You don’t really need a large amount of VRAM unless you want to use GPU ultra feature extraction. With my NVIDIA GForce RTX 3060 8GBVRAM I can do GPU feature extraction with my images to 4050pixels in the long dimension.
See: Image size change for Ultra feature extraction?
Some say using ultra is unnecessary, and in any case, CPU feature extraction can give better results.
However, in dense reconstruction of the point cloud the GPU is used, but with very low VRAM required, so a 4GB card will work just as well as one with 24GB VRAM, if they had the same number/speed CUDA cores.
I get a [WARNING], that my video card is recognized, and the CUDA, is recognized, but WebODM, cannot use my CUDA(?), and ask if my driver is up to date … I checked the NVIDIA site, and I’m informed I have the latest driver.
I don’t have a problem getting another video card, but I want two things …
The use of my 4K monitor(I’m only getting 2K), and the use of CUDA, which should speed up that stage of computations.
I just think there should be a way to know if one, or the other might work, before I buy it.