PDA

View Full Version : 9.6 to 2015



regenerator
11-28-2015, 07:06 PM
Hello!

I downloaded 2015 demo the other day to make sure my PC can handle it as suggested by tech support. I got Dell Insipron All-In-One (6GB RAM, NVIDIA GeForce GT 525M, Intel Core i5-2400 @2.5 Ghz (4 of them). So far so good but when I render a scene from 9.6 the render times are astronomical and I got all these specks all over the scene. I see the anti-aliasing is different than 9.6. I usually set it to 17 and was good to go. I also see Shading and Lighting Samples from some reading I've done are very important. Not sure ISBG samples play an important role as well. Would like to upgrade but if the render times are this bad then maybe my machine just can't hang with this version. Not able to post screenshots of the scene yet, sorry about that. It's displaced terrain over 6 million polygons. Uses about 400+ MGB of RAM with some fog. Also, Image caching was off in preferences so I don't think it's that. Any advice would be appreciated. Maybe the demo is purposely slow? What took 9.6 a few hours takes 2015 an estimated 15!

For 2015 I set the anti-aliasing or what's called Minimum Samples in Camera Properties to 200. Yes, it's very high but I had really bad specks so was just playing with the values. I never use Adaptive Sampling so that is off. I set Shading and Lighting Samples to 7 just to see. I may have gone buckwild with the settings.

ernpchan
11-28-2015, 09:40 PM
Look into this unified sampling thread. http://forums.newtek.com/showthread.php?137003-Unified-sampling-question

A lot has changed since 9.6 so settings you had there don't necessarily translate in 2015.

And of course all that will change when 2016 comes out since the render engine is completely different.

regenerator
11-29-2015, 06:43 AM
Thanks for the info ernpchan. Last night I used some rational sampling values and the render time was close to 9.6 but image quality was still a little off especially the Hypervoxels (plankton) in the scene.

regenerator
11-29-2015, 08:25 AM
Here is a render from 9.6:
131258

And here is the same file for Lightwave 2015. Notice the difference of the plankton (Hypervoxels) when in over the terrain as opposed to when they are floating above. Don't know how to fix this.
131259

Here are the settings:
131260 131261 131262

prometheus
11-29-2015, 10:11 AM
Volumetric radiosity which you have turned on is a huge Rendertime clogger, unless really needing it..turn it off.
importance sampling will be much slower than without.
directional rays may also cause slow down sometimes..

you got extreme high aa of 200..no wonder, you got fog and hypervoxels with volumetric radiosity, no wonder, also turn of volumetric AA in the volumetric tab..it increases rendertime a lot, and if you are not using volumetric lights, thereīs really no need for it.
you also got a 6 million polyterrain displaced, the system has to freeze it..and no wonder it takes time.

Thereīs a lot of places you need to optimize I think.
Canīt tell anything about the difference with hypervoxels unfortunatly.

regenerator
11-29-2015, 12:13 PM
Hi Prometheus,
Agreed about the 200 AA. Was experimenting. Did a render of a much lower AA today and it looked better but no cigar. Going to apply your recommendations today. Maybe that will resolve those splotches Hypervoxels. Who knows. Thanks for the advice. Much appreciated.

prometheus
11-29-2015, 02:44 PM
regarding terrain or ground displacement, think about the movie prometheus and the terrain resolution of around 25 millions of polygons, with very high resolution texturs(a bit unknown)
took around 8 hours per frame..using cinema4d, (and probably also some good hardware too) cinema4d which I also believe handles the geometry much better than lightwave( though actual rendering speed comparison is unknown)

regenerator
11-29-2015, 08:38 PM
Loved the CG in 'Prometheus'. BTW I solved the Hypervoxels problem. For some reason in 2015 having Unpremultiply Alpha in Output causes the strange artifacts. Set it to Premultiply and they look fine. Now finding a way for the terrain not to look grainy. Playing with the Bump values. They could be too high.

prometheus
11-29-2015, 09:28 PM
yeah bump values are tricky for adding detail, could that you have too strong bump, but it could also be that you are having simply to large bump scaling..for sediments it should hardly be noticable, accompanied by other a bit larger fractals that is the destortion of flows winds etc..but also here one need to find the right fractal and not too large.

regenerator
11-30-2015, 09:58 AM
Problem resolved! Thanks ernpchan for the link that contained Rebel Hill's video explaining unified sampling. And thanks prometheus for your advice as well which drastically improved render speeds. Was using brute force settings. Think I'll upgrade today. Love this forum.

Snosrap
12-09-2015, 07:46 AM
Yeah a lot has changed -for the better- in the LW11 to 2015 releases. You should see render times of the same or lower in most cases, but it is very scene dependent. Unified sampling is a bonus, but you will find that some plugins offer to override the unified system. (Some of DP's plugins come to mind.) Upgrading from 9.6 to 2015 should be a no brainer decision. :)

lardbros
12-09-2015, 10:17 AM
By the way... if you took your image (rendered out WITH an alpha channel) using the un-premultiplied alpha setting... it won't look right just viewing it as you have done...

BUT if you take it into a compositing tool, like AfterEffects and then multiply your image with black on import (using the interpret image), it'll come out as you'd expect/as per your other image :)