improvements possible?

[ThePriest]

File doesn't open in r9.6/10

[mm149047]

File doesn't open in r9.6/10

hmm ... it'S R10
is there a chance i can make it readable to you?

[kerbouci]

just explain the the scene components and the render settings. as well as your machine characterestics

[vid2k2]

I could never open any file you posted via your site.
It comes in as html code.

Zip the file :)

[mm149047]

I could never open any file you posted via your site.
It comes in as html code.

Zip the file :)

= zipped :)
www.markusmooslechner.com/test.rar
very strange though

[vid2k2]

Much better.
I'll have a look at your v10 file

[mm149047]

just explain the the scene components and the render settings. as well as your machine characterestics

well, i have a spline set up of ascii data points. some 800 or so. every point has its own radius, so the nurb the spline sets up varies its radii accordingly. the scene is rendered AO. no additional light. i set the spline to adaptive and 5 degrees. AO to 10% accuracy, minmax samples (10-128) contrast 1, the embedded circle to uniform and 7 units. this setup is for a 640x480 preview render.
my machine: pc, amd2500xp, 2gigs ram (the rest of the renders usually work fine)
markus

[vid2k2]

[QUOTE=mm149047]hi everybody!

i would be very happy if someone could spare some time and look over file below and check the setup for possible improvements as to render speed. the file as it is now renders very slowly !

thanx for your time :)

markus

QUOTE]

A few things:
- no need for render option settings of 15-5-15
-those can be 5-1-5 ( actually, no need for reflection)
-AO schnell/fast can be 16 rather than 128 (even lower)
-I'm not sure if sweep nurb splines need their own AO.*
-Perhaps general AO (Acc70%-Max Samples 16)may be faster over all*
(however, I've not tested this*)

I tend to be a 3D purist and avoid post work, but much of the spartan
look may be better adjusted in PS? I've not worked with this type of scene
before, so, I'm not sure if this would improve your time.

The attached file ran quickly with minimal settings

[mm149047]

[QUOTE=mm149047]hi everybody!

i would be very happy if someone could spare some time and look over file below and check the setup for possible improvements as to render speed. the file as it is now renders very slowly !

thanx for your time :)

markus

QUOTE]

A few things:
- no need for render option settings of 15-5-15
-those can be 5-1-5 ( actually, no need for reflection)
-AO schnell/fast can be 16 rather than 128 (even lower)
-I'm not sure if sweep nurb splines need their own AO.*
-Perhaps general AO (Acc70%-Max Samples 16)may be faster over all*
(however, I've not tested this*)

I tend to be a 3D purist and avoid post work, but much of the spartan
look may be better adjusted in PS? I've not worked with this type of scene
before, so, I'm not sure if this would improve your time.

The attached file ran quickly with minimal settings


thanx for going over my file
doing a quick render with your settings - can the splines be improved?
unfortunately global AO does not produce the relevant results :(

[georgedrakakis]

www.markusmooslechner.com/test.rar (http://www.markusmooslechner.com/test.rar)
hi marcus,
is the link dead?
i couldn't download anything, just a blank screen.

ps: your gallery is very-very-very interesting!!!
are all these set ups of ascii data points?

[mm149047]

hi marcus,
is the link dead?
i couldn't download anything, just a blank screen.

ps: your gallery is very-very-very interesting!!!
are all these set ups of ascii data points?


the link should in fact work - plz try again
thanx for your compliments ... the "heartbits" area (the AO images) are asciis - the rest ist nonlinear math art
:)

[vid2k2]

I rendered your file as was and then as I described with
global AO. That cut about 30% off the render time.

screen capture v10 demo
rendered on a slow 1.8 imac / 2 GB

[mm149047]

I rendered your file as was and then as I described with
global AO. That cut about 30% off the render time.

screen capture v10 demo
rendered on a slow 1.8 imac / 2 GB

that WAS useful - had to up the max samples to 128 though - 10 is too grainy for me :)
thanx for tweaking
markus

[georgedrakakis]

hi mm149047,
the link is working fine (4 clicks= 4blank screens -> 4 downloads!).
this is a more ''clean'' render (but more flat & boring ) using the following setting, as showed in the attached file (min/max -> 10/15 in global AO settings reduces render-time a lot, 33mins on a 1.83 Mhz dual-core, 1Gb ram & almost full hd laptop).

http://i122.photobucket.com/albums/o271/studiomoccoro/ascii.jpg

@ vid2k2:
i like your render, grain is definately my style ;)

[mm149047]

[QUOTE=georgedrakakis]hi mm149047,
the link is working fine (4 clicks= 4blank screens -> 4 downloads!).
this is a more ''clean'' render (but more flat & boring ) using the following setting, as showed in the attached file (min/max -> 10/15 in global AO settings reduces render-time a lot, 33mins on a 1.83 Mhz dual-core, 1Gb ram & almost full hd laptop).

hi george!
thanx for your experiment - i think i will have to play with the settings a little more - lowering the settings too much reduces the desired AO effect ....

[mm149047]

hi again!
i am still working on this project - got to do a lot of tweaking so to make it renderable bigscale. it is going to be a still - so now i am trying to tweak the splines, for here i think lies most of the possible render time reduction. still all those fidderent intermediate point settings give me a hard time. as far as i have gotten so far i think that subdivided is a good choice. i use 20 degrees and 0.1 m / unfortunately there are similar effects upping the max length on smaller degrees. :( can someone help me out here on last time? the problem is that i will render the object from different perspectives - so the settings should work on all of them. is there a chance?
i added a neutral material so to make previewing easier. it will be rendered in AO though.
www.markusmooslechner.com/test.rar

thanx for your time
markus

[LucentDreams]

right now your using geometry AA on a scene with a lot of fine geometry. Geometry AA works only on the Edge of the geometry but at 16 samples which make sfor a realy sharp edge, but often slower than necessary, and doesn't help the grain you may get when using AO. Try placing a big flat plane right in front of your camera with a material that has everything off except transparency. This will eliminate any geometry AA need as all geometry is rendered through the transparent shader which is handled by the best AA. using best at 1x1, 2x2 or even 1x1 4x4 should give comparable speed, a satisfactory edge, and also smooth out the grain of your AO. So while it isn't really going to be much faster on its own you should be able to get away with a grainier AO setting.

[mm149047]

right now your using geometry AA on a scene with a lot of fine geometry. Geometry AA works only on the Edge of the geometry but at 16 samples which make sfor a realy sharp edge, but often slower than necessary, and doesn't help the grain you may get when using AO. Try placing a big flat plane right in front of your camera with a material that has everything off except transparency. This will eliminate any geometry AA need as all geometry is rendered through the transparent shader which is handled by the best AA. using best at 1x1, 2x2 or even 1x1 4x4 should give comparable speed, a satisfactory edge, and also smooth out the grain of your AO. So while it isn't really going to be much faster on its own you should be able to get away with a grainier AO setting.

OMG - how can someone possibly know this ! you helped a lot - i bow in front of you :)

[mm149047]

right now your using geometry AA on a scene with a lot of fine geometry. Geometry AA works only on the Edge of the geometry but at 16 samples which make sfor a realy sharp edge, but often slower than necessary, and doesn't help the grain you may get when using AO. Try placing a big flat plane right in front of your camera with a material that has everything off except transparency. This will eliminate any geometry AA need as all geometry is rendered through the transparent shader which is handled by the best AA. using best at 1x1, 2x2 or even 1x1 4x4 should give comparable speed, a satisfactory edge, and also smooth out the grain of your AO. So while it isn't really going to be much faster on its own you should be able to get away with a grainier AO setting.

kai - can i bother you again ??
i am sure i am doing somehting wrong, but with your settings everything seems to be a lot slower than with geometry AA. i placed a transparent plain plane in front of the camera - material transparent, AA->best 1x1-4x4 and variations. the output IS smoother but a lot slower - even when i tried downing the AO shader settings. --- any idea??

[LucentDreams]

well AO will always be slower with AA best than without AA best, even if its behind glass, but with AA you should be able to use lower settings for AO as it should appear somewhat smoother and less grainy.

like many optimization techniques it depends on the content in your scene. Like the Acene motionblur trick its only advantageous under the right conditions. Scene motionblur it typically slower to get comparable (often better results) but if there is any movement in the scene scene motinblur will reduce flicker and give motionblur to the animation, so an extra minute per frame may be worth the much better final product, but in some scenes it could actually be faster than rendering individual stills to.

The transparent best AA trick works great with geometry intense detail as the edges all render faster, but since AO is slower with best results will vary depending on the ratio of geometry detail to AO detail.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.