a question ive always had in mind is what makes AO bakes/renders take so long? and what could i do to improve my PCs performance to increase the speed of my bakes?
it concerns me when my pc starts venting like a jet when baking AOs xP
Using GPU baking with xnormal will decrease the render times *dramatically*. Bakes that used to take upwards of an hour happen in under a minute. Just change the renderer dropdown from "Default Bucket renderer" to "Optix/CUDA Renderer (alpha/experimental". I can't remember if I had to change any options in the plugin manager.
AO takes so much time because lots of rays ( ==billions ) need to be traced.
For instance, a 2k x 2k map with 4x AA and 256 ao rays(medium quality) takes exactly
2048 x 2048 x 4 x 256 = 4,29 billion rays.
The best way to accelerate that is:
1. Get a better CPU or overclock your existing one.
or
2. Use the Optix/CUDA renderer with your GeForce card.
or
3. Use the OpenRL renderer with a Caustic Graphics's RT2 card ( or use the new improved OpenRL's CPU mode that xn3.18.2 optimized ).
or
4. Render per-vertex AO using the xn's tools and then render an AO map with interpolation enabled.
As a little tangent, when you can't realistically increase the speed of AO bakes (can't buy new hardware, etc), it's a good idea to try to remain productive while they're happening. Bake your normal and material masks first, for example, and set up your base materials in PS while the AO is still chugging. Minimize time spent doing nothing waiting for a bake
As a little tangent, when you can't realistically increase the speed of AO bakes (can't buy new hardware, etc), it's a good idea to try to remain productive while they're happening. Bake your normal and material masks first, for example, and set up your base materials in PS while the AO is still chugging. Minimize time spent doing nothing waiting for a bake
Replies
Alternatively you could try Knald to generate AO from your nm, which will only take a few seconds.
For instance, a 2k x 2k map with 4x AA and 256 ao rays(medium quality) takes exactly
2048 x 2048 x 4 x 256 = 4,29 billion rays.
The best way to accelerate that is:
1. Get a better CPU or overclock your existing one.
or
2. Use the Optix/CUDA renderer with your GeForce card.
or
3. Use the OpenRL renderer with a Caustic Graphics's RT2 card ( or use the new improved OpenRL's CPU mode that xn3.18.2 optimized ).
or
4. Render per-vertex AO using the xn's tools and then render an AO map with interpolation enabled.
Also always do lower rez low quality ones to make sure the settings are okay. a 256x256 will show you if you have any major errors.
Yes, to all of this.