I haven't bothered profiling DAZ's various memory usage during operation, but there are a handful of useful things to reference that have only been touched on above, VRAM, Shared Memory, RAM, and Virtual Memory/Page File. I think it's useful to make a distinction because it think some comments and responses may be conflating VRAM and Virtual Memory, which are wildly different.
VRAM is the the RAM on your GPU, shared memory is a portion of your physical RAM that is made available as VRAM should an application (that is able to use it) need more video RAM than is available from the GPU. Shared memory is set by either drivers or bios, this is a configuration of your hardware as opposed to windows, and most systems make 50% of RAM available to be shared to the GPU. While still fast, shared memory will be slower to the GPU than its own VRAM.
Virtual memory, or the page file, is an amount of disk used to dump info from memory to disk in order to free up memory from without having to complete or terminate processes. My personal recommendation is to put this on your fastest disk, and hard set it to double your RAM + VRAM, which enables windows to fully dump what it's doing in memory to disk and fully use the RAM. While it's probably better than it used to be, I don't trust windows to handle this dynamically and would rather always have a contiguous file set aside in advance, though I'm not sure if windows cares about this being contiguous anymore with SSDs. The big catch here, this probably doesn't help your renders and paging really only benefits when swapping between threads.
TL;DR - Run DxDiag. On the system tab make sure that Memory is 2x the page file, though this only matters for separate threads. On the display tab note that the "approx. total memory" is what DAZ has available to it for a render. The VRAM can't be changed without swapping out your GPU, however you may be able to increase total memory by adding RAM and/or changing what is configured to be available as shared memory.
Alternate consideration, look into using canvases to render in multiple layers, which will reduce what needs to be loaded into memory for any single render.