Geometry viewers are rendered ‘in-place’ onto the screen. They aren’t an offscreen render where the results are displayed on the screen the way a Render TOP (indeed, all TOPs) are.
The amount of pixels they take up on the screen is in direct proportion to their rendering cost. This is different than any TOP, where the size of the TOP viewer has nothing to do with the cost of the TOP.
A Geo Viewer open means the geometry data has been uploaded to the GPU also. If you close the viewer so the geometry is no long visible anywhere (and it’s render flag isn’t on), then the GPU memory will be freed.
That is to say if you have a chain of 10 SOPs, and all of their viewers are open, then you are taking up 10 times as much GPU memory as you would if you had only one viewer open (assuming the geometry is about the same in each SOP).
I was aking cause you once recommended to only use a single renderTOP, and for additional renders we should use renderpass TOPs. Is this still valid? It would make my setup much easier if I could have 500 render TOPs (only 3 are active at a time, the others are disabled (render parameter = 0) and have their resolution reduced to 2x2)
Or I could have 1 master renderTOP and 500 renderpassTOPs (2 of them active). That is, if all of them can target the same master render TOP? I vaguely remember that they need to form a chain which wouldn’t be possible. Is that still the case?
All the limitations from the Render TOP are still the same as they used to be. I know it’s annoying so am trying to find a good way to solve them, but it’s not done yet.