521-advanced-graphics-extensions-for-vulkan.pdf

Both BVH build and tracing are regular commands, and likely the command buffers containing them work on any queue. So, in theory they can be executed async.

Should be similar in DX12?

Edit: Oh, likely the queue has to be a gfx queue... now i understand your question. Anyways, doing compute while RT should be possible already i hope, if it makes sense.