Hi,

As I have access to a machine with 8GB memory and 8 cores and not enough time to
create any new POV images for over a year now :-( I have been trying to
re-render one of my old images at very high res.. I have been trying to
generate radiosity stats (save them to a file) at 2625x1750 but the image
contains black patches shown below.

Any ideas what could be causing this.

My current theories are:

max_trace_level (set to 15)

Something to do with the radiosity settings:

    pretrace_start 0.08
    pretrace_end   0.01
    count 500
    nearest_count 20
    error_bound 0.08
    recursion_limit 4
    low_error_factor .9
    gray_threshold 0
    minimum_reuse 0.015
    brightness 0.8
    adc_bailout 0.01/2
    save_file "museum.rad"

Something to do with the amount of memory being used:

1.5GB (VM size 1.93GB)

Any ideas... Also if anyone knows if the radiosity data will still be OK or if I
need to stop the render and restart once the issue is fixed would be good (this
has been running for 11 days so far).

Sean

Post a reply to this message

Attachments:
Download 'temp.jpg' (86 KB)

Preview of image 'temp.jpg'

I forgot to add that the machine is only running XP 32 bit and 32 bit POV which
is why I wondered if approaching the 2GB memory limit may be causing the
problem.

Sean

Post a reply to this message

s.day wrote:
> Hi,
> 
> As I have access to a machine with 8GB memory and 8 cores and not enough time to
> create any new POV images for over a year now :-( I have been trying to
> re-render one of my old images at very high res.. I have been trying to
> generate radiosity stats (save them to a file) at 2625x1750 but the image
> contains black patches shown below.
> 
> Any ideas what could be causing this.
> 
> My current theories are:
> 
> max_trace_level (set to 15)
> 
> Something to do with the radiosity settings:
> 
>     pretrace_start 0.08
>     pretrace_end   0.01
>     count 500
>     nearest_count 20
>     error_bound 0.08
>     recursion_limit 4
>     low_error_factor .9
>     gray_threshold 0
>     minimum_reuse 0.015
>     brightness 0.8
>     adc_bailout 0.01/2
>     save_file "museum.rad"
> 
> Something to do with the amount of memory being used:
> 
> 1.5GB (VM size 1.93GB)
> 
> Any ideas... Also if anyone knows if the radiosity data will still be OK or if I
> need to stop the render and restart once the issue is fixed would be good (this
> has been running for 11 days so far).
> 
> Sean
> 
> 
> 
> ------------------------------------------------------------------------
> 

I can eliminate two of the options for you.

With maxTraceLevel set to 15, you likely would not see blotches like 
this.  You won't get 15 levels of reflection in this scene and even if 
you did, you shouldn't get this blotching.

I can't imagine that insufficient memory would cause the image to appear 
blotchy.  It may slow down the render by using virtual memory instead of 
physical memory and it may fail altogether but I can't imagine a 
scenario in which memory would cause differences in the image without 
crashing.


I can't comment on the radiosity settings - I'll let other people in 
this group address that.

David Buck

Post a reply to this message

"s.day" <s.d### [at] uelacuk> schreef in bericht 
news:web.47d915fa72169bca7eb3d110@news.povray.org...
>
> Something to do with the radiosity settings:
>
>    pretrace_start 0.08
>    pretrace_end   0.01
>    count 500
>    nearest_count 20
>    error_bound 0.08
>    recursion_limit 4
>    low_error_factor .9
>    gray_threshold 0
>    minimum_reuse 0.015
>    brightness 0.8
>    adc_bailout 0.01/2
>    save_file "museum.rad"
>

Your combination count / error_bound might be faulty. You could try (for a 
first pass on a small image without aa):
error_bound 0.2
count 1000 (or 5000)
recursion_limit 3 (or 4)
pretrace_end 0.01 (or even lower)

and for a second pass go for:
error_bound 1
count 100
recursion_limit 1
pretrace_start 1
pretrace_end 1

It is mainly your very low error_bound that makes the render to take so long 
I believe and the count for that is far too low. From what I have learned in 
these ng's, you hardly ever need such extremely low error_bound values.

Thomas

Post a reply to this message

> As I have access to a machine with 8GB memory and 8 cores and not enough 
> time to
> create any new POV images for over a year now :-( I have been trying to
> re-render one of my old images at very high res.. I have been trying to
> generate radiosity stats (save them to a file) at 2625x1750 but the image
> contains black patches shown below.

Those look like some problems I was having with isosurfaces a while back, 
try increasing accuracy and max_gradient if you are using isosurfaces.

Post a reply to this message

> I forgot to add that the machine is only running XP 32 bit and 32 bit POV which
> is why I wondered if approaching the 2GB memory limit may be causing the
> problem.

It definitely won't cause the problems on the image; but what a waste! 
8GB of RAM and you can only use 2GB of them? Get a 64-bit system :)

Post a reply to this message

>     pretrace_start 0.08
>     pretrace_end   0.01
>     count 500
>     nearest_count 20
>     error_bound 0.08
>     recursion_limit 4
>     low_error_factor .9
>     gray_threshold 0
>     minimum_reuse 0.015
>     brightness 0.8
>     adc_bailout 0.01/2
>     save_file "museum.rad"

I find that if the surfaces in my scene have a bit of texture, I can get by with
much lower count settings, since a little blotchiness blends in with the texture
anyway.  That's a very low error_bound setting!!

Is there any media in the scene?  I had a scene a while ago in which I had
allowed interaction between media and radiosity, and some strange things
happened.

Post a reply to this message

"s.day" <s.d### [at] uelacuk> wrote:
> Hi,
>
> As I have access to a machine with 8GB memory and 8 cores and not enough time to
> create any new POV images for over a year now :-( I have been trying to
> re-render one of my old images at very high res.. I have been trying to
> generate radiosity stats (save them to a file) at 2625x1750 but the image
> contains black patches shown below.

I rarely use radiosity, and my first guess is that the blotches must relate to
your radiosity settings.

However, I have encountered black blotches in a scene in which I used a mask,
let me see, I used an image_map in which a photograph of a butterfly had been
modified so that the butterfly alone remained, and all else became a black
background. I set the black to be transparent in POV-Ray, so that all one would
see was the butterfly itself. Then three copies were placed near each other, and
moved along a spline curve in an animation.

Sometimes, then, one of the three butterflies would have a black blotch, shaped
like the rectangle bounding the image_map.

The fix was to change max_trace_level from 3 to 8.

Also, may I say that, if there are not multiple reflecting objects in a scene,
one can set max_trace_level to 100 or whatever without any speed penalty. So it
should be easy to test if max_trace_level is the culprit.

But your blotch problem looks like it's got to be radiosity to me.

Post a reply to this message

> Also, may I say that, if there are not multiple reflecting objects in a scene,
> one can set max_trace_level to 100 or whatever without any speed penalty. So it
> should be easy to test if max_trace_level is the culprit.
> 

max_trace_level affects radiosity.

Post a reply to this message

Wow! thanks for all the replies. It seems the general consensus is that this is
likely to be caused by the radiosity settings, I think I will try adjusting
these settings first (higher error_bound as suggested by Thomas and Kirk).

The scene does not have any isosurfaces (yet that will be after I collect the
radiosity data) otherwise this would have been a prime candidate as I have seen
these kind of problems with isosurfaces before.

The textures are very basic for this trace, no image maps etc..

No media in the scene to be causing any issues.

I will run some tests and let you know if it fixed the issue (don't hold your
breath even at low resolutions this is a slow render).

Thanks to everyone for the suggestions.

Sean

Post a reply to this message