Very slow location

[simon.snake]

Hi

I'm trying to render the following parameters and on my aging laptop it's taking much longer than it ever did previously.

I realise it's quite a deep location, and I'm trying to render at 7680x4320 which is pretty large, but surely it shouldn't take so long.  I interrupted it at 1hr 10 minutes and it still hadn't drawn the first full screen of picture.

Can anyone else try the image for me and let me know how long it takes them.

[ellarien]

I've started my Core i7 desktop on rendering just the first frame at that size. I'll check back on it in about 8 hours and let you know if it finished yet. 

(It only seems to be using one thread, though. It's been a while, but I thought KF could make better use of the processor than that! This is the latest version, and I haven't really played with it yet.)

[ellarien]

Update: 9 h 45 minutes later we're on reference 36 and still trucking. It was using the whole processor at least some of the time. It's a pretty location, but at ~3 million iterations all over I'm not surprised it's taking a while.

[simon.snake]

Thanks.

On my laptop I usually put it in to hibernation overnight, and then KF continues from where it left off, but this morning I switched on and KF just sat there with the clock completely stopped.  It had already been going an hour or so last night.  I'll have to stick with running at a lower resolution image for now.

[ellarien]

It's all done at just under ten hours.

https://www.dropbox.com/s/wgy9ujhrrxzjylm/ssnake_slow.jpg?dl=0

Enjoy!

[quaz0r]

(It only seems to be using one thread, though. It's been a while, but I thought KF could make better use of the processor than that! This is the latest version, and I haven't really played with it yet.)

unfortunately even with perturbation you are still limited by whether you can calculate the reference points in a reasonable amount of time, which is a purely linear task, iterating a point from beginning to end at full precision.

[Kalles Fraktaler]

Yes that is a pretty location, I rendered it in 20 minutes and 11 references in 640x360
The number of approximation terms is related to the resolution.
The more terms the longer time for calculation of the approximation, which is unfortunately single-threaded (also in MM...)
The larger resolution, the more time can be spent on approximation that will be saved for less calculation per pixel.
Unfortunately I haven't done any big investigations of the optimal relations of this
For 7680x4320 a full 120 terms will be used, compared to 10 for 640x360, so it might be worth trying to manually set a lower value (in the Iterations dialog)

[ellarien]

The number of approximation terms is related to the resolution.
The more terms the longer time for calculation of the approximation, which is unfortunately single-threaded (also in MM...)

Ah, I guess that explains it. I'm not used to rendering at such a high resolution -- probably the single-threaded phase isn't  noticeable when rendering at the resolutions I usually use.

[simon.snake]

For 7680x4320 a full 120 terms will be used, compared to 10 for 640x360, so it might be worth trying to manually set a lower value (in the Iterations dialog)

I've noticed that if I leave KF set to auto iterations (which I usually do), the number in the iterations box can go really high, but the maximum for the frame is much lower, therefore taking longer to render.

If I then change it manually (to a much lower value) for one render, the next zoom in starts with the original (much higher) value, not what I've set it to.

Is this the normal behaviour or can it be overridden?  I don't mind setting it lower, but there must be a way to have auto iterations take that lower value and increment that next time.

Hope that makes sense.

[TheRedshiftRider]

Decreasing iterations is a good solution sometimes
When I rendered some of my 16000x9000 images on a 64bit laptop I had a completely different problem where the program itself did not react at all. When making large images depth and density matter, when there are too many iterations it can bug out the calculation process. Lowering the iterations did not have a result of black areas even in those images and it gave a massive boost.

[Kalles Fraktaler]

The number of iterations is always automatically adjusted since I haven't encountered any problems with it myself.
Since KF doesn't calculate more iterations than necessary as reference, i.e. stops at bailout, the only issue with entering an unreasonable high value would be that the array of iterations use unnecessary lot of memory, not that the render get slower.
So I think you both encounter slow render due to too many approximation terms?

[TheRedshiftRider]

Now that I think about it. I only have the problem with the iterations. But I will have a look at the approximation. I don't have problems with long render times. So I guess I have to use higher approximation to be able to render with large iteration values, but to let it take a bit longer?