 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Just quick and random, but I like it so I thought I should share. I'll
continue tweaking and add some randomness to the color also.
Post a reply to this message
Attachments:
Download 'rsweep.png' (262 KB)
Preview of image 'rsweep.png'
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 24-6-2012 12:42, Eero Ahonen wrote:
> Just quick and random, but I like it so I thought I should share. I'll
> continue tweaking and add some randomness to the color also.
I like that! Especially the "roughly glued" joints which look extremely
realistic.
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> On 24-6-2012 12:42, Eero Ahonen wrote:
>> Just quick and random, but I like it so I thought I should share. I'll
>> continue tweaking and add some randomness to the color also.
>
> I like that! Especially the "roughly glued" joints which look extremely
> realistic.
>
> Thomas
>
I think that those "roughly glued joints" are realy sphere_sweep
artefacts...
But, in this case, those do look good.
Alain
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Alain wrote:
>
> I think that those "roughly glued joints" are realy sphere_sweep
> artefacts...
> But, in this case, those do look good.
>
>
> Alain
>
Yes, indeed they are just that.
I made the object a bit transparent, added some lights to make some
colorization and changed to 3.7 to get more threads on quad-core system.
...the new version has now been rendering for 1d, 5h and 50 minutes to
reach 32%. I guess it came out a bit too heavy and won't pay off the
extra load, but now I really DO want to see what will come out of this,
so I'll let it run.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Eero Ahonen wrote:
>
> ...the new version has now been rendering for 1d, 5h and 50 minutes to
> reach 32%. I guess it came out a bit too heavy and won't pay off the
> extra load, but now I really DO want to see what will come out of this,
> so I'll let it run.
The original picture has 20 joints. Here's one with 10 joints from 3.6,
apparently it renders in under 9 minutes, even though 3.7 is taking it's
share of the CPU resources at the same time. Also 3.6 is now going in
33% after just 18 minutes. Even though the different versions seem to
render the scenes differently (3.6 in lines, 3.7 in pieces) and possibly
to calculate the percentages a bit differently (or hitting more load at
one time to be precise) due to this, it seems that this scene is way
harder for 3.7 to render. OTOH the visible part of the image seems to
say that 3.7 makes a lot better quality (mostly concerning the
sphere_sweep artefacts).
Post a reply to this message
Attachments:
Download 'rsweep-lite.png' (68 KB)
Preview of image 'rsweep-lite.png'
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Am 25.06.2012 21:57, schrieb Eero Ahonen:
> The original picture has 20 joints. Here's one with 10 joints from 3.6,
> apparently it renders in under 9 minutes, even though 3.7 is taking it's
> share of the CPU resources at the same time. Also 3.6 is now going in
> 33% after just 18 minutes. Even though the different versions seem to
> render the scenes differently (3.6 in lines, 3.7 in pieces) and possibly
> to calculate the percentages a bit differently (or hitting more load at
> one time to be precise) due to this, it seems that this scene is way
> harder for 3.7 to render. OTOH the visible part of the image seems to
> say that 3.7 makes a lot better quality (mostly concerning the
> sphere_sweep artefacts).
Yes, fixing a good deal of the sphere_sweep artifacts came at a price.
We found the problem to be rounding issues in some optimization code; as
it would have been difficult to fix, we decided to turn it off for now.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
clipka wrote:
>
> Yes, fixing a good deal of the sphere_sweep artifacts came at a price.
> We found the problem to be rounding issues in some optimization code; as
> it would have been difficult to fix, we decided to turn it off for now.
Yes, that indeed explains it all. I could see a possibility to choose
the behavior with some option of the object to be nice, but if there is
none, I'm more happy with more precise results, even if the cost is this
huge difference in rendering time.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
