View Full Version : Native Renderer versus rest of the world

10-23-2012, 03:30 PM
I just started experimenting with KRAY and right away I noticed different take to GI in KRAY. Calculations of GI seams to be much faster than in LW3d but rest it is very similar.
In KRAY first pass of light coverage of the scene is much smoother. My question is that different type of GI method or just different realization of the same method ?
I have only demo of KRAY because right now I do not have money or reason to buy it but it looks promising. Integration with LW3d is seamless and speed is awesome.
LW3d is still better in open air scenes in my opinion and is for sure more flexible but it make me wonder what we can expect in rendering department in new Lightwave ?
Can we expect much from LW12 in rendering engine ? Just smoother GI will be enough. I have no clue about technical/mathematical background of this area but I
noticed few additional methods like photon mapping and others that is not available in LW3d ... maybe this is the secret of better GI.
I`m sure that it is all secret but few small details will be big help, for example in planing what to buy in the future.
As for the other render engines I`m interested only in those that can be integrated with LW3d so all those strange exporters, material editors and unicorns will not interest me.
Good thing that for this moment I do not have any projects that involve interiors because LW is so slow in that area !
On my blog You can watch few tests and examples I made while working on new project.

BTW: FPrime .... is it dead ? Can we expect anything soon from Worley ?

10-23-2012, 09:57 PM
I'd love to see GI improvements to LW's renderer too. I like it's simplicity and output, but there is a lot of room for improvement. FPrime is dead as far as I'm concerned. It really offers nothing more than what LW already has.

10-24-2012, 02:50 AM
Photon is like little direct light source giving per cent of energy that traditional direct light source has. It's forward ray-tracing - calculations are done from light source to scene, then bounces couple times that you want, and are reaching target. Photon supporting renderer can even convert direct lights and luminous surfaces to photons at the beggining and then no distinguish between them in the main shading routine.

LW GI is working completely reverse doing backward ray-tracing- ray from camera hits geometry, then in the all directions are send rays (primary GI rays), checking what color is there and averaging result (so GI has no normal vector), then again the same from these evaluated spots until reaching GI bounces limit (it looks like tree where camera location is root, then what camera see are first branches and so on). Without adaptive AA and without GI interpolation you can even calculate how many rays and surface evaluations you will have in the scene (without reflections and refractions):
width * height * aa passes * ( primary GI rays ^ GI bounces ) =
1920 * 1080 * 9 * ( 128 ^ 3 ) = 39,137.0 billions rays
(that's why MC GI no Interpolation is so slow ;) It's no cheating)