Displacement map hi-res render

[hellobard]

Hi,
I have scene with a displacement map on the floor. Rendering at 700x390, the scene renders in 1 minute. When I render at 1280x720 the light cache stage takes forever, and after 10 minutes finally ends up crashing Maya. Obviously upping the resolution will add to the render time, but this is absurd.

When I render without the displacement map, the scene renders just fine at both resolutions.

Is the displacement map render output size dependent? Should I be altering the settings for my DM for a higher resolution? If anyone has a good tip on how to turn down the render time, I would be very happy to hear it

[hellobard]

Alrightee, fixed it by setting my Max Subdivs in the Default Displacement and Subdivison part of Settings from 256 to 12.
720p resolution renders in a minute now ^___^

PS: what's the advantage of having 256 set as default? Isn't it insanely high?

[joconnell]

256 is a big max alright, but what's more important is the edge length. What it does is literally tell vray how long in pixels you want the edges in your final mesh to be. If the edges are more than this value, they get subdivided in half until they're the same value as you've specified. If you've got displacement settings that work well for a small res render, then when you move to a higher res, you can probably turn down the edge length to keep similar quality.

In terms of it being size dependant, the main parameter is "view dependant". If you have this turned on, the edge length that vray is measuring is done in pixels as the edges appear in the final render. This means that if your edge length is set to 4.0, vray will keep subdividing edges in your mesh until they are 4 pixels long in the final render. The main benefit of this is that edges that are further away get less subdivisions, you won't see any difference in the detail anyway. If you have view dependant turned off, it means that vray uses world units to measure your edges. This means that every edge in your model is measured in world units and again subdivided until they match the edge length value or less. The problem is that with this method, an edge in front of the camera will be treated the same as an edge much further away. The problem with this is that the far away parts of the mesh will be much higher quality than you can see and are going to use lots of memory and time.

In terms of having 256 as a default, it's okay as vray more than likely won't use subdivisions that high except in very rare circumstances. Once the correct edge length has been reached, vray will stop subdividing the edges so it might only use 24 subdivisions rather than always going up to the max 256. If you have a flat plane that filled the entire screen and it only had two segments so that each edge was half the length of your full rendered frame, then vray might go up to 256 subdivs. In cases like that you're better off adding in geometry at the start with a turbosmooth or similar though, it ends up being less intensive than only relying on the vray displacement.

[hellobard]

thank you very much taking the time to write that incredibly detailed and illuminating answer, joconnell! I really appreciate it

[joconnell]

Good karma