Poll: Which engine are you using?

A forum especially for contests and Hall of Fame.
Post Reply

Which engine do you use most of the time?

Adaptive (BSD)
34
10%
Unbiased (TR1)
51
15%
Unbiased (TR2)
24
7%
Presto (AO)
102
30%
Presto (MC)
124
36%
Adaptive (AMC)
9
3%
 
Total votes: 344
User avatar
jc4d
Thea Beta Tester
Thea Beta Tester
Posts: 1202
Joined: Tue May 12, 2009 3:13 pm
Location: Tornio, Finland
Has thanked: 20 times
Been thanked: 2 times

Tue Feb 09, 2016 6:25 am

Mohammad Nael wrote:
Hieru wrote:
JQL wrote:EDIT - It would also be nice if we could define materials or areas of a model where to filter more than others as filters will take away accuracy
+1.

Maybe this could be done in the same way you can set the Tracing Depth for individual materials?
No, because Thea has a professional Extension Depth that you may lower the main depth and keep Extra Depth for transparent/SS/clip high @ 15 by default.
The question is that, Can we give some materials more sampling importance, I sometimes do a region for some places!!!!
I agree, having some extra sampling control would give so much flexibility when we want to get rid of persistent noise.
I would like to have control per object and/or per material.
User avatar
Hieru
Licensed User v1.5
Licensed User v1.5
Posts: 1950
Joined: Fri Jun 26, 2009 11:47 am
Location: UK
Has thanked: 54 times
Been thanked: 22 times
Contact:

Tue Feb 09, 2016 11:44 am

Although, as others have pointed out, it's quite often areas with indirect light where you get the most problems.

So, you might have noise in the shadow falling on 10% of a wall. Having extra passes per material would apply to the whole wall rather than just the small shadow area.

What we kind of need is a brush or polygonal lasso that works a bit like a live region render.
User avatar
Pixero
Thea Beta Tester
Thea Beta Tester
Posts: 1721
Joined: Tue May 12, 2009 9:20 am
Location: Halmstad, Sweden
Has thanked: 25 times
Been thanked: 127 times
Contact:

Tue Feb 09, 2016 1:17 pm

Hieru wrote: What we kind of need is a brush or polygonal lasso that works a bit like a live region render.
+1 for that...or something even better that Giannis can come up with.
User avatar
JQL
Thea Beta Tester
Thea Beta Tester
Posts: 4980
Joined: Sun Jun 23, 2013 6:34 pm
Location: Lisboa
Has thanked: 13 times
Been thanked: 71 times
Contact:

Tue Feb 09, 2016 7:37 pm

The problem with lasso is that it works for stills but not animations. Something that could discover an indirect light difficult area in the model would be better. Or something like a rotoscope selection like we have in movie compositors that is aware of a selection in 3d space.
www.casca.pt
Visit us on facebook!

Win 10 64bit
Sketchup 2017 64bit
Intel i7 5820K 3.3Ghz 32Gb Ram
Nvidia Titan X 12Gb
User avatar
jc4d
Thea Beta Tester
Thea Beta Tester
Posts: 1202
Joined: Tue May 12, 2009 3:13 pm
Location: Tornio, Finland
Has thanked: 20 times
Been thanked: 2 times

Wed Feb 10, 2016 7:57 am

JQL wrote:The problem with lasso is that it works for stills but not animations. Something that could discover an indirect light difficult area in the model would be better. Or something like a rotoscope selection like we have in movie compositors that is aware of a selection in 3d space.
That´s one reason because I think that having it per material and/or object could be better workflow.
If a corner is the problematic zone, cut the model geometry, duplicate the same material and assign the new one with higher sampling to the geometry that you just cut and hit render (for example)
User avatar
giannis
Thea Developer
Thea Developer
Posts: 7988
Joined: Sun Jul 06, 2008 8:25 pm
Location: Athens, Greece
Has thanked: 11 times
Been thanked: 6 times

Wed Feb 10, 2016 8:43 am

I follow the discussion; having a kind of "adaptive" sampling for areas is not that easy, or putting more effort to a certain material. First of all, it is usually hard to know a-priori. So it requires manual intervention.

One way could be an automated adaptive way to put more effort until some criterion is met. In general, it is tricky to have really consistent way of doing that. What I also found in some experiments, is that while you may design a kind of local adaptiveness, this will reduce also performance (especially from GPUs).

Changing back the discussion a little bit: it is very interesting for me to see here how we have adapted (as a userbase) from a CPU renderer to a GPU one. So far, it seems that most of the people are using Presto. Of course, that's a chicken and egg situation because that was also the engine that got more effort in the past 2 years.

Greetings
Science may limit knowledge but should never limit imagination - Bertrand Russell
User avatar
Mohammad Nael
Posts: 1257
Joined: Sun Feb 05, 2012 10:22 am
Location: Syria /Aleppo
Has thanked: 0
Been thanked: 1 time
Contact:

Wed Feb 10, 2016 10:45 am

jc4d wrote:
JQL wrote:The problem with lasso is that it works for stills but not animations. Something that could discover an indirect light difficult area in the model would be better. Or something like a rotoscope selection like we have in movie compositors that is aware of a selection in 3d space.
That´s one reason because I think that having it per material and/or object could be better workflow.
If a corner is the problematic zone, cut the model geometry, duplicate the same material and assign the new one with higher sampling to the geometry that you just cut and hit render (for example)
I believe if this is achievable, Giannis would have implemented it since the first day of Presto.
But for me I am really in short of a biased method supporting GPU+CPU, but how, I don't know. My customers does not understand unbiased or scanline, they just want a noise free and detailed image.
Regards
User avatar
JQL
Thea Beta Tester
Thea Beta Tester
Posts: 4980
Joined: Sun Jun 23, 2013 6:34 pm
Location: Lisboa
Has thanked: 13 times
Been thanked: 71 times
Contact:

Wed Feb 10, 2016 10:59 am

Mohammad Nael wrote:I believe if this is achievable, Giannis would have implemented it since the first day of Presto.
But for me I am really in short of a biased method supporting GPU+CPU, but how, I don't know. My customers does not understand unbiased or scanline, they just want a noise free and detailed image.
Regards
Displacement isn't there since Presto's day one of but it was achievable.

We all want our images to be noise free and detailed as fast as we can, and a GPU+CPU BSD engine would be nice to see, but in wich way would a noise reduction algorithm based on area be against a noise free image?
www.casca.pt
Visit us on facebook!

Win 10 64bit
Sketchup 2017 64bit
Intel i7 5820K 3.3Ghz 32Gb Ram
Nvidia Titan X 12Gb
User avatar
jc4d
Thea Beta Tester
Thea Beta Tester
Posts: 1202
Joined: Tue May 12, 2009 3:13 pm
Location: Tornio, Finland
Has thanked: 20 times
Been thanked: 2 times

Wed Feb 10, 2016 10:59 am

Mohammad Nael wrote: I believe if this is achievable, Giannis would have implemented it since the first day of Presto.
There is another company that has this kind of extra sampling per material, but maybe Giannis have found so far that there is not so much speed gaining using this method??
Mohammad Nael wrote: But for me I am really in short of a biased method supporting GPU+CPU, but how, I don't know. My customers does not understand unbiased or scanline, they just want a noise free and detailed image.
Regards
I would love to see that as well, honestly there is a "I don´t know what" in the images rendered using biased method. But I think this will fall into the adding a new render engine part, so I´m not so hopeful to see this coming.
User avatar
Hieru
Licensed User v1.5
Licensed User v1.5
Posts: 1950
Joined: Fri Jun 26, 2009 11:47 am
Location: UK
Has thanked: 54 times
Been thanked: 22 times
Contact:

Wed Feb 10, 2016 11:07 am

Aside from it being such a great engine, I think most people have adapted so well to Presto because they either upgraded or replaced their hardware in anticipation of what Presto (or GPU rendering in general) could achieve.

When Presto first came out I could only use it with CPU. Part of the motivation for building a new rig was the need to get a high spec GPU and take full advantage of Presto. I did however go for a pair of Xeon chips, so I exclusively use Presto with CPU + GPU.

It also helps that there was little to no learning curve when it came to the differences between using Presto and TR1/TR2 :thumbup: .
Post Reply