so i've read mental ray in maya2008 is faster with displacement and ambient occlusion rendering. can anyone confirm this and possibly give me an idea of how much faster it is?
surely mental ray's displacement rendering can't compete w/ renderman, yet....can it?
thanks.
QUOTE(casey33 @ 10/01/07, 09:48 PM) [snapback]274168[/snapback]
surely mental ray's displacement rendering can't compete w/ renderman, yet....can it?
Well it depends on the situation, they work differently the fine displacement that facsimiles the way reyes does this is pretty efficient already in 3.5 and is actually at same speed marks as prman in many instances. But only works on nurbs properly. This said prman displacement elents have a certain inbuilt flaw in them that mr does not have (on the other hand mentalray is a artifact magnet in on certain edge situations). However the order of shading execution differers from prman so making a qualitative comparasion is hard. As prman encapsulates displacement also with final sampling system weirdly.n But no prman 13.5 is amazingly fast but beyond that prman can do blurred reflections that are raytrace local without raytravcing as well as fg so you could fg fur without super problems and times.
But for some effects they clock same speeds but prman just uses lot less memory (thus you can put lot more detail in)! Its like comparing a dog with a cat.
Still havent got enough qualitative data to make a educated guess, but they are faster in certain conditions to 3.5 yes. Mainly see the difference in baking speed. (also something prman is better imho)
PS: if i make sphere with lambert and default light put in a dsplacement of Bulge and do a comparable quality mr does 16 secs while prman 2 (which is mostly export,and with quite much over quality! i cam up the quality to beyond mentalray and still be at 2 but when i turn raytracing on the situation changes a bit to mrs favor, tough i dont have 13.5 to test on. even so doing 2 2 secon renders and a 6 second preprocessing yields decent and stable occlusion without ray tracing still faster than what mr does. I wish maya would have something to present sets in for mr better so tuff like raytrace sets would be easier to do.)
I haven't done enough test to confirm this as well.
And for the comment on prman... you'll never get prman speed with displacement in mental ray if you use raytracing ... nerver.
Displacement in prman with raytracing will be as slow and memory hungry as Mental ray. It's a simple principle really... prman can create the displacement geometry, render it then throw the geometry away... with raytracing (mental ray, prman with raytracing, and any other raytracer) the geometry is create then kept in memory... this will take more memory and everything will be slower because every single raytrace ray will need to check for intersection with this displaced geometry.
QUOTE(Kobayashy @ 10/02/07, 12:34 AM) [snapback]274174[/snapback]
Displacement in prman with raytracing will be as slow and memory hungry as Mental ray.
Well almost, prman uses a potenttially different tessellation for raytace lookups than it does for the visible mesh. So theres potenttial savings there (and it ses quads as bases). Off course you can do this with mental ray too. But setting it up in mayas pain.
In fact mr has a slightly faster intersection lookup than prman. and yes mr is afaster raytracer. But prman tessellates faster.
But yes, its a rule, more memory you use the faster the intersection tree, the coolest trick in prman is ray traced effects without ray traced data. By pre solving the octree, trick is not so cool in that it is faster than elementary raytracing but that it can be resued partially or fully form frame to frame wich makes it even faster. Off course its NOT ray tracing so its also slightly less accurate.
this reflection stuff looks like the interpolate reflection of the mia_material in mental ray.
Prman use occtree, MR use FG map... same difference
just curious....what would you estimate the difference in speed to be between maya7's mental ray and maya 2008's mental ray?
QUOTE(Kobayashy @ 10/02/07, 05:43 PM) [snapback]274198[/snapback]
this reflection stuff looks like the interpolate reflection of the mia_material in mental ray.
It isnt its not doing any ray trace at all each point samples , its like a mipmap lookup with directional samples (its a 3 dimensional image map but unlike a intersection tree it contains averaged copies of results on lower level so you can sample many rays worth of hits with one equalient to raycast). But yes you can do interpolate samples in prman too but that would use raytracing in fact interpolating many pixels is a built in feature of the renderer so all shaders can do this, which is what they want to avoid altogether but it would end up being way slower. (the interpolation works by sending rays every so and so pixels and then interpolate result this just pre solves the world).
Now since prman has very fast at scan lines you can actually afford to render the scene many times for the cost of shooting some rays.
Idea is that you render without raytracng 2 times. By the way pixar had also a nice discussion on how to optimize the photon emission coherence with 2 renders.
Just on example of baking speed is that when i needed to bake a shader to disk it took me 1 minute to bake it in prman and 1 minute in gelato and it took me 12 minutes to bake in maya+mr.
But its not so much mr's slowness but the rather idiotic maya implementation. because if i make it myself by writing the mr stream i can do it again in about 1 minute.
IHMO render man is the better tool, but mentalray is simpler to use and is more straightforwardly effective. But engineering a rednder in renderman is actually much easier.
QUOTE
just curious....what would you estimate the difference in speed to be between maya7's mental ray and maya 2008's mental ray?
maya 7? i was comparing it to maya 8.5 but basically fg is way more optimal after 3.5 so a 300 sample rays for fg is about the same quality as 1000 rays in maya 7. for 1/3 of the cost.