How to improve the MR FinalGather speed!

#1

i often use “save”,when rendering 320*240 pixel pic,i saved the FG map,rand load it in rendering 1600,but i found it’s always helpless to improve the speed of rendering FG,why~i turnoff the rebuild yet:)

#2

Hi,
does your Max calculate FG again when you change to 1600pxls or it uses stored FG map? Max should use it. It is very big diference in render time between 320p and 1600p, could be more then 20 times, even without FG.
If you render in smaller resolution doesn’t mean that FG is faster, it is the same, depending of sample value. 320pxls U can use for prewiew shots if you doing animation, but for final, there is no reason to render all first in 320 just to build FG map,you’ve got the same number of samples and calculation time is the same. That is my oppinion, and seems quiet logical for me. If someone think different, he/she can post.

#3

just as U said,i can’t find it will be faster when loading small FGmap,so what’s the use of FGmap?goodlooking?or be used in same resoultion~?

#4

generally people save FG maps to speed up animation renderings… so instead of recalculating the full FG map every frame, we set FG Freeze or Rebuild to ON so mray will calculate fg rays for the first frame and then mray will calculate FG only for the places where modifications occured

hope u understood my bad english

-cpan

#5

U mean the FGmap only used in animation?
U mean if saved the FGmap in first frame,FG will add calculate to the different between 1-2frame?

FGmap is really helpless in still frame??damn!

#6

U mean the FGmap only used in animation?

usually yes, but you can use it also for stills with Rebuild OFF so it only calculates the places where changes occured (eg. changed the color of a material)

U mean if saved the FGmap in first frame,FG will add calculate to the different between 1-2frame?

yeap, kinda

FGmap is really helpless in still frame??damn!

nope, as allready said u can speed up renderings with the Rebuild set to OFF… and u can also visualize the map in your viewport (don’ know if in max u can this) so u can see how the rays are distributed (helps in setting the min/max FG Radius)

#7

I already turn off the rebuild button~! helpless to the speed! if i change the resoultion!
setting the radius i often use the 1/100 of scene!

how the FG speed slowly is! shit!
why not learn it from VR? VR is more faster than MR at this!

#8

yashu
generally people save FG maps to speed up animation renderings… so instead of recalculating the full FG map every frame, we set FG Freeze or Rebuild to ON so mray will calculate fg rays for the first frame and then mray will calculate FG only for the places where modifications occured

this is not true. MR will use one map and will NOT recalculate modified parts. If ya dont believe create floor and wall (in gray) put the sphere (in red) near wall and create Fg map. Check use map and then remove sphere and hit render. Red reflecton is still on the wall althougth there is no sphere- MR used map!!!

COCLUSION: MR clculate whole image (if rebuild is turned on) or use one FG map for whole animation (if use map is checked) and dont recalculate only changed parts.

#9

you have just proved that mr can save a fg map, and re-use it…(probably rebuild was off…)
fg its wiew dependent , and i bet you didnt move the camera..... rebuild must be turned off in order to let mr use a saved fg map , and since there are changes in camera move , like the manual says , it will only compute those changes ....adding those changes to the initial fg map. thats easy to see if you check the size of the map, at first frame , and at subsecvent frames…it wil grow…

#10

not only calcalute the changed parts,but almost all~when 320pixel to 1600
but i find set the correct FG Max/min can improve the FG speed faster
i lost the hope in load~ing,helpless~

#11

Okay maybe im a bit late for it but here are my views about that and other Mr points.

1.) Fg map

In my view its not true that MR will change the FG Map
if something changes in the scene
(he changes the map but its totally wrong, and not useless.)
I tested it with a normal room (white) and a box red.
I save the FG Map then i change the box color to yellow.
Render againg nothing changed, move the camera 90 degree
then suddenly half of the room is red and the other half yellow.
So i dont know how you doin this but for me its definately not working.

As i work now for a firm and doing some interior scenes
my chief has warned me of some MR issues.
But i will mention that later on.

2.)
I tried a lot of things in order to speed up the render process
in Mr.
One of the best methods is to play a lot with the different param in
Mr (especially the RADIUS !!!)

Its not true or better its a bad way in order to get good results
only to put the Fg samples high - bad decision.

I had a scene with a sample of 300 and a render time beyond goodness. But playing with the param i could get a much!! better result with an Sample set of 35!! but changed the radius.
(Almost the same with the photon setting for GI)

Also if you are planning in using FG , you can set the GI sample
way down (Photons per sample)

I almost use something like 500 - 1000 when using FG.
But if you have much higher Gi Photon set you must increase the Gi sample too ( perhaps 1000 -2000)

But nomally i get very good result with 500 (perhaps less!)

Another important parameter is the FASTLOOK UP option.
I dont know why “nobody” uses this but u can belive me in some
situations i reaaally speeds up the gathering process.

Its very hard to come in the MR optimization process but in my view
its only a question of a proper setup.

After studying the parameters i could speed up my renderings
generally almost to 600%…

So whenever i can use the FG / GI Map i use it.
For me its like “Knight Riders Super P. Mode” in
combination with a proper GI sample and Fg Radius setup.

My two cents guys…

P.S Perhaps another interesting point would be to dicuss about
the very common flickering affect in MR.
I got some solutions for that, would be interesting what you guys think ?

#12

FG is view dependant, so if used in animation, it should be recalculated each frame.
You can turn FG not to rebuild in each frame (thick off rebuild) but youre getting odd results.
I am not sure if FG is a good idea for animation, is very very time comsumming to avoid flickerind due to fg recalculation. For stills, although you can optimise it as naik detailed.
You have an extense discussion here
cheers

#13

a verry simple way to lock the fg map for an animation is to render the mostcomplete fg map you could get , letting the camera to see every visible faces of the geometry in the scene,as the fg map grows,and save it to a file, then …simply go to the location where your fg map is saved and set the file to read-only(right click/properties…) and set mental ray to use that file, rebuild off.
in this way your animation will be flickering free and a lot , a lot faster…bien-sur without moving objects in the scene
it is true that mr will try to add information to your fg map, but that one beeing read only…the trick will work , and you will force mr to add only a small amout of time to fg calculation…

for stills the fg time is not that critical…
hope that helps:)

#14

So i almost use FG for my interior animations.
Its true that MR will flicker very often.

but there are ways to handle it.
One is the mentioned FG MAP to use, but in my view there will be a bit of flickering still visible.
(Especially at the edges of objects when the cam is rotating around it)

I can tell you guys, the best “trick” is not to save the
animation as an AVI or whatever file.
You have to save it as single / Seq Image files, with no compression.
After that combining all pics into an animation file.

The embedded MAX AVI Algorhithm is nasty, i can tell ya.

One thing to too, if your are using sophisticated
Phenom Materials from the MR Lib. then its absolutely normal that the render time is very comsuming.
There is no way around it, or maybe i dont even know…

#15

hello naik

100% right about not rendering avi-s straight out of max,

thats common sense almoust. if you have blurred reflections, lots of shaders , maybe it coud be useful to use aray type ` shader , to switch fg calculation(and not only) to a simpler version of your shaders…
being careful with ray offset could help too, (simplerversions of scene objects visible only for secoundary rays, fg ray for example, and real objects visible to camera but invisible for secondary rays… etc

those materials from mi lib are verry fast versus a standard material they are calling way less shaders during renderings+ they respond far better to light, but must be used only for specific purposes,you are doing interiors so a good way to test that is to replace the material for …those wall i supose you have in… with a mental ray lambert, or a Ward one …you will see the difference.

one neat trick is to use an ambient occusion shader anr to compose it later in post, this way your fg setting can go a lot lower…or use it as a mask between two versions of the same shader…

cheers sorin

#16

Hi PSV,

There are some interesting points you have mentioned…
But one thing is unclear.
What do you mean with Mi lib ?
Do you mean the lume shaders? (Metal,glare,glass)
or something else…

one neat trick is to use an ambient occusion shader anr to compose it later in post, this way your fg setting can go a lot lower…or use it as a mask between two versions of the same shader…

I must admit, i didnt tried this out yet, but i will give it a go.
Also i dont really got problems with Fg.
Im almost satisfied with the results.
Also i can not understand guys using Fg sampled about 1000-2000 or even higher ??!! and after that beeing disappointed about the
Killer Render Time… Or beeing annoyed about it.

Hey whats up?? If you rendering a multi Mill. Poly scene thats the same!

I have never ever rendered a scene ( it was never necessary)
to go higher then 500 - 600 FG samples…
As i mentioned before, in MR its a question of the right settings for the proper scene.

PSV did you work with the Ambient Occlusion Shader?
How about the difference in quality. That would be interesting…

Regards…

#17

ok naik ,

those wrong spelled( mi lib…)where in fact “mib_illum_lambert” , “mib_illum_phong”, etc wich if you are using 3dsmax can be found in C:\3dsmax7\mentalray\shaders_standard\include
for max 7, and in a similar dir for max6 *ive forgotten the exact name

those shaders must be un-hidden , they came with max by default but are hidden …max help claims thats for a bunch of reasons from duplicating existing functions in max shaders , wich is true, but limitating from certain points of view..., till undocumented , or untested functions,,,,like glare from lume colections.... go to max help/ aditional help/mental ray shaders and youll get a full list of shaders existing in that mi file…
be AWARE that if youre doing that on a work computer , withaut backing-up the files , things can go wrong…thats a common question on mental ray shaders thread on max resouces here on cg talk.....search hidden and youll find the exact procedure…
i use them all the time …

ambient occlusion shader packed in max , maya etc is good but dirtmap is faster for the same quality…in my experience

fg computing time is nothing compared with pure ray tracing time …for large:) scenes and larger resolutions…:scream:

and an humble advice…bear in mind that max it isnt the only program out there with mr as a renderer......go on maya rendering threads here on cg talk and youll be amazed…
how things can be the same…and there are some guys wich are real pro…wathever…:slight_smile:
most of those mr shaders designed for maya work in max:thumbsup:
hope it helps
Sorin

#18

Hey PSV, now i got it, was a little bit confused first.

Yeah actually i though i had unhidden all “shaders” but there are some
which are not hehehe…

“mib_illum_lambert” , “mib_illum_phong”,

These two babys here, i never worked with but after your tip i will give it a go. THX.

Yeah the last days i have read a lot about the AO and the dirtmap.
In my view probably very handy or time saving.

bear in mind that max it isnt the only program out there with mr as a renderer......go on maya rendering threads here on cg talk and youll be amazed…

hehe i know buddy, (I think Jeff had converted some shaders from Maya to max, and other guys told me too that many things in MR are the same on other apps.)

Thats true… Therefore we can all take advantage of that instead of
inventing the “wheel” …

Thx again for the tips.

BTW why is this thread in this forum and not in the main one?
I dont understand the logic here sometimes in CGtalk…
(Headline for this forum :“Post your links to shaders/Materials and plugins here”)

I would rather prefer dicussing that in the main forum like other did about MR issues…

#19

yes AO can save loads of time
the neat thing about those maya shaders is that you dont have to convert anything,ok just adding a litle apply type in mi file…not always
those two babes …are more , lambert , ward included.:slight_smile: …lambert is fastest for simple surfaces…like a white wall…

ok ,ciao

#20

I don’t undertand, Max 7 has MR 3.3 and if you have the subscription MR 3.4, why are these lumes hidden, and are they hidden in Maya and XSI. Someone please clarify this for me.