Page 3 of 4
Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 7:29 am
by Daniel79
radiance wrote:the new GTX260 with 2GB is cheap and can do a pretty neat render at large resolution.
Radiance
Maybe you spoke about GTX460??...ther's GTX 260 have 2gb??
Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 7:33 am
by SurfingAlien
there was a 1.8Gb GTX260 but Radiance is talking about the new GTX460 2Gb for sure
Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 8:08 am
by thedreamers
SurfingAlien wrote:there was a 1.8Gb GTX260 but Radiance is talking about the new GTX460 2Gb for sure
i think that card will work great with octane...a lot of memory (more than 480) and still 336 cuda cores on it

Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 9:31 am
by radiance
Hey, sorry i made a typo, GTX460 i meant

it's because the GTX260 is a word i've typed a lot, it's in my fingers
Radiance
Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 11:56 am
by Chris_TC
n1k wrote:Those are smoothing artifacts caused by low poly geometry. When such things occur one must add more polygons to his/hers models and problem will be gone.
radiance wrote:because those objects are quite low poly.
When adding polygons the lines only become thinner. Also, having to add polygons when dealing with an Export-Import workflow is sub-optimal.
In one of my car renders I had already maxed out the amount of polygons to what my system could handle, and yet the car seats (at
100,000 triangles each) still had black outlines which I had to manually fix in Photoshop.
Needless to say that rendering the same geometry with a quarter of the detail (25,000 triangles) doesn't show black edges in either Mental Ray or 3Delight.
Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 12:50 pm
by n1k
Well, radiance said that geometry is going to have smaller memory footprint in near future octane builds. Also precomputed Catmull–Clark subdivision surface is announced which is going to reduce exporting time for subd models greatly.
Cheers,
n1k
Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 1:15 pm
by Chris_TC
That's nice and all, but this is still a huge issue that needs to be addressed.
Not just because it adds unnecessary exporting time, memory consumption and rendering time. But also because these ugly artifacts are difficult to predict and foresee.
The car seats I had mentioned looked very nice and smooth in my viewport (even with less subdivisions), and yet Octane added a black outline.

Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 1:38 pm
by radiance
Chris_TC wrote:That's nice and all, but this is still a huge issue that needs to be addressed.
Not just because it adds unnecessary exporting time, memory consumption and rendering time. But also because these ugly artifacts are difficult to predict and foresee.
The car seats I had mentioned looked very nice and smooth in my viewport (even with less subdivisions), and yet Octane added a black outline.

Hi,
those are normal smoothing artifacts, due to the fact that the edges on the chess piece haven't got smooth groups or edge-split geometry in octane,
and interpolating/smoothing 2 faces with a high angle in between gives an artifact.
the render in mental ray probably uses smoothgroups or uses an angle based splitting.
Radiance
Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 1:57 pm
by face
Angle based.
Import the obj in Softimage and change the discontinuity angle to 30° under geometry approximation.
Then reload it in Octane and you will see, the black artefacts will gone on some parts.
face
Re: 6000x6000px render test with new 2.3 res limit (8192x8192)
Posted: Fri Jul 16, 2010 2:39 pm
by Chris
On some parts, but not all?
That is a problem.