Yesterday, Ive found something mysterious (for me): My PC was running at 17fps and its was using 60% of CPU power only (by Task Manager). Timing node has shown me that the biggest part is “Prepare Graph Time” with around 0.0470 s.
In help file its written that:
prepareGraph (encapsulating all cpu work for preparing the next frame)
I’d like to get my PC up to 25fps, which seems possible. Hopefully, by reducing prepare graph time.
Some question to the devs:
-So is this “the bottleneck” CPU to GPU?
-And if so, I use Pipet once, DynamicTexture twice, AsVideo twice. Are this the reasons for this big values?
-What about using two graphic cards? One only for my crazy pixel shaking (like DynamicTexture, which is coming back again with AsVideo - actually former thread). Is it possible to set a destination graphic card for this somehow?
-I use some EX9.renderers for debugging purposes (about 20), switching them and quads and groups off usually. Do they still grab render time? Or is there a performance gain by deleting this nodes?
-I use some shaders in a row (as there is just one pass possible), i.e. for gaussian and erode/dilate. So, some go in DX9.Texture out and into a shader again. Is this done by graphic card internally (what I reckon) or is it somehow influencing Prepare Graph?
Thanks in advance