Hi guys,
I want to create a traffic light and sign system. For the signal system I use UDIMs to toggle through the different states and symbols. This works totally fine, but I do have my concerns about the memory efficency - especially when moving over to the signs.
At the moment, for testing purpose, this is how the texture looks like for my signals:
As you can see, there is a lot of texture space wasted, but doing it this way it’s easier to manage … and here comes my question.
How do UDIMs work internally? In this case, I do have 27 files that get loaded as a single texture. There is a lot of black space that is unused. Is this space really wasted because internally it’s just one big texture, or is this just a representation of the texture?
Using virtual textures, the GPU just takes the parts of an texture that are needed and load these, right? So when there is a lot of unused space like here, does it tax the memory?
If not, this would be really important for me to know, because the next step is it to implement every sign that is available in germany. These are roughly 500 different signs.
In my dream scenario I create on big UDIM texture that can handle all scenarios. I do not need every sign to any time, but in case I need, I can maintain the correct order. So I wonder, in conjunction with virtual textures, can I do this? So just the parts that are needed get loaded and everything else is “gone”?.
Thanks for your help and have a nice day 
