 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Scene contains 6 frame level objects; 0 infinite.
Statistics for triforce.pov, Resolution 640 x 480
----------------------------------------------------------------------------
Pixels: 312000 Samples: 312000 Smpls/Pxl: 1.00
Rays: 20049467 Saved: 579731 Max Level: 10/10
----------------------------------------------------------------------------
Ray->Shape Intersection Tests Succeeded Percentage
----------------------------------------------------------------------------
Box 7642950539 5029906249 65.81
CSG Intersection 1606762390 586731732 36.52
Bounding Box 4711634819 3005395253 63.79
Light Buffer 8364355053 5795374111 69.29
Vista Buffer 1202459 901250 74.95
----------------------------------------------------------------------------
Calls to Noise: 0 Calls to DNoise: 10
----------------------------------------------------------------------------
Media Intervals: 24236446 Media Samples: 690560487 (28.49)
Shadow Ray Tests: 1197618743 Succeeded: 575085574
Reflected Rays: 10840576 Total Internal: 5663492
Refracted Rays: 6429680
Transmitted Rays: 2408330
Number of photons shot: 58881
Surface photons stored: 563400
Media photons stored: 25211616
Priority queue insert: 159359134552
Priority queue remove: 93253179332
Gather function called: 3608877
----------------------------------------------------------------------------
Smallest Alloc: 25 bytes Largest: 327704
Peak memory used: 516015973 bytes
----------------------------------------------------------------------------
Time For Parse: 0 hours 0 minutes 1.0 seconds (1 seconds)
Time For Photon: 0 hours 2 minutes 28.0 seconds (148 seconds)
Time For Trace: 88 hours 45 minutes 32.0 seconds (319526 seconds)
Total Time: 88 hours 47 minutes 55.0 seconds (319675 seconds)
----------------------------------------------------------------------------
CPU time used: kernel 208.67 seconds, user 272669.25 seconds, total
272877.92 seconds
Render averaged 1.13 PPS over 307200 pixels
An i7 ought to do this in 3.5 hours, not 3.5 days. I'm not even using
antialiasing or very high media/photon settings!
--
Tim Cook
http://empyrean.freesitespace.net
Post a reply to this message
Attachments:
Download 'triforce.png' (298 KB)
Preview of image 'triforce.png'
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Tim Cook <z99### [at] gmail com> wrote:
> An i7 ought to do this in 3.5 hours, not 3.5 days. I'm not even using
> antialiasing or very high media/photon settings!
weird indeed!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Scene contains 6 frame level objects; 0 infinite.
>
> Statistics for triforce.pov, Resolution 640 x 480
> ----------------------------------------------------------------------------
>
> Pixels: 312000 Samples: 312000 Smpls/Pxl: 1.00
> Rays: 20049467 Saved: 579731 Max Level: 10/10
> ----------------------------------------------------------------------------
>
> Ray->Shape Intersection Tests Succeeded Percentage
> ----------------------------------------------------------------------------
>
> Box 7642950539 5029906249 65.81
> CSG Intersection 1606762390 586731732 36.52
> Bounding Box 4711634819 3005395253 63.79
> Light Buffer 8364355053 5795374111 69.29
> Vista Buffer 1202459 901250 74.95
> ----------------------------------------------------------------------------
>
> Calls to Noise: 0 Calls to DNoise: 10
> ----------------------------------------------------------------------------
>
> Media Intervals: 24236446 Media Samples: 690560487 (28.49)
> Shadow Ray Tests: 1197618743 Succeeded: 575085574
> Reflected Rays: 10840576 Total Internal: 5663492
> Refracted Rays: 6429680
> Transmitted Rays: 2408330
> Number of photons shot: 58881
> Surface photons stored: 563400
> Media photons stored: 25211616
> Priority queue insert: 159359134552
> Priority queue remove: 93253179332
> Gather function called: 3608877
> ----------------------------------------------------------------------------
>
> Smallest Alloc: 25 bytes Largest: 327704
> Peak memory used: 516015973 bytes
> ----------------------------------------------------------------------------
>
> Time For Parse: 0 hours 0 minutes 1.0 seconds (1 seconds)
> Time For Photon: 0 hours 2 minutes 28.0 seconds (148 seconds)
> Time For Trace: 88 hours 45 minutes 32.0 seconds (319526 seconds)
> Total Time: 88 hours 47 minutes 55.0 seconds (319675 seconds)
> ----------------------------------------------------------------------------
>
> CPU time used: kernel 208.67 seconds, user 272669.25 seconds, total
> 272877.92 seconds
> Render averaged 1.13 PPS over 307200 pixels
>
> An i7 ought to do this in 3.5 hours, not 3.5 days. I'm not even using
> antialiasing or very high media/photon settings!
>
> --
> Tim Cook
> http://empyrean.freesitespace.net
May be worth looking at your media. Do you have intervals set to
anything larger than 1? Bring back to 1 and adjust samples as needed.
intervals 10 samples 5 is much slower than samples 100 intervals 1.
How many dispersion_samples?
It don't looks like a memory isue, unless you have a ridiculously low
amount in your computer, as you use less than 500 Mb.
Alain
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Alain wrote:
> May be worth looking at your media. Do you have intervals set to
> anything larger than 1? Bring back to 1 and adjust samples as needed.
> intervals 10 samples 5 is much slower than samples 100 intervals 1.
>
> How many dispersion_samples?
>
> It don't looks like a memory isue, unless you have a ridiculously low
> amount in your computer, as you use less than 500 Mb.
Yeah the media's one thing I need to fix and save back to the library...
Really, the thing that's making it slow is the refractive objects. I
could slap no_image on them and get the nice media/photon effects in a
few minutes. Turn off the media and photons and have the tetrahedra on
and it takes a few days.
Is there any way to speed that up? How do commercial renderers compare
speedwise for refractives? What do they do differently?
--
Tim Cook
http://empyrean.freesitespace.net
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
4b67af7b$1@news.povray.org...
> Really, the thing that's making it slow is the refractive objects. I
> could slap no_image on them and get the nice media/photon effects in a few
> minutes. Turn off the media and photons and have the tetrahedra on and it
> takes a few days.
>
> Is there any way to speed that up? How do commercial renderers compare
> speedwise for refractives? What do they do differently?
>
> --
The keyword that comes in mind is adc_bailout... did you try setting it
higher?
Marc
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Alain wrote:
>> May be worth looking at your media. Do you have intervals set to
>> anything larger than 1? Bring back to 1 and adjust samples as needed.
>> intervals 10 samples 5 is much slower than samples 100 intervals 1.
>>
>> How many dispersion_samples?
>>
>> It don't looks like a memory isue, unless you have a ridiculously low
>> amount in your computer, as you use less than 500 Mb.
>
> Yeah the media's one thing I need to fix and save back to the library...
>
> Really, the thing that's making it slow is the refractive objects. I
> could slap no_image on them and get the nice media/photon effects in a
> few minutes. Turn off the media and photons and have the tetrahedra on
> and it takes a few days.
>
> Is there any way to speed that up? How do commercial renderers compare
> speedwise for refractives? What do they do differently?
>
> --
> Tim Cook
> http://empyrean.freesitespace.net
Then, what is your max_trace_level and adc_bailout?
Often, adc_bailout 0.01 is good enough.
Your main problem looks to be many total internal reflections. With
dispersion, you can get total internal reflection for blue, but not for
red. This leads to partial doubling of the rays, blue is reflected but
cyan to red goes through with some added reflection, and the spawning of
still more dispersion rays.
What's the value for dispersion_sample? A somewhat smaller value could
be enough.
One other thing that you can try, rotate the tetrahedra a small bit. It
could be all that's needed to prevent extremely long internal paths.
Some comercials products seems to have quicker bailouts. Other may skip
on the dispersion, or fake it even more than POV-Ray does, or only apply
it for the first or last refractive surface encountered before hiting an
opaque surface. Some don't even support dispersion.
Alain
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Alain wrote:
> Then, what is your max_trace_level and adc_bailout?
> Often, adc_bailout 0.01 is good enough.
Was at default, trying small render with 4 and 0.02 now...
> What's the value for dispersion_sample? A somewhat smaller value could
> be enough.
Was 40, trying 20.
Reducing resolution to 400x300...
Fiddling a few other photon settings...
--
Tim Cook
http://empyrean.freesitespace.net
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Oh wow. Just a bit faster (just under a minute)...needs work to get
closer to previous result, but have a starting point.
--
Tim Cook
http://empyrean.freesitespace.net
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Tim Cook <z99### [at] gmailcom> wrote:
> Oh wow. Just a bit faster (just under a minute)...needs work to get
> closer to previous result, but have a starting point.
from 88 hours to 1 minute is a long way. :)
radiosity and photon blocks also have their own adc_bailout, keep that in mind.
BTW, you surely is really making use of all those threads and cores with 3.7,
right?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
nemesis wrote:
> BTW, you surely is really making use of all those threads and cores with 3.7,
> right?
I'm still crawling along with 3.5, because I use Moray and it doesn't
work with newer. Here's hoping the team decides to fix it first after
final 3.7 release before waiting for POV 4.
--
Tim Cook
http://empyrean.freesitespace.net
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
