How to improve the MR FinalGather speed!

[psv]

search in max reference …lume shaders are shipped with max, maya have a converted and unoficialversion from maxs one..and xsi i dont know but being the first program with mr as a renderer they should have them too…LUME shaders are in max from max6 version…mr3.2 , and not all are hidden …in fact some are well exposed like metal , or glow, etc
mr beeing mr after all , disregarding the OEM those lume shaders should work on all platforms, with slight adjustements…

max help/ aditional help/lume manual

[steven_xu]

special thank you naik!! thanks a lot~

[steven_xu]

naik!!
you often Open FG+GI or only FG？ whic h can take the best effect?i often use FG only,it’s correct?some one said if U open FG only you will find the blacknoise on rendering pic

another Question:)
i set the max FG radius as the 1/10 of the sence and the min is the 1/10 of max，the max radius is the most reason which effect the speed of FG,and the min is effect the FG pic quality,so i often use 1 or 0.1 as min,how about U?

[cpan]

you often Open FG+GI or only FG？ whic h can take the best effect?i often use FG only,it’s correct?some one said if U open FG only you will find the blacknoise on rendering pic

when U use FG only, mentalray doesn’t calculate any secondary diffuse/specular bounces so you won’t get realistic lighting. This has been addresed in mentalray 3.4 (maya 6.5+, max 8) and now you can use FG with the Secondary Diffuse Bounces to ON in order to get extremly realistic results (much more realistic in many cases than the GI + FG method) and fewer RAM consumption

the GI + FG method works like this:

• the GI/Caustic pass calculates the Secondary Diffuse/Caustic Bounces
• then the FG pass use the GI/Caustic information to create the final illumination (the difference between FG only and FG+GI/Caustic is that with the second way, mentalray also computes secondary light bounces).0

another Question:)
i set the max FG radius as the 1/10 of the sence and the min is the 1/10 of max，the max radius is the most reason which effect the speed of FG,and the min is effect the FG pic quality,so i often use 1 or 0.1 as min,how about U?

well, i never use the Scene Size FG option but instead Pixel Size. This means FG points will be calculated uniformely “on the screen” instead of scene. So to be more specific, Pixel Size FG will greatly improove FG speed & quality especially on big scenes (in scenes where geometry is far from the camera). Another advantage of Pixel Size FG is that you can use the same settings for min/max in every scene, not depending on it’s size… i usually use 1.0 / 10 (aka 1 pixel min and 10 pixels max) for detailed illumination. On the other hand, Scene Size Dependand FG is highly recommended with animations ( and Pixel Size for stills) cuz you can speed up renderings by freezing FG points

take care
cpan

[naik]

special thank you naik!! thanks a lot~

Your absolutely welcome my friend !!!

I saw this thread very late… besides i dont understand why this thread is in this forum here, and not in the main…???
I would like to see a sticky thread about such MR issues…! :bounce:
Cause there are many many things we can discuss…hehe

naik!!
you often Open FG+GI or only FG？ whic h can take the best effect?i often use FG only,it’s correct?some one said if U open FG only you will find the blacknoise on rendering pic

Okay buddy here is my view (as i see there are many opinions about that…)

I try always to use Fg + Gi…or better say GI + Fg…
Fg was brought into that Global illumination process in order to
support / assist the GI, and in my view, cause of the common
speckles / artifacts. But at the end buddy our aim is to get a good result, so why not breaking the “rulez”.

Also im sure, that the combination of both will increase the quality of the renderings, absolutely
( GI Photon gives the specific clear hard look and FG will smooth it out and make it softer… - very generalized)

another Question:)
i set the max FG radius as the 1/10 of the sence and the min is the 1/10 of max，the max radius is the most reason which effect the speed of FG,and the min is effect the FG pic quality,so i often use 1 or 0.1 as min,how about U?

Dont do that… i always manually set the radius, and never ever use the default setting or this 1/10 rule, never.

As i mentioned before the radius function is sometimes the key for good and fast renderings in combination with GI. After 100 of test renderings i made this, as one of my MR rulez…

the max radius is the most reason which effect the speed of FG,and the min is effect the FG pic quality,so i often use 1 or 0.1 as min,how about U

Sorry but i disagree with that, you can not divide Max and Min in my view, they are dependant from each other. I always set both parameters which stands for a limit which Fg rays will be in the radius for the gathering process.
I can hardly explain that, sorry…
As these params are scene dependent its hard to give generalized values. But i often use 2m for max and 1m for min, then i start to tweak…

As yashu mentioned, with MR 3.4 the new diffuse Bounce parameter has been added which is very very cool. And i always make use of it.
But be warned, it will increeease the rendertime!

There is also some other stuff “THE FALLOFF” which will limit
the ray-length distance of the light rays for the regathering process.
In the past i always built up a invisible big box or a sphere which will enclose my scene cause of the effect, that Photons / rays were shoot into the open space and heaviliy increase the render time.
Make use of that new function its awesome and will boost with the correct setup your rendertime too…

Okay to the end one comment to the radius.
I always use the scene radius and not the pixel based radius.

This means FG points will be calculated uniformely “on the screen” instead of scene.

Im dont 100% agree with that yashu, it is correct that the pixel radius is uniformely but the scene radius is that too.
I dont know why it should not be uniformely…
Perhaps there is something that i dont know… but im quite sure about that, otherwise my animation would flicker like Hell…

On the other hand, Scene Size Dependand FG is highly recommended with animations ( and Pixel Size for stills)

How about that? i know that the scene radius is good for animation but why not for stills? I use scene radius 99,9% for my renderings.
Can you explain that more detailed?
In my view for MR users it is more obvious to work with the scene radius cause of the fact that the values are not abstract as pixels.
If i see a scene and have the size realtionships i can almost imagine which values (scence radius ) would fit very well without tryin around.
Also i dont think that the pixel scene radius would fit for all renderings… cause if the scene is totally different (objects) or the render size, that will definately change your set? What do you think?

@Steven xu --> i hope you can use it buddy.
Also nice would be having once scene and experimanting with that!
Perhaps you with your normal way and settings and i or other mates
with their procedures, that would be awesome!

Regards…

[cpan]

naik:

Scene Dependant FG allways creates more points on surfaces that are far away from the camera (they look smaller but their size is what matters for Scene Dependant FG) and Fewer on surfaces that are closer to the camera. This way FG will take more time to compute and the results are awkward (better details far away and poor ilumination details closer to the camera). So, Scene Dependant FG is uniformely distributed, but on the “scene objects” not on “screen space”… hope u get the idea :shrug:

On the other hand, Pixel Size FG will compute the same amount of points, not depending if the surface is closer or further than the camera so you get uniformly detailed ilumination and faster FG computing - this is only good for stills cuz on animations the Scene Dependant FG can be reused when objects/transforms are animated in the scene.

In my view for MR users it is more obvious to work with the scene radius cause of the fact that the values are not abstract as pixels.
If i see a scene and have the size realtionships i can almost imagine which values (scence radius ) would fit very well without tryin around.
Also i dont think that the pixel scene radius would fit for all renderings… cause if the scene is totally different (objects) or the render size, that will definately change your set? What do you think?

well, the Pixel Size FG is much “easier to use” than the Scene Dependent FG. It doesn’t mather what size the scene has at all, it’s just pixels wich mather (for example with Pixel Size, if you change the render resolution to a larger one, the FG computation time will increase.). So with pixel size, NOT depending on the scene size, you will allways get good details with 1/10 pixels for example

take care
cpan

[steven_xu]

i test the FG radius with i test,look it below

this scene is 50005000cm,and the car is 1500600,i set the FG max to 10 and min to 0.1,cal by pixel,add only SKY and effect with Xdof~ i think it looks not so bad:)

and i test it in another scene without SKYlight and backround,just only use spotlight
the parameter is the same,so i guess the parameter of FG’s radius is right

[steven_xu]

i render the pic above in 5 mins, before these,i test the FG’s Radius,look the sample below:

1:FG100 max60 min1

2:FG100 max600 min1

Now you can see,the max over 60 the result is wrong,even the speed is so fast with 1min

3:fg100,max600 min10,result is even looks like a shit

4:i turn off the FG manual and set it auto,
it looks the car want to fly away :)))

5:when last i set the FG’s radius to max10,min0.1 , it takes a good look with 3 mins

with this parameters,i begin to add light/material/backround… in scene

i want to know what is the real onnection or proportion between scene and FG size
i can’t believe there is no way to help the setting:( only testing and testing???

damn~

another Q:i often render a big scene with 200-400W faces with much raytrace
when i only add skylight and 1-2 spotlight to render it,MR’s speed look very good~
faster than VR and FR,LS…
when the light amount over 10,such as omni(even with falloff),the mr speed will turn very very slowlyslowlymore slowly than VR…

how about U?

[psv]

HEY STEVEN XU

if you wan`t a clear relation between fg radius size and scene size, you should add another parameter ,scene consistency in other words …how the geometry is spread across the scene…a sphere on a plane ,or a forest… for the same scene size you will have two setups…

as yashu mentioned pixel size fg will work better if you don`t want to do testing…again and again…

once you have tested 10scenes you will be able to spot the right settings i`m sure about it…

[steven_xu]

PSV,i even think only use pixel to calculate is the simplex way to resolve the FG’s problem,but it depend on the much time,i only want to find a balance or rule to setting FG,not only click the pixel button…

[naik]

Okay guys my view about that…

Yashu i certainly understand your explanations ,but there
is one point which is inconistent.

Scene Dependant FG allways creates more points on surfaces that are far away from the camera (they look smaller but their size is what matters for Scene Dependant FG) and Fewer on surfaces that are closer to the camera. This way FG will take more time to compute and the results are awkward (better details far away and poor ilumination details closer to the camera).

In my view the pixel radius will not work properly for objects which are far away because the radius is based on pixels and not on the scene, so im sure that the scene one is more accurate for that reason.

If not than this would be inconsistent too…

Here are two pics where i tried to compare it.
The first one is scene based with Max 1m and min 0,1m

This one is pixel based ( i tried almost hit the same values in Pixels and Meter scenes based on the processing points)

In my view they are almost the same, but…

First i dont think its true that with the scene radius the objects which are near to the camera are less accurate.
Yashu compare both pics. The Objects in the front are almost 100% the same ( regarding to the P points)

But in the rear of the pic there are slightly differences, one is that in my view the scene one is more accurate (more points)
See the left wall when is goes black as an example…

Perhaps its not a good example but for a rough comparison.

@Steven xu

Nice Pics buddy, but im sure if i had the scene on my pc
i would need perhaps 2 maximun to get a good result and one more to do it “perfect” for my needs.
So its hard to say, in my view its an experience thing.

“once you have tested 10scenes you will be able to spot the right settings i`m sure about it…”

What psv said is true, but definately for the scene readius too.

STEVEN XU

The new pics with the red Golf.

Why do you use such exaggerated values for the maxium??
60 ?? Which units do u use? Meters?
as i saw 600 my heart was bumping like hell mate :eek:
hehehe…

All things considered, im sure if there is one experienced MR user who only worked with the scene R and another who only worked with the Pixel R, that at the end the result is absolutely the same…
Just knowing your tools and how to tweak…

regards

[steven_xu]

lol~~special thanks three friends(naik,PSV,YASHU) to discuss this,it’s a so nice time,and i learn much from you:)

BTW: YASHU & PSV: said use pixel to calucate the FG,click the pixel as radius,not setting the radius blow?
not setting max & min? by default? 1,0.1?i think it will works too slow by default…even the result is the same.

naik:
i can’t see your pic,maybe the pic address is wrong…can you check it?
U said why i set the max radius to 60…hehe,i set 1/10 as scene,bigger can improve the speed of rendering time,but bad effect,you know,always client not given U so much time to render…why i question this,just want to find the way to speed up FG,it’s too slow even with perfect result…

[naik]

naik:
i can’t see your pic,maybe the pic address is wrong…can you check it?

REally???

Impossible buddy…

YOu still got the problem?

ohohhh…

[steven_xu]

look at this…

[naik]

okay checked the address and it is okay (imageshack)

When i see the pics, then you MUST see them too… very strange…
Hope its visible now…

60…hehe,i set 1/10 as scene,bigger can improve the speed of rendering time,but bad effect,you know,always client not given U so much time to render…why i question this,just want to find the way to speed up FG,it’s too slow even with perfect result…

Hmm, but what is your working unit in max?
Do you use Meters , centim… Inches ?

It seems to me that your scene is like i call “GARGANTUA”…hehe

lol~~special thanks three friends(naik,PSV,YASHU) to discuss this,it’s a so nice time,and i learn much from you:)

Cheers buddy

[naik]

look at this…

heee??!
How?
Its not on my server either in my pc…
Its on the imageshack server…

So when i update the cgtalk site it must be invisible for me too…
very strange man…

Hmmmm

[steven_xu]

damn,i even can’t see your pic…
what’s meaning of “GARGANTUA”…???..
BTW:i used CM as default

[naik]

damn,i even can’t see your pic…

Grrrrr!!!

Okay i will post it againg - same order - first one scene R second one pixel R

Hope it functions now…

what’s meaning of “GARGANTUA”…???..

HAHA! your so cool buddy - GARGANTUA means a mega mega huge scene (regarding to the scene unit)
somethin like 600Meters * 600 Metes - for a small room…

[steven_xu]

naik~~if i can’t see the pic,can U mail me the pic…thanks,my mail is steven_bchs@yahoo.com.cn

because i work for citroen…so,U know,a scene with 8-20 cars will be a horrific thing especially in one scene with other things…so speed is important…

[naik]

Still the problem??

MAn that starts nerving me!!!

If it is still not there i will send it to your email…

regards…