Logo by slon_ru - Contribute your own Logo!

END OF AN ERA, FRACTALFORUMS.COM IS CONTINUED ON FRACTALFORUMS.ORG

it was a great time but no longer maintainable by c.Kleinhuis contact him for any data retrieval,
thanks and see you perhaps in 10 years again

this forum will stay online for reference
News: Support us via Flattr FLATTR Link
 
*
Welcome, Guest. Please login or register. March 29, 2024, 06:29:24 AM


Login with username, password and session length


The All New FractalForums is now in Public Beta Testing! Visit FractalForums.org and check it out!


Pages: [1]   Go Down
  Print  
Share this topic on DiggShare this topic on FacebookShare this topic on GoogleShare this topic on RedditShare this topic on StumbleUponShare this topic on Twitter
Author Topic: What parameters slow down fractal rendering most of all?  (Read 7997 times)
0 Members and 1 Guest are viewing this topic.
Sfumato
Safarist
******
Posts: 94



« on: August 07, 2011, 08:30:43 PM »

Dear friends, please share Your experience on how You work with MB. What I mean is when I start exploring a new fractal I have to wait for long when the first 'outlines' of the image rendered and I can understand what next step should be undertaken. I. e. each step takes minutes or even more.
What am I doing wrong?
Thank You in advance,
with kindest regards,
Sfumato.
« Last Edit: August 08, 2011, 08:23:19 PM by Sfumato » Logged
Sfumato
Safarist
******
Posts: 94



« Reply #1 on: August 08, 2011, 07:56:53 PM »

Of course I decrease max-iterations in 'Ray-tracing parameters' to <150, to say nothing of volumetric light (i. e. it is 'switched of' ;-)
P. S. Still can't figure out what a hardware should I use to render images with volumetric light enabled - ?!
Logged
taurus
Fractal Supremo
*****
Posts: 1175



profile.php?id=1339106810 @taurus_arts_66
WWW
« Reply #2 on: August 08, 2011, 10:53:11 PM »

rendering with mandelbulber needs first of all processing power. the sandy bridge core i platform or the (old) phenom II hexa cores provides the best price/benefit ratio for now. if you want to render print resolutions, 8gig + of memory and a 64bit operating system are useful. the rest is rather not important for rendering speed.

the "ray step factor" is an important value for rendering speed. make sure to optimize it before rendering.

to primary question: check the image size. 480p are quite enough for developing. make sure time cosuming params like dof, volumetric light or ambient occlusion (not the post processing one) are turned off or only used when you've found the right pov.
« Last Edit: August 08, 2011, 11:05:27 PM by taurus66 » Logged

when life offers you a lemon, get yourself some salt and tequila!
Sfumato
Safarist
******
Posts: 94



« Reply #3 on: August 08, 2011, 11:44:43 PM »

Thank You very much, taurus!
My PC is Intel Quad 9650 (3.4 GHz) with 16GB DDR3, OS is 64-bit.
I'll be more attentive to the parameters You mentioned, grateful for Your soon and comprehensive reply!
With kindest regards,
Sfumato.
Logged
junkdog
Guest
« Reply #4 on: August 09, 2011, 06:39:29 PM »

What kind of fractals do you typically work with? Some formulas take much longer to render and don't really play (as) nice with one another.

I've never touched max iterations for the sake of reducing render time; like taurus66 said, the DE step factor (raystep factor) plays a huge role in render time.

In general: plain, unrotated, mandelboxes are rather quick. As are KIFS. FoldintIntPow is also rather snappy, under most circumstances.

I've tried and failed working with volumetric lights. If your parameters exhibit any DE Stepping errors, volumetric lights are a no-go (or I'm wrong, since I've never actually managed to get them working for any of my own creations - but they all have some DE stepping errors).

If you have the screen estate, try running two instances of MB. One can easily copy-to/copy-from the clipboard. I find it very handy when exploring or tweaking parameters; makes for a more interactive experience.

For setting up lights and colors for images that you plan to render using ambient occlusion, I typically render it at a lowish resolution (480p or 960p) with ambient occlusion quality set to 2. It's only slightly slower than screen space ambient occlusion but gives you a good sense colors and light sources.
Logged
Sfumato
Safarist
******
Posts: 94



« Reply #5 on: August 09, 2011, 06:52:54 PM »

Thank You, junkdog! Gratefull for Your prompt and valuable tips. Btw all my 'creations' have DE stepping errors and I was just about to ask Forum folks about them: What am I doing wrong?
With kindest regards,
Sfumato.
Logged
junkdog
Guest
« Reply #6 on: August 09, 2011, 07:23:01 PM »

Happy to be of assistance smiley

As for the DE stepping errors; *sometimes* it's possible to get rid of them by lowering the DE step factor - but after one goes below 0.1, it just takes too long to render imo. Also, for hybrids, you should always check linear DE stepping (sorry if I'm stating the obvious).

Some things that typically cause DE errors: rotated boxes (or even worse, boxes with XYZ offsets), any change to the fractal constant factor, additional folding (under the mandelbox and the fractal tab)...  or just hybrids in general.

Sometimes, replacing a formula for another "similar" one rectifies some of the DE errors without noticeably changing the geometry.
Logged
ihcn
Guest
« Reply #7 on: August 09, 2011, 07:40:46 PM »

The DE step error only matters if you actually see artifacts in the rendered image. What kind of error are we talking about? I regularly get 0.5% on images that look perfectly fine.
Logged
junkdog
Guest
« Reply #8 on: August 09, 2011, 07:44:03 PM »

Thanks for clarifying, I meant it in relation to volumetric lights. Any DE errors seems to cause the render time to increase exponentially when using volumetric lights.
Logged
Buddhi
Fractal Iambus
***
Posts: 895



WWW
« Reply #9 on: August 09, 2011, 09:17:48 PM »

If "DE error" is high (above 1%) it means that raymarching algorithm is not enough accurate. You will observe some artifacts on the image (noise). When it is lower than 0.5%, the noise is almost invisible on images, especially where there is a lot of details. Always try to play with DE Step factor, because it has the biggest influence on rendering speed. You can use LQ (Low quality) or HQ (high quality) buttons to find "optimal" value of DE factor (it uses DE error value to find optimal DE factor). Then you can tune DE factor manually (check quality of image, not the "DE error").

Try not to change Max. iteration. In most cases It has no influence on rendering speed. Difference is only visible when you are rendering some cross-sections (using limits). Normally "rays" never reach high iteration regions, because may-marching threshold is based on the distance (not on maximum number of iterations). You can watch it in histograms window. On left side there is number of iterations histogram (scale is up to 64 iterations). For mandelbulbs and mandelboxes average number of iterations is between 5 and 20 (peak on histogram is on the left or in the middle).

More details about this topic on: http://sites.google.com/site/mandelbulber/user-manual#TOC-Ray-tracing-parameters

About volumetric light it increases "DE error" value, because calculating of light rays uses not accurate ray-marching algorithm (for faster rendering). Fractal details will be still accurate. Don't matter about this. If image looks good, just ignore this. To get faster rendering of volumetric light you can decrease quality of this effect (default is 5) for instance to 1.0. rendering will be 5 times faster but there will be more noise on rays. On high resolution image noise is less visible.

About the formulas, the fastest is "Tglad's folmula Mandelbox". The slowest are Xenodreambuie's formula and all hybrids.
One remark about hybrids. When in hybrid sequence the most of iterations are Mandelbox or some kind of IFS formulas always use "Linear DE Mode". Then ray-marching is much more accurate and you don't have to reduce DE Step factor (rendering is faster). Sometimes you have to try both modes and compare the result.

The slowest effects:
- volumetic light
- not DE shading mode - it uses special algorithm for calculating normal vectors. It's very slow but surface is always smooth.
- ambient occlusion based on rays (not this from Post effects) (speed depends on quality parameter - higher value = higher number of rays = much slower). Sometimes quality = 2 is enough.
- additional light sources. Rendering time = normal time * number of light sources (sometimes faster when lights are not so bright - there is some optimization)
Logged

Sfumato
Safarist
******
Posts: 94



« Reply #10 on: August 09, 2011, 09:45:25 PM »

Thank You very much, Buddhi, for Your substantial reply!
With kindest regards,
Sfumato.
Logged
taurus
Fractal Supremo
*****
Posts: 1175



profile.php?id=1339106810 @taurus_arts_66
WWW
« Reply #11 on: August 10, 2011, 10:02:01 AM »

thanks from me too buddhi. a lot of useful knoledge about tuning the de step effectively!  O K !!
Logged

when life offers you a lemon, get yourself some salt and tequila!
Xazo-Tak
Navigator
*****
Posts: 77



« Reply #12 on: September 05, 2011, 06:01:37 AM »

Thank You very much, taurus!
My PC is Intel Quad 9650 (3.4 GHz) with 16GB DDR3, OS is 64-bit.
I'll be more attentive to the parameters You mentioned, grateful for Your soon and comprehensive reply!
With kindest regards,
Sfumato.
It takes minutes a step to render with THAT?! Are you trying to make a super high res large poster or something?
Logged

I lost the game!
tryptophan
Safarist
******
Posts: 86


« Reply #13 on: September 08, 2011, 08:15:51 PM »

I'm running a 4.5g ghz quad core sandy bridge and I find working with just about any fractals (fractal software) excruciatingly slow except for fragmentarium which uses the GPU to render. Anything slower than .5 secs I find difficult to work with but the fractals mandlebulber creates are worth the wait. I'm super stoked Buddhi is working on a GPU preview. Mandlebulber running as fast as Fragmentarium (for preview) and then software render for final output == awsome!

Logged
Pages: [1]   Go Down
  Print  
 
Jump to:  


Powered by MySQL Powered by PHP Powered by SMF 1.1.21 | SMF © 2015, Simple Machines

Valid XHTML 1.0! Valid CSS! Dilber MC Theme by HarzeM
Page created in 0.196 seconds with 28 queries. (Pretty URLs adds 0.008s, 2q)