Deep zooms

[David Makin]

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.

Can you post an example UPR ? i.e. that exhibits <100% CPU use when disk rendered ?

[Dinkydau]

Does that mean Ultra fractal supports 32 cores?

[David Makin]

Does that mean Ultra fractal supports 32 cores?

Quote from the manual:
------------
If you have a processor with HyperThreading, Ultra Fractal will recognize it as a dual processor and split up its calculations accordingly. In most cases, this will result in a modest speed improvement. Otherwise, you can force Ultra Fractal to use just one processor.

Open the Options dialog and go to the Fractal tab. In the Advanced calculation options area, the Minimum number of threads option sets the minimum number of threads that Ultra Fractal will use for a single fractal window. If this is set to 1, calculations will not be subdivided. Typically, this should be set to the number of processors in your computer.

Ultra Fractal will typically use one thread per layer. If the number of layers is low, extra threads will be added (subdividing one or more layers) to reach the minimum number of threads setting, which ensures that all processors are used. However, if there are many layers, this would create a large number of threads, which would saturate the system.

The Maximum number of threads option limits the number of threads for a fractal window, making sure that complex fractals will not start an unlimited number of threads. By default, this is set to four times the number of processors in your computer. If you would like more layers to be calculated simultaneously, you can increase this setting.
------------

So I'd guess UF supports up to 32 full cores or 16 hyperthreaded but I'm not sure if the situation is exactly the same for the sections being rendered in a disk render as with a full image within UF.
I'd also wonder what would happen if you set min threads to 32 on a system with 32 hyperthreaded cores - is it smart enough to use one thread per core or would it use 16 cores with 2 threads each.

[Dinkydau]

Yes I'd guess that as well. Thank you!

[JackOfTraDeZ]

Someone said:  "I am currently rendering an image at 1.8047813E1100 which takes only 5 hours".

An image at that depth with fractint would take weeks to do at least; I would not even attempt it, and I'm no "fractal wussie" !

I have UF and it can import PAR files like the ones for my deep Mandelbrot zooms, but it consistently renders them much SLOWER than fractint, even though it's windows and a faster machine. That is why I never played much with UF for my animations. WHY is this? Do I have to tweak settings or something?  Doing something wrong ??

Can you send me or post this PAR and let me run it on my fastest machine with UF, see what time I get.

Thanx

[Dinkydau]

Well I must say that only a few thousands of iterations are used for that.
Here are some parameters:

E1103 {
::iJtJAin2tbZQvNuNQU47Bw/HE89N7MkcI50Cfpot36petALUtVSEWb5AZF0N/77HlymWgero
  92msI2rInHn59ezI+wc/xl+zf3u766WGXOPcY/PpqE339HjnWe6QOJdPNM+4TLHSVp7c/rDz
  3OodPM3fZ4Qybxdce404ytD7/xxpP/6p+X++6HDC/T897ubNkV8P2/8y41pD7/h+jf+x5rvM
  daf31n7POu86BVkd3dp/5nHnecb3DTLDzH+gev/v+nfrBk/fwPfDovB0/vAVTaxE3tSWCZJL
  atEDuahSKWr8nYWLeh1tiEXByksGkYKLJJXyeyMXLaoqqXFPaRpWCWOqhQycxBY6KLROMrmC
  gbaFoYsECmyZoWxBPOINJkMJrUIZqmFCcseOXzq5BeiU0YuYVNkI88GQ5YSsaNGNlDwjiErK
  JelMITZGcLRMhiq1QORdJSISJVMHkriprAxhVECRCegygyPHdPDjYSgyTqhaSqRNSNZKfQRt
  eKWJ488kvxREi4qrpG/EqFvBJT1ILJhULahYh8LlyhcsUVpYxYM5hShFUN7hNgKsjahiXlkJ
  kTuXgfA9GfD3afkhVSLVc0lEHZp4hoVqlcmaOtJ/oIBvAGeGKqQq2kMk/CsKCslVUCI6YcNj
  jFTh6Rr0KqaRdzWBqqVPUrWVyBj9ZZ+1rQXQykgwD5iYpcu41APloN+Owh0JVI+NgK5q5JOM
  Li6zGyc6V8kZHbVRakng9MDo+apieTOLYEQbLxk+VVjtoknSKrAKVUA2BCJliiuSVlaMjlIp
  BFNwzkE84YFDQEvcdFIZt69sF1q1MeYh912gWKqQRZRIFzakXc0qenXPBZBFQk8Gs6bqG7g4
  D0Tg2lrFepUKDnmIuM6DOpGNt2vUC0o4Z3whFR41UjSLbABNmzJgCTB2Va8gUDNrSgfb2Hwt
  gSLRcHhKkC4q5QQobDXfcrXTbmDcvSr/icKhH2CkLgqWoVgiN10B8EYPAEyaCCKsq0S1q2NO
  CGs26AxcRDJTNwwRGRiZOCklCUPSOt2OgmBwsJSomY0sG6bcEUdNTfKCDh2SNMtNRtRt8c2f
  sGcmtQdh5NIwQ0s1oK6GadfbzjcJ26y4sxExHEdCiuZlqNmFar6JOXEJ8p8HvlWOfVoHNCJv
  CUiRBEF5EOOUAGNlhog9jUSIoUr4eY8nRHM7gKjeDpRdGuADafbULZOj7gwZVivSfYT/lc0s
  WdSbOtgAFnH9EhK+7UbqqCbLtns5sx6QDM5Bm415RMIGuEOGSAXQ7/h43GBZkoasBal54YsJ
  1NxiblGqKGehBOkkN+EjBjSxjADxQxMSdaNDYwPozgE6IhGJBscTnKbZE6cChICWN6JSfOsg
  2sgEYSxIZQGJmz0G4xeFt9BvsIgFY4DIvbkd3l+HnOY37MDKETr3/b3dPcd+yLn7XvL2l+vM
  2uMGWDp75h5jPNc8zH4GcP+U3DjnHmaXGc/vu0Pdqf+09v8wl9tw4GczveY/P3fb5XYphz77
  e+T3W6nXOIfEg+0v3Pe+6Lc1y75yfjT3GPNsdT05+pbPwBOddaY3dsl/5SnHnG6n3d3jz9nG
  5kWX92lrX5WrvOcrb+6S/61N/AMTtbkz/L8dEyujXPfd+QjWY457rwuWv356iYp/rFiybPtN
  WmR3vvS013BrNzJ3A4tFxq81oq0O/1HXTv9UjXDwIbqutbA/3z/prvtd597HHMb3d/pi9FaC
}


Of course this does not look like somthing special.
My goal is to zoom in deep enough to end up seeing a mandelbrot set.

[JackOfTraDeZ]

I copied your post and saved it to a text file and named it "test.par", then navigated to it with UF 4.02 version.

That took only 1 hr 45 min to render at 640x480 on my Alienware Athlon64, my fastest system. I think it was using only 1 core, looking at task manager display.

I was amazed, then I saw the iteration count was less than 3000.

You will never see a mini-brot at that depth with so low iterations.

You need at least a million, and don't keep zooming into the "center" of the image; go off to one of the large standing waves where 4 points intersect.
From my experience, you are a long way to go.

[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.

Can you post an example UPR ? i.e. that exhibits <100% CPU use when disk rendered ?

Yes, this one for example:

HermanZoomAuxEntournures {
::08WYMjn2tr1SLuRSS47G8/Bh8ZVKjIfPDCML0wcwrPs9ecATZpq7uWrWVTJp12zv+9LzseES
  tZwePLhBrqUmRGfR8FPyg+h+6tnq3/bv9NLWcq90+mNL/jm+nrPs4v66eeR95vto5wpuz9HO
  33cc5iv2u70TbcG1inaaf8pTbMB1i91fvp/4GOJkH6rfu54GShPlH2EJs8HeBLIUp8KybN6g
  N6sp1vtvZX7pjbWW3fs5QzvHWz0aWpi/+nb/Meis5nW+23kPlsiut+lTtdH2s8DpXtgWuo7l
  6ttn++ml4cf/7U3ff+/Jn3GZe8JvKG4kkWs45m+HbeubHg7T1972ngC23H/o4phdj1/c9Lv0
  e4xyZDzRT/mlq1lDSVBtz7VBl2ocOryvO9KlPGUByTxgx78QaWNeus+/MJoojjug2yRbw4CB
  ccedZvBYwAo5AU3Y0FNKjlsv/daTwaCDnZa7ZB9rJDbMGckQGJRW0oArNO2qDsyOJELEvGYw
  oZnHvRrf/78xom8XLDNgbRjMMOqY0YNXLHfkwX8OCfUWLsLRrzbc3/xrlVWQETejJYCeCf9V
  yiNwoDlO6tshToDI17vUzKaUW9+VlGp0w2aHEXWQDaIMzalXbtWKehgYs7AhfKGBATiwC7A9
  D0IrNigA2G8e+CRoYCEVy5ABymFRAesrM35dX4R/Mig0OWXYrUgogzGhiZ51rybLLILzGNQm
  3l3RAeddeH6Qw7M4XBROfMIcC0KrDLArkpIOoisLCS7Zw7wBxmyBwuE1lCGYfCJpzOnTVMsJ
  jVwphaHNacmRl2rKQTbNlzLRrRAE5UBweNImygIrY0OFVLEF8aRnWDNwp8+iGp0TCCHNkry4
  8QSIIx5zuavB/jLJHqfE5Wc5wbGR5WryNlERF8W393bspHAZhwRjzRXpjzB52q8xCiNQW4Oi
  mjdtwRUBtfOSkrMRQcU686GjtMVKESPEHEvjMzhKmqgKqh74OJlHaAj0VBTWMTUXN8fRriMZ
  BdHpEcSqKC4r1uhXPwzAKiVwQZSJxJJ9J6qs6irhtSWiHZ2h5WwH+w9cwWBaoJK83EwACZSK
  oJOYjUxkmA2KjQ07IhP9D+kPlBP+OSP5juL7kqP8YqYVJJ8KyZqYOZkGdBxIsHUY2wvi8Vc0
  4nN8F2qHaJ4SzW+VRTl1QT2bwsIO5wGt0BHEUQLMyDJrCVIgLKNwWVFiTkBxEMVRYrEGPKoG
  Nbl8wpNC8gQchpbFZdVqy5icUFzGnryhwrENUD1mHMfDJ0zrxX515SQCO+xT9Nn2+EKYWKXO
  b4EcVBJVwMnsM5zo88glRS8IpdQaAkIvg7SmVXY4FjcFJQFwkEB90Vsijfp5rjkClooro2po
  EooSmwbn1GRpEZdAZCdZmZZCzpQgB7jSmvTgJprTJTkJ9UP01/8591/WJv03az9ekiLzQMqk
  CE5kkN9wKl6Sn+LN9bfqZ7X209wDLeodfzhU7ZLrfqr68DPvM1pXP6h6P+XoZnP9PrXu4lP9
  y2n6a32sJHullxn2+Sq7nCvRteIrntkBO9hWHre/7YwFzVFSrIVNwT5HmK3kXb2vA9fazltP
  8V8zGyZzMomVhxfI1FDQpTXi7m/NHSLqRNsA8kYNOjWNUzL9zF2s3bESzi2QgHswAjEhwSnD
  LK8DUoUIA6LgMpItUVDAbNqgEubFpH1tRoBhosGDUDv3i6mC5VEYm7xJ9zUhyaKDHJkVBFvh
  441FNaaL5PsD8WkXGcTFKHSSZm+EFWuiZKxpLdb4uax0rXLosa+vZZjQT2YktZFpuUU58feU
  G3lqDiiYgN6dBQGIAtkOO7+TfCBsUd6soUJeRU1Pe3iNW8aXtbUwQz/M7WsxSJNZmJ1abJOG
  A5+7rsrDX01CtOlFH+eP8ZwwlTvqjFNy4HaZkiaHaFBtTShIn3Da/JVDE9fk6WOF8aV286zR
  SIUAVHHTClfFCnhTPFLut7w/58j1na2coLFa29VkMgx3eqt898r31s/U9GCsIXJe9z1t77Of
  azKCP8Y7/t5MCgr3vfxuuDNJ5+Xfq9Q7pNojY8AKM0+CuA3hu3+m2DHb31UubYf9hjPkPjDI
  PAk3r/ptn/cTfX3JRql7PVfYHuSV15t7nyvc/zYRPtcGCrVXqppKp4VP2XvrF7JfKHzbaz3b
  Omeqv7Ud5Kgr4hchrsu5E+rQH+5cQ6ENEmiooRgUUJyrU6MMFGZDzpGWZnDzXZUz1BybbqZX
  q0/X62d2h88rsykxOZJrVp8Rpns2cjkW5tLXso9wumvBo4L1i1irLGoZYwGRp5UbVjJHx1te
  fWJmTwFB89UKFUq/iwMmggF1rLNnNgJtJpXlYhMwSLQ2uxKmHhRSdytnMAjFb7231vBdShrh
  6CCUZKNSTRR9Y048Ms8+ZYxKeGWalZoW8Is4kGOCL2IyIzK3MsySZGWJ9s0h2EsiOBs0kAVa
  L9aUZtpcIGBqy8r09DksuMVcEWZT3EucBBuK9dqmxlO35/AuwFLF4KnufEXugAXG3o5ZCXsj
  F4K6F4CdB8Kche6Zv1KpgsyUimcqLimcyoJvWGNNhsh4Ar5qopBctKKjsMyILJsQS1rh1qkI
  FsQKKAGpetDD59jWp7ClUH8XsAWSW4qooxXtR05Rxd5mBViwOCqkpYqrgg42YykFwOnRTpp8
  inKIRUJkfCR5e8vERpbPD5hMvlJdJzJi0+FYqmmDGn8q/gBm9Pq3+lH77OfY3v6MzuNDsbzA
  72MwuNDsbzA72MwuNDsbzA72MwuNDsbzA72MwuNDs0PtrBLLfXKXJrk1Jy3GciMuRnYoSu5k
  V3lh7yf9xp9v7bzZmKp3GNJVK5A1uj/Znn2YzExLv//F3UWAt89qLQbcMYsfOgPlMZMGnlDT
  LNyt5xxkfaq22QpQxEzYjInulu8vljhr4zqhy4i64BvYeMyOQ1sW08pyNUvbaiZpp3NiB0Vq
  AEsoUP6ORgicoXpLopRkxWZxILPr5D3rHN8j7VIgh2McZcrwBwsY+LsVY/NOeGI6hbVJnlka
  GI4mdzARLph5RkNBEcH3xJwMNVMn0Doj0rAS6aFecjRBU+/nKBj5Vek8oYn8I/YS1FcKteo3
  ihx86+bYRz+CctqLGyXoABvYqRC1nFtWzGRMOB6VZgsTzJKICKYRlPSO1oS6wRIYyzFrAgxR
  DZY5IaSj1+KQg7RhW5V/SDK6/VZXI2N=
}

[Dinkydau]

I copied your post and saved it to a text file and named it "test.par", then navigated to it with UF 4.02 version.

That took only 1 hr 45 min to render at 640x480 on my Alienware Athlon64, my fastest system. I think it was using only 1 core, looking at task manager display.

I was amazed, then I saw the iteration count was less than 3000.

You will never see a mini-brot at that depth with so low iterations.

You need at least a million, and don't keep zooming into the "center" of the image; go off to one of the large standing waves where 4 points intersect.
From my experience, you are a long way to go.

Yes the preview takes less time to render. But as I said, only a few thousands of iterations were used. I must say I have no idea of how many iterations I will need to render the mandelbrot set below, as I have no idea how long it will take until I get there, but I do think 1 million will not be necessary.

"don't keep zooming into the "center" of the image"
That was my plan actually, I wanted a mandelbrot set as deep as possible. (not in theory but in reality of course )

`From my experience, you are a long way to go.`
Certainly, I´ve been zooming like forever. I started zooming into one of the centers at E640.

[David Makin]

Yes, this one for example:

For that UPR on my P4HT with min threads set to 2 and max to 32 CPU use on either "core" never drops below 97% either within UF or when rendering to disk - I'll try it on my Core2Duo laptop next time I fire it up

[David Makin]

Well I must say that only a few thousands of iterations are used for that.

Here are some parameters:

E1103 { <snip>

Of course this does not look like somthing special.
My goal is to zoom in deep enough to end up seeing a mandelbrot set.

Boy do I need a new computer - I tried pasting that one (on this P4HT) and was promptly given an ETA of >30 hours

[Dinkydau]

I have no idea of what you could be doing wrong. How many iterations are used? How many GHz are you using?

That took only 1 hr 45 min to render at 640x480 on my Alienware Athlon64, my fastest system. I think it was using only 1 core, looking at task manager display.

That is actually amazingly fast. I need little over 1 hour for the preview using 3 cores.

[fractalwizz]

Dinkydau, boy have I been looking for someone as enthused in deep zooms as i am.

i don't think we've met before.

i make deep zoom animations and publish them on utube

yes, deep zooms take forever - i have a separate computer that is rendering a deep zoom into a special area at a magnification of e+243
so far, 6  months have passed

[Dinkydau]

I am one of your subscribers, my name is KesavaAVP.

[David Makin]

I have no idea of what you could be doing wrong. How many iterations are used? How many GHz are you using?

I am not doing anything wrong It's just that my PC is only a 3GHz P4HT that overheats so probably only actually runs at about 2GHz under load so the ETA of 30hrs could well just be "correct".
I tried the E1103 UPR on my 2GHz core2Duo and the ETA on there was just over 9hrs which seems consistent.

I also checked the Herman zoom UPR on my core2Duo with much the same result as on my P4HT - both cores were used to over 95% inside UF and when disk rendering without AA - with AA on "normal" the behaviour was the same except at a couple of points CPU/FPU use dropped briefly to around 70%.