Deep zooms

[Dinkydau]

Hello,

On YouTube you can find lots of videos of deep zooms into the mandelbrot set. When I open Ultra Fractal and start zooming, at some point it is going to take over half an hour to render a 640×480 image. The magnification is only 6.1367913E127 which does not seem to be extremely high according to the YouTube videos.

I use the Mandelbrot (built in) fractal. My processor is a 4-core 2,66 GHz. All cores are used when rendering.

Either I am doing something completely wrong, or those YouTubers must have like supercomputers to render their animations. Half an hour of 10,64 GHz of processing power for only one frame! 

Any suggestions?

[David Makin]

I think you just need to be patient.

Note that before now I have done disk renders that take up to a month - in fact that was just a static image at 8000*6000.
I quite often do animations that take up to a week to render and all but the shortest ones I've done take at least a day.

Also UF isn't specifically designed for optimum deep-zooming, I think there is other software that may be better equipped for the task - in fact I'm just guessing but I think some of the deep-zoomers may be using their own private software

[lkmitch]

Dave's right--patience is key when deep zooming.  I haven't looked at any of the zoom animations on YouTube, so this may be a moot point--the number of iterations is probably more important than the zoom level in assessing how long a render will take.  That is, I've done regular zoom (magnification of a few billion) renders that took days because they used a few billion iterations.  Conversely, one can do a deep zoom in a day if there are relatively few iterations (zooming in to a feature on the spike, for example).  But, essentially, you're asking for a lot of computations from your rig.  As Dave says, UF is not (to my knowledge) optimized for deep zooming--I think we trade that for formula flexibility.  So, be patient.

Kerry

[Dinkydau]

I think you just need to be patient.

Note that before now I have done disk renders that take up to a month - in fact that was just a static image at 8000*6000.
I quite often do animations that take up to a week to render and all but the shortest ones I've done take at least a day.

Also UF isn't specifically designed for optimum deep-zooming, I think there is other software that may be better equipped for the task - in fact I'm just guessing but I think some of the deep-zoomers may be using their own private software

Dave's right--patience is key when deep zooming.  I haven't looked at any of the zoom animations on YouTube, so this may be a moot point--the number of iterations is probably more important than the zoom level in assessing how long a render will take.  That is, I've done regular zoom (magnification of a few billion) renders that took days because they used a few billion iterations.  Conversely, one can do a deep zoom in a day if there are relatively few iterations (zooming in to a feature on the spike, for example).  But, essentially, you're asking for a lot of computations from your rig.  As Dave says, UF is not (to my knowledge) optimized for deep zooming--I think we trade that for formula flexibility.  So, be patient.

Kerry

Sorry for my late reaction

Yes, the iterations are certainly very important, but even then...

I'm used to wait a month for animations to be rendered, but this would take much and much longer. That is to get a really, really deep zoom of course.

Is there any free software that you'd advise me to use instead of Ultra fractal?

[David Makin]

Is there any free software that you'd advise me to use instead of Ultra fractal?

I can't really help there, I generally avoid zooming further than is possible with normal double precision - usually sticking to <*1e14.

[Dinkydau]

Thank you for at least trying to help. FractalExtreme seems to render animations really fast, but the trial has got its limitations... and of course I want to make the deepest zoom ever without paying anything for the software. 

[bib]

Hi there 

Just like Dinkydau, I find that arbitrary precision is too long in Ultrafractal.... so no deep zoom videos! But, I have already calculated several days that I don't want to lose (the remainder of the calculation would take months...)

So does anyone know if it's possible to interrupt a render that is taking too long, without losing all the long hours (days...) already calculated? Is there anyway for example to tranform a .urj file direclty into an .avi ?


Thanks

bib

[JackOfTraDeZ]

whimper snivel whine!  my current animation to E+112 is 15 months ongoing now, estimated another 6 to complete.  the frames take about a day each.  patience my friend - good things come to those who wait.  besides, 4 systems with 24/7 fractal displays makes a nice permanent lightshow in my living room, especially at night...

[Dinkydau]

Omg, how does that take 1 day per frame? Are you talking about the mandelbrot set? I am currently rendering an image at 1.8047813E1100 which takes only 5 hours.

But yes, I guess patience is the key, 10 times more patience than what I was used to, and what the people around me call me crazy for.

Edit: Does anybody know how many processors or processor cores Ultra Fractal supports?

[David Makin]

Edit: Does anybody know how many processors or processor cores Ultra Fractal supports?

Quite a lot - I believe as many as the OS will support

[bib]

Regarding the number of supported cores by UF, its not crystal clear. In preview mode, my 8 logical cores are at 100%, but when rendering, some fractals use 100% CPU, especially deep. Zooms, and some others use less than 10%, even multilayer images. Does anyone have a clue?

[David Makin]

Have you tried changing both "min threads" and "max threads" ?
e.g. if you have 8 cores and a 5 layer fractal then I would set min to 8 and max to 40.
To be honest I've never checked what happens with respect to full use of the CPU/FPU either when rendering inside UF or to disk.

[bib]

Yes. Min threads is 8, max is 32 and I hardly use more than 3 layers by image. It seems to depend on the formula. For example deep zooms on simple formulae and your MMF-4D are using full CPU, and in general, Herman fractals are using about 20% when rendering to disk.

[Dinkydau]

Have you tried changing both "min threads" and "max threads" ?
e.g. if you have 8 cores and a 5 layer fractal then I would set min to 8 and max to 40.
To be honest I've never checked what happens with respect to full use of the CPU/FPU either when rendering inside UF or to disk.

In other words, the minimal value is the amount of cores, the maximum is the amoun of cores times the amount of layers?

Edit: Does anybody know how many processors or processor cores Ultra Fractal supports?

Quite a lot - I believe as many as the OS will support

So if I have 1024 cores (extreme numer for example) and set the minimal threads to 1024, it uses all 1024? If you don't know that doesn't really matter, I'm just interested. If it supports 16 cores that would be enough I think for the time coming.

[David Makin]

Have you tried changing both "min threads" and "max threads" ?
e.g. if you have 8 cores and a 5 layer fractal then I would set min to 8 and max to 40.
To be honest I've never checked what happens with respect to full use of the CPU/FPU either when rendering inside UF or to disk.

In other words, the minimal value is the amount of cores, the maximum is the amoun of cores times the amount of layers?

Edit: Does anybody know how many processors or processor cores Ultra Fractal supports?

Quite a lot - I believe as many as the OS will support

So if I have 1024 cores (extreme numer for example) and set the minimal threads to 1024, it uses all 1024? If you don't know that doesn't really matter, I'm just interested. If it supports 16 cores that would be enough I think for the time coming.

Serves me right for not checking - it seems that the maximum you can set either min or max threads to is 32 - unless that limit is based on detection of how many "cores" (plus hyperthreading) are available or specifically on my system in which case the limit for 2 cores or 1 hyperthreaded or for my system is 32.

I don't actually know the details of Frederik's diisk rendering algorithm but I *think* it renders in sections, each section essentially the same way rendering within UF renders a full image i.e. when a thread has done it's part of the current section it does part of the section initially allocated to threads that haven't finished yet.

Even if rendering just one layer if you have min threads set to 32 then it will use 32 threads (by splitting the area) as 32 cores if you have them