Title: Baby Monster Post by: Dinkydau on December 04, 2014, 11:31:25 PM (http://th02.deviantart.net/fs70/PRE/f/2014/338/8/e/baby_monster_by_dinkydauset-d88olq0.png) (http://fav.me/d88olq0)
Mandel machine, mandelbrot set This is an enormous intricate and symmetrical structure found at at depth of 2^6901.5 in the mandelbrot set. There are two types of trees, there's shape-stacking with fork-shapes going on, and the main structure it itself a tree. I estimate that the pixels in this image have iteration counts around 35 000 000. I rendered this a few weeks ago at larger resolution, but mandel machine crashed and I have not been able to reload the file containing the progress. You can click the image to see the 4000×3000 image on deviantart. Named after the baby monster group which is also an enormous intricate symmetrical structure. Magnification: 2^6901.591 3.8539742140068607747447368372813 E2077 Coordinates: Code: Re = 0.4026223242141852148037824724584025835755135011342881027042525627039569502881670946745557412930327115393258794913548834454537746442641230778661250740808696667685167031459830818673342528174897920124209545137508922550549968221550363823064128549715028301461622593069839496780093952808237573205197164572673594064416772605893796305663235635527880132594672201779519565231079208219169990178251347386228988056919160862667394171741551928464324300754658056232719271701044207283382298845257395452425798232763791743779314353854383722879333193167946859425243707261854429090471722384741345982162350257163211439072535377347476863063971161548476045589432821456788895124531684440296396587503362225268140306547818930627963692042601813606866625158984034539104729252270036522838631376204929633117350514392534098621950298874921561848480721715466716034612641357006597673782157926191463771077735121499521415799630479650549024192802530771425233126786755033877940979179549256821540572419329726522803422394291516882903607750877675838298417678464016201757067174161515701977269714418733430534088597594663546370996913330820306039197466323342547246232849027813752870111841998012412400077840383443533115591420661427063968799614034835064533287768378088050419077884398682152379426978057892246066882890819414748458464211672718229513050617638747371353729091501912075437789263019440768014217748069621815249161576890441895414789868155359854876653884160567578413850597585526839599148529091786964892859271522516203113708568784835766360055829328924133199024188688844937378174198793904588311188393798698747503071556711010350663949412136302614381521786902283300674108990630100130448100638953468850976478379826240732566854132975350164428623726436941877225968956042433264823693750422381553726384657388053048021285450658976425595261910844568695172761539871359485880883688256972852102360571149799163601150837375544386512748326612079156006358192675222805435405708616040869538348696626733367203267216061599286266159729225868003852298996109238906739358339973449675179531666048724844664052031271387710787740075700568712301627930122446 Title: Re: Baby Monster Post by: Chillheimer on December 05, 2014, 08:52:31 AM I love your deep zoom pics.. this one is great!
you must be a vers patient guy ;) I never reach those depths.. I've been to e1000 just 3 times now. I tend to zoom into high densitiy areas, leading to high iterations very soon.. Title: Re: Baby Monster Post by: TheRedshiftRider on December 05, 2014, 09:16:58 AM This one looks very interesting.
Title: Re: Baby Monster Post by: Kalles Fraktaler on December 05, 2014, 09:28:52 AM Amazing!
What do you do when you sculpted an interesting shape with deep Julia morphing? You just continue sculpting a new shape with the previous shape just as a small part of it! Fantastic! And it was already there infinitely long before the universe was born in the big bang. As all the others are, that haven't been discovered yet. Title: Re: Baby Monster Post by: Chillheimer on December 05, 2014, 09:40:33 AM And it was already there infinitely long before the universe was born in the big bang. yeah!As all the others are, that haven't been discovered yet. and yet, at the same the m-set is 'just' a simple mirror, depicting a "lowres"-version of all that is around us in the cosmos.. fractals are just so omnipresent and ... awesome! I will never get over the fact that so few people care, that there is no serious scientific discussion about the big WHY fractals are everywhere. god here, religion there, wars fought, science debunked, new theories emerge.. yet, no one seriously researches the obvious.. :-\ but being here and exchange with you guys makes it at least a little less annoying, and to be honest not feel so alone with these views.. and it's also very cool - as I stated in another thread- to be the very first to see these 'places'.. one of the few realms left, where everyone can be the very first explorer. sitting at home, staring at beauty on a screen. :) and: if some people keep saying that just zooming into fractals (without any post-production) isn't real art because it has all been there, it's pictures like the deepzooms from you dinkydau, that prove different. with the shapestacking techniques you use, it is as much creative sculpting as chiseling a block of marble. the final statue has also "been in there" and creating the right palette is as much an artform as choosing and mixing "real" colours to paint a picture. out of curiosity: how many hours/days did it take you to reach that location? Title: Re: Baby Monster Post by: SeryZone on December 05, 2014, 03:50:22 PM This is awesome! Really... Your work is legendary.
Title: Re: Baby Monster Post by: simon.snake on December 05, 2014, 04:42:39 PM That is one pretty special deep zoom image. Just a slight change in the palette or colour mapping parameters would change this into something nowhere near as interesting. Part of the fun is to find the image, the other is to hone the palette to reveal maximum detail.
Totally outstanding. If I were to possess even 1% of your skill and talent I would be so happy. Very well done. Title: Re: Baby Monster Post by: youhn on December 05, 2014, 05:23:18 PM Wow.
:cantor_dance: Title: Re: Baby Monster Post by: Dinkydau on December 05, 2014, 05:57:19 PM Thank you very much everyone
I'm very happy with so many comments. Chillheimer indeed it's special that it's possible to be the first explorer of various things in the mandelbrot set (and other fractals). That makes it very worthwhile to do for me. Some time ago, SeryZone found (likely) the first S completely made up of other Ss, and Kalles fraktaler was (likely) first to explore the stretched parts of some other fractals. I didn't record how much time I have spent exactly. It took me more than a month to zoom to this. Some days I zoomed for a few hours, some days not at all. I started with kalles fraktaler and went to mandel machine when it was improved with greater maximum depth. Near 2^6900, calculation of the reference point took more than 10 minutes each time. The first render at 25600×19200 took slightly longer than a week on my old computer, and this one (16000×12000) about one day on my new computer with i7-5820k. simon.snake Revealing details in a suitable way was a challenge with this location. The main shape easily blends with the details in between, but I also didn't want to destroy the details to make it clearly visible. Otherwise the shape-stacking would have been done for nothing. Title: Re: Baby Monster Post by: Adam Majewski on December 06, 2014, 01:41:12 PM (http://th02.deviantart.net/fs70/PRE/f/2014/338/8/e/baby_monster_by_dinkydauset-d88olq0.png) (http://fav.me/d88olq0) Magnification: 2^6901.591 3.8539742140068607747447368372813 E2077 What means second parameter after magnification ? Title: Re: Baby Monster Post by: Chillheimer on December 06, 2014, 05:50:34 PM What means second parameter after magnification ? its the zoom depth Title: Re: Baby Monster Post by: Chillheimer on December 07, 2014, 01:42:31 PM hey dinkydau,
I'm wondering: -did you go past minibrots, or do you try to avoid them, to keep iterations lower? -do you zoom "blindly", just searching for the middle, without waiting for the full image to be calculated? -what imagesize do you use in deeper depth to explore? -do you really plan the shapestacking or is it kind of random most of the time? I mean, did you have this baby-monster in your mind and did every zoom on purpose to reach this? regards! Title: Re: Baby Monster Post by: Adam Majewski on December 07, 2014, 03:58:11 PM its the zoom depth Isn't it the different form of magnification : Code: 2^(6901.591) = 3.8539742140068607747447368372813E2077 = 3.8539742140068607747447368372813 *10^(2077) ? Title: Re: Baby Monster Post by: Dinkydau on December 07, 2014, 04:25:38 PM Isn't it the different form of magnification : Yes, that's it. Some programs use a power of 2, others use scientific notation.Code: 2^(6901.591) = 3.8539742140068607747447368372813E2077 = 3.8539742140068607747447368372813 *10^(2077) ? hey dinkydau, I don't really avoid minibrots. I zoomed past one minibrot here for the fork-shapes but I didn't need more.I'm wondering: -did you go past minibrots, or do you try to avoid them, to keep iterations lower? -do you zoom "blindly", just searching for the middle, without waiting for the full image to be calculated? -what imagesize do you use in deeper depth to explore? -do you really plan the shapestacking or is it kind of random most of the time? I mean, did you have this baby-monster in your mind and did every zoom on purpose to reach this? regards! For long periods of zooming in the center, I first calculate the depth of the next shape where I have to go offcenter, so that I know where to stop. Then it's safely possible to zoom blindly. The image size I have used during zooming was 1280×960. I planned everything. My idea was to make an S full of two types of trees and forks in between. Then I decided how long the forks were going to be and which types of trees to make. When I was on my way I realized that making an S would be undoable because the speed of the calculations was just too low to go that deep (in kalles fraktaler). With mandel machine I have been able to continue up until here and it was hard. So yeah, I had this baby monster in my mind, but I had not planned to stop here. Title: Re: Baby Monster Post by: SeryZone on December 07, 2014, 07:07:38 PM If mandel machine will update to infinite zoom, will you zoom it further???? That's would be crazy!
Title: Re: Baby Monster Post by: quaz0r on December 07, 2014, 09:15:00 PM good work sir. its cool to see you guys really getting into it. :)
Title: Re: Baby Monster Post by: Dinkydau on December 07, 2014, 11:47:12 PM If mandel machine will update to infinite zoom, will you zoom it further???? That's would be crazy! I could try it.Title: Re: Baby Monster Post by: quaz0r on December 08, 2014, 12:02:45 AM have you guys made a collection of the deepest locations people have been to or anything like that? that would be cool both for testing and exploring.
Title: Re: Baby Monster Post by: Botond Kósa on December 08, 2014, 12:21:52 AM The first render at 25600×19200 took slightly longer than a week on my old computer, and this one (16000×12000) about one day on my new computer with i7-5820k. Congrats for your new computer, with 6 HT cores it seems 2-3 times faster than your old one with 32 cores, right? More and more people are having Haswell CPUs supporting AVX2, I guess I should soon implement the FMA routines. Title: Re: Baby Monster Post by: 3dickulus on December 08, 2014, 12:48:16 AM The last frontier? it's an infinite frontier for sure. with each push to deeper and deeper depths entire new realms are available to explore.
Cheers to the trail blazers :beer: Title: Re: Baby Monster Post by: Dinkydau on December 08, 2014, 01:02:00 AM Congrats for your new computer, with 6 HT cores it seems 2-3 times faster than your old one with 32 cores, right? More and more people are having Haswell CPUs supporting AVX2, I guess I should soon implement the FMA routines. Reference points are calculated 3,1 times faster. I don't have something to compare how good it is at computing pixels without full precision, but I'm pretty sure this one CPU beats the two opteron 6272s that I had. The speed-ups are at the cost of almost half the performance in fractal extreme.I think AVX2 can be used to speed-up reference calculations. Is that correct? Title: Re: Baby Monster Post by: Botond Kósa on December 08, 2014, 09:08:30 AM Reference points are calculated 3,1 times faster. I don't have something to compare how good it is at computing pixels without full precision, but I'm pretty sure this one CPU beats the two opteron 6272s that I had. The speed-ups are at the cost of almost half the performance in fractal extreme. That seems reasonable. The i7 CPU offers 3.1x single threaded performance on 6 cores instead of 32, resulting in 3.1 x 6 / 32 = 0.58x multithreaded performance compared to the Opterons. But this is only true for integer code. (References in MM and full precision calculations in MM and Fractal Extreme are calculated using integer numbers.) Per-thread floating point performance on the i7 should be about 4x faster than the Opterons, resulting in more than 2x multithreaded performance.Could you make a comparison of the two systems by calculating a deep and large image? Floating point throughput is displayed in the Statistics box under GFLOPS. Title: Re: Baby Monster Post by: Botond Kósa on December 08, 2014, 09:22:50 AM I think AVX2 can be used to speed-up reference calculations. Is that correct? Not really. AVX2 contains the MULX instruction that may simplify the integer multiplications used in reference calculations (details can be found in this article (http://www.intel.com/content/dam/www/public/us/en/documents/white-papers/ia-large-integer-arithmetic-paper.pdf)), but I only expect a speed improvement below 5%. But the FMA instructions could speed up floating point calculations by 50-100%. Title: Re: Baby Monster Post by: Dinkydau on December 08, 2014, 05:17:10 PM That's less than I had imagined. On wikipedia AVX2 is described as an "integer version of AVX".
Title: Re: Baby Monster Post by: Botond Kósa on December 08, 2014, 06:43:36 PM That's less than I had imagined. On wikipedia AVX2 is described as an "integer version of AVX". That's right, AVX2 extends all integer SSE instructions to 256-bit wide. But the full precision reference calculations in MM use no SSE instructions, they are based on the 64-bit MUL instruction that multiplies two 64-bit integers and produces a 128-bit result. Unfortunately there is no SSE version of the 64-bit MUL. |