Question About Render Speeds on Zoom

[Iggshuybiryel]

Hi,

Been a long time since I visited these forums. I had stopped doing much with the software for a couple of years, then recently installed a version and re-familiarized myself with the program. The reason I'm here today is because I just realized there was a newer version of Mandelbulber and updated again a few days ago. Some of the changes make sense or don't cause me too much confusion. But I'm finding when I click to zoom in my render times seem to start increasing dramatically, more and more as I go it seems. Under the last version I used, I could set the resolution and quality settings and it would take about the same length of time to render as I kept zooming in. I'm not sure if it was because I was doing something differently before or not, but I seemed to at least have a sense of how long it would take to render on subsequent zooms. I'm finding that now, however, it varies depending on where I click sometimes, and others, I can discern no apparent cause for the increase in render times. Yesterday, for example, I zoomed in to find the render time increased from hours to days in one or two clicks. I canceled that and tried something else, so I don't have settings or image to share. Is there something very different about this version of the software from previous versions causing something like this and is expected, or is it likely something unique to my settings, hardware, and experience? 

Thank you in advance for any helpful insights you may offer.

[mclarekin]

From my experience I have always found, as general rule, zooming in increases render time.  I think the reasons are to do with the way Distance Estimation works.

The first situation is when you are zoomed far out.  A screen showing an object one pixel in size is the fastest. As the proportion of the screen that displays the fractal object increases, the render time increases.This is because calculating DE for the blank areas is fast.

Once you are inside the fractal the zooming will continue to increase render time.  The DE  algorithm includes a reduction in accuracy for a surface in the distance, so for those parts the DE calc works quicker saving time.  When the distance to the surface is really small then there is no large savings in render time.

There maybe other reasons as well  (Buddhi ??)

The fundamentals are described in this documents:

http://cdn.mandelbulber.org/doc/Mandelbulber_Manual_v091.pdf

increased from hours to days in one or two click

I generally avoid anything that takes over an hour at 3200 x2400. This depends on your computer and the  actual fractal that you choose, as well as the zoom depth and effects.

If you post your settings here, I could take a look at them, and see if  they are set to the optimum.

[Iggshuybiryel]

Thank you, mclarekin. I will check the manual more closely and continue to play with test renders and settings. As for the experience I described with a few fractals, I didn't keep the settings. The fractals I've explored today behaved pretty much in the manner you described. The ones before were using hybrids with the same basic combinations, so perhaps it had something to do with the complexity of the particular combination choices I made.

One other thing, although it isn't related to my original post--I'll ask here anyway to avoid clutter in the forum. I'm noticing that Mandelbulber is not using my GPU to process the renders at all. I thought that feature was added some time ago. But GPU-Z shows that it clearly does not use it. My monitors use the amount needed to power their use, but nothing added to it on my GPU during renders at all. Can anyone shed light on this? Was I mistaken in thinking that GPU use was added, or am I missing something?

Thank you again.

[mclarekin]

Yes the amount and type of function used in the formula has a lot to do with render time. Menger Sponge types are fast they have very little maths, and it is all +-* which are fast functions.

Mandelbulber V1.21 OpenCL  was the last GPU type. Then Buddhi rebuilt the program in QT (which is a long process) and for various reasons openCL has not been implemented. Mandelbulber V2.10 should  be released soon, and hopefully OpenCL will come with the next version, V2.11.

Using the old V1.21 with openCL is still a lot of fun, but you need to use the custom formula part, to really get the most out of it.

[Iggshuybiryel]

Okay, got it. Thank you so much for the answers. My mind's much more clear on several things now. Take care!