koobas.hobune.stream
Beast Lightmapping, Baking texture step ( past GI calculations)
Hey, I'm finding that the act of writing Unity/Beast writing down the gi calculations into a lightmap takes WAY longer than doing the calculations themselves. I've always seen on videos when the status bar is shown during the process than GI calculations always take a lot longer than baking ( although with speedups in video I can't really be sure 100%) but for me, the GI calculations are probably taking 10% of the time of a bake. I don't think that I have crazy settings, but I've tried from extremely to high enough and it's always the baking texture step ( not the edge dilation or whatever that final step is called by Beast before Unity imports the lightmaps.) Is there something I can change somewhere? Would it run faster if Unity was installed on an SSD? It's been killing me for a while and I would really love to find out if it's something that everyone experiences or not ( Otherwise, I'll think of migrating back to doing lightmaps in Mental Ray / Vray and try to figure out how to render radiosity lightmaps outside of Unity.)
there is an asset to extend beast features so you can dig into granular settings.. works quite well. perhaps you can tweak some GI settings with it.
http://forum.unity3d.com/threads/lightmapping-extended-advanced-lightmapping-settings-editor.161319/
I would check some of these in GI settings:
Rays (1000 is default I believe) Bounces (3?, with 2 bounce boost?.. not sure what the defaults are there)
.. and I'm pretty certain many of the tweaks are "almost" exactly like Turtle...
$$anonymous$$y issue is not really related to that. There is a thorough lack of documentation about Beast and it's integration in Unity. When you bake you get the following steps displayed on that status bar GI Calculations Baking Texture Bilinear filter The GI calculations step finishes at various settings from extremely quick, to longer ( a few hours). This seems to be the step where all the rays are fired from the lights, the bounces and/or AO is calculated. Then it moves into baking texture, which I would think is the step where that data calculated is being written down to pixels in the lightmap 32 exr. This step for me completes probably 10 to 20 times slower than GI calculations. I figure that it is writing it in RA$$anonymous$$ and then transferring that data to my hard drive. That's why I'm asking if getting an SSD would speed up that writing to disk function, or if there's a way in windows to quicken that process.
what is your operating system? There are some advanced techniques you can use to monitor system performance to see if caching is an issue... I have done this with mental ray, but using windows 8.
With regards to putting Unity on the SSD, it honestly will only help Unity when the app starts up.
One technique that can be done is to put in a SSD for use as a cache drive (this is what I do), and assign all cache (%TE$$anonymous$$P%, et al.) via environment settings to be located on the SSD. You may also consider adjusting the paging file for your system and also look at what options are available for moving or setting any application specific caching to the SSD (I also do this for photoshop).
However, you really need to first find your bottle-neck. And this can really only be done through rigorous testing and utilizing the data collector features found in performance monitor.
