Help understanding factors that determine render speed

[Weber]

I'm rendering out an animation that has over 12,000 frames and in the last 12 hours I was only able to render 220 frames...  This is almost unbearable since I've got a decent system, so I don't understand why it's taking over 3 minutes to render a single frame especially with mid settings?  So (3 mins x 220 frames = 11 hours give or take).  

I've posted my system specs and render time for this benchmark here: http://www.fractalforums.com/mandelbulb-3d/recommendations-on-cpu-selection-building-a-rig-dedicated-for-mb3d-rendering/15/

And here are the times for my current project, but at some point it takes even longer

Main calc time: 2.41.1
HS calc time: .23
AO calc time: .05

My formula stack is simply: MixPinski4/ABoxVaryScale.
  
I'm rendering at 1920x1080 at the following settings:
DE stop 8.8
Raystep multiplyer .4
Stepwidth limiter .75
Stepcount for binary search .8
Smooth normals 1

Post Processing:
Hard Shadows/Ambient Shadows

So what factors cause render times to go up?  I've rendered other videos at even higher settings which have taken less time with.  I suppose this is why I haven't made very many videos, since I only have one machine to render with which is usually done over night while I'm asleep, or away from the house etc..

[xahhax]

raystep multiplier does impact speed...(higher, less time). I think also #iterations in each formula and lots of other parameters

Your machine isn't particularly fast looking at this thread (I also posted there).
I render in 4K (3840*2160) and it takes me on average 1min per frame on a 12 core machine. so this seems about same factor.

I think it is difficult to avoid a dedicated rendering rig if you are doing lots of animations...

[cKleinhuis]

some images need small raystep settings to avoid the overstepping, the iteration count is usually very high and 10 to 15 iterations is usually enough, the iteration is always reached when bumped in a solid area, but most of the time the ray marcher is outside the max iterations, nevertheless fine tuning lowering the iterations may help a bit

check your values, when you resize the pic mb3d adjusts these values, disable that options, and fine tune everything for the dedicated movie render resolution

[Weber]

Thanks for the help guys, and the iterations is actually set to 1 on each formula.  And I've tried adjusting other settings such as Raystep multiplier and others, but I end end with too noisy an image(I hate little black pixels).  I think it's as good as its going to get and I'm just gonna have to accept it the way it is.... I hit 890 frames so far...

[xahhax]

Keep it up :-)

[cKleinhuis]

it actually is accepting the reality, i have done a few renderings that all took more or less two weeks to render like 2500 frames in fullhd but when deep inside a highly complex object the render settings just go through the roof, some formulas behave especially nasty in regards to rendering time, e.g. the random ifs boxes, one of the formulas you included takes more processing power

[Weber]

This is not a normal box looking fractal, it's a very noisy one with lots of rotations used to get it to look the way I desired it to.  It has an earthy look to it, and as a result, some normals will only appear smooth with very low raystep multiplier settings.  And if I adjust the DE stop to try remove the noisy pixels I end up also losing the precious detail that I desire as well.  I think I actually found a happy medium though, I'm now rendering at 2:10 total time per frame, so I saved some time there and I'm happy enough with how it looks.  The raystep multiplier is now .7 with a DE stop 9.2.

[zebastian]

@kleinhuis
Totally agree: i once did a animation render in mandelbulber. Disabled fancy effects decreased image size and made a testrender. Worked really fast, good to go for the final render.
I changed the settings back, rendered an hour and calculated the finish for ~10 hours. I went to bed and expected a finished animation. Nada.
What happened was that after roughly 2 hours i got really close to the surface. I got a dof effect applied which increased the render time (square)antiproportional and the last 5 images took the whole time.

So now for estimating the render time i do:
1. Render full animation with smaller image size. Measure after 5 frames or 10 frames and at the End.
2. Then Start the high res render and check after 5 or 10 frames.
3. make an educated guess

[Weber]

I decided to start rendering another animation instead.  The new one is rendering much faster and I have a whole bunch of cool ideas.  It's also gonna be my first time using positional lighting. 

I am currently rendering 4k 3,840x2160 @1:45 per frame, some frames are taking even less time.  This new one also has 4 formulas instead of two. 

[Patryk Kizny]

Guys, I suggest you look into Mandelbulber 3D + OpenCL or GLSL stuff. That renders som much faster!