So I am rendering a exercise room for a client. I am using the new path tracer featured in the new IRender nXt. I am getting a lot of noise in the shadows. How do I get rid of that? ie: Setting adjustments, more passes, etc....
How many passes are you using ? - the path tracer option is relatively new and fairly experimental - your render will require 100s of passes to complete - see this post : http://irendernxt.com/forum/topics/path-tracer
I had it set to unlimited passes. The image above was at 300 passes @ 15hours when I stopped it. I will check out the link you provided and see what it has to say. Thanks for the quick response. If you come across any more info regarding path tracer issues let me know.
I have been experimenting lately with path tracer on similarly complex interiors and 300 passes will not even touch it! - if you look at Al`s post he is looking at 1500 passes before he gets near parity - again this is is early days and the developers probably need more time to refine - In commercial terms I for one am sticking to the standard packet method (which ain`t bad anyway) until things develop a little further. On some of the other sites\posts there is talk of implementing nvidia cuda technology to improve speed - we probably need to see if Roy ( one of the original authors can do anything with this).I think there is a lot of cover on the autocad accurender forums but our Richcat monitors them more than I - he may have more reference
Thanks! This is all good info.
I will be sure to do more rendering with path tracer and share my results.
I will also keep my eye out on the forum to be of assistance.
I appreciate the support.
Here is a rendering of the same room I did prior to the one I originally posted with 800 passes. I guess you can start to see the different with the noise.
Nope, there are no noise reduction settings when using this engine-- noise is a fact of life and will gradually work its way out of the images-- and yes-- it's more prominent in the shadows.
Some post-process filters can do a nice job of removing noise-- Selective Gaussian Blurs with low threshholds work well for this. Beyond that, throwing as much compute power at the image as possible will help.
Do keep in mind that the standard engine can get pretty close to this result-- it's usually much faster to converge.