PDA

View Full Version : Gigantic MilkyWayi in Lightwave 3D - Tech Demo



3dhotshot
07-28-2018, 05:06 AM
Stumbled upon this guys video simply amazing !!!

https://www.youtube.com/watch?v=OUpTnklTWZ0

prometheus
07-29-2018, 01:14 PM
What isnīt impressive is that it takes a long long time for vpr to iterate, vpr does often iterate render the scene for fast previous, then you can assume the final renders gonna take a much much longer time as well.

dlvphoto
07-30-2018, 03:39 PM
What isnīt impressive is that it takes a long long time for vpr to iterate, vpr does often iterate render the scene for fast previous, then you can assume the final renders gonna take a much much longer time as well.

Considering this video is in LW 11.5 and is 5 years old, you're watching VPR iterate on what could well be a 5 year old mid-range system in an older version of Lightwave. I'd like to see the scene rendered on a current cpu or even rebuilt to use the current 2018 volumetrics, etc..

What I see on his video is darned impressive and I give him all the respect he's due. I'm not going to hold the refresh rate of an older version of Lightwave on a system from five years ago as any kind of performance benchmark for this scene.

Scott

prometheus
07-30-2018, 04:28 PM
Considering this video is in LW 11.5 and is 5 years old, you're watching VPR iterate on what could well be a 5 year old mid-range system in an older version of Lightwave. I'd like to see the scene rendered on a current cpu or even rebuilt to use the current 2018 volumetrics, etc..

What I see on his video is darned impressive and I give him all the respect he's due. I'm not going to hold the refresh rate of an older version of Lightwave on a system from five years ago as any kind of performance benchmark for this scene.

Scott

Donīt think it got to do that much with the age or the lw version...do you know how many volumetric lights he cloned? a not very efficient practice unless you use them in sprite mode,
you could try this in 2015 and it would still be darn slow.
2018 isnīt even to consider for comparison since itīs completely different in volumetrics and lights.

jwiede
07-30-2018, 11:12 PM
I'm not even sure you could still do what he did with volumetric lights in LW2018 -- at least, and still have VPR converge in any sort of humanly-tolerable timeframe. Without sprite mode, hundreds of volumetric lights cause VPR to converge very slowly. I just tried an experiment with a few hundred point lights, and it was painful.

Being restricted to the same scattering medium for all volumetric lights in scene poses a problem as well: He adjusted the different populations of star types to use different volumetric scattering params (yielding different-sized stars), but you can't really do that with LW2018. I tried using some nodal trickery in an attempt to adjust scattering for different groups of lights, but it slowed down the already-painful VPR convergence quite a bit further. I was already down to a fairly small limited region, and it took minutes per step for that to converge.