GI Rendering Speed - FPrime is how much faster than LW?

#1

All,

I know different benchmarks will yield different results, but “about” how much faster is FPrime than LightWave’s native renderer?

I’m rendering out GI character animation that’s pretty optimized for quality & speed, but 2 hours a frame is hurting my mojo and would be a huge issue on longer format productions.
Is a speed increase of 3X (I’m guessing wildly) realistic?

Realtime adaptive rendering isn’t my main concern, but would be a nice bonus if FPrime is a screamer.

BTW, I’m currently using (and like) the the fake SSS chanlum shader. Any decent alternative in FPrime since it doesn’t handle LW shaders?

Thanks. -Mike

#2

This really depends, but in many cases Lightwave’s native renderer is actually faster than fprime. FPrime is faster in very few occasions these days and I’d say that it’s main attraction today is it’s awesome previewer. All this is about the latest version of Lightwave of course, FPrime beats the native renderer hands down when compared to 9.0 and earlier versions.

:edit:

Oh, and be sure to check Except’s awesome radiosity guide if you are getting slow rendering speeds with 9.5 (2h per frame is really bad for modern hardware if your scene isn’t super complex). http://www.except.nl/lightwave/RadiosityGuide95/index.htm

#3

FP is starting to show its age as a previewer also. It can’t display many nodes, so unless you want to stick with traditional layers surfacing, its a bit limited. I still love it and use it all the time, but an update is due.

#4

Thanks guys!!!
Haven’t had time to read that link yet, but sounds like going from LW9.0 to 9.5 will do the trick. :slight_smile:

-Mike

#5

Just is case people are interested in my current settings:

My character I’m rendering is around 200,000 polys, with close up framing, full monte carlo GI, 1 bounce, 3x9 rays, 35 passes (for pass based DOF & MB), and no adaptive samping. The background plate isn’t used in the render, but very low poly boxy geometry is used to throw light at the charater. Light is from one point light and the backdrop color.

Render time at 800x600 on a 2.8GHz Dual Core is right at 2 hours using LW9.0.

-Mike

#6

Please post the rendering time after you upgrade to 9.5, so that we can see the difference :smiley:

#7

…and think about using PR motion blur and DOF…

B

#8

I thought rendering speed in LW increased dramatically from 9.0 to 9.2 - in both MC and Final Gather. I know that FG was extremely fast in 9.2. I would assume that it was faster still in 9.3 and 9.3.1. And hopefully faster in 9.5 as well.

#9

Those settings seem rather extreme…especially 35 passes!

Since using 9.5 I’ve opened FPrime once…discovered that hardly anything works in it anymore and it’s much slower/lower quality and haven’t touched it since.

#10

OK, I’m sold on 9.5 (which is a big deal for me because I’ve been looking hard at Modo).

VirtualFM,
I’ll try to post some times, but that won’t happen anytime soon.

BeeVee,
I’ll give the new MB & DOF a serious try.
I’ll be quite happy with a GI speedup of even 3X, which “sounds” pretty easy going from 9.0 to 9.5.

biliousfrog,
My FStop is 1.0, and even 35 passes actually isn’t enough for the background to blur correctly. Besides, using the older versions of LightWave (without the new MB & DOF) the number of passes you use shouldn’t be a “huge deal” in GI render times since using more passes allows you to lower your ray settings…the new motion blur & DOF should make this workflow obsolete in 9.5 though.

-Mike

#11

Without interpolation for GI?

#12

Well, your mileage may vary… I did some comparisons between 9.2 and 9.5 and had mixed results… Overall I hade better times with 9.5, but also had a scene where 9.5 was 2 times slower with no apparent increase of quality or reason to justify the increased rendering time. OTOH, there was another scene where 9.5 was almost 6 times faster (19m 9s versus 3m 17s in 9.5)

#13

unless you have lots of transparent glass adn also need lots of color bleeding, this is overkill imho. the upgrade to 9.5 will get you seriously fast final gather, along with a cached radiosity solution (if you’ve got objects moving though, mc is required, but can be run in interpolation mode as well).

the previously linked guide is essential, but even with a core2 duo, one shouldn’t be that much over 10mins a frame. maybe look at breaking up the scene to re-assemble the render-passes in post if there’s that much going on in-scene.

#14

Any chance you could submit that scene to lw-bugs (at) newtek (dot) com as a bug report? We’d like to get a look at it if possible. Thanks!

#15

Fprime is in need of an update that’s for sure, and for interiors with lots of bounces and limited light sources it’s not appropriate, but it’s still faster than LW for what it does.

Fprime will guarantee a No flicker animation render as it is a brute force method that is accurate and stable between frames. Lw’s GI can do this but is substantially slower.

LW’s interpolated modes aren’t appropriate for animation. It’s a cheat (and a very good one for stills and flythrough’s) but it’s not achieving the same thing that Fprime is.

#16

I’d have to say I completely disagree. More than a year ago we used LW’s Final Gather for a beautiful flicker-free animation that took leaps and bounds less time than FPrime did. We had quite a few lights with lots of bounces and while FPrime was good, it in no way apprached the speed of FG. And using the same setting in FPrime it took quite a bit longer and didn’t really look that great.

Fprime is wonderful, but it is VERY long in the tooth and needs an overhaul when it comes to nodes and speed. I still use FPrime every day, but when it comes to animations, IMO it falls short.

I guess it depends on exactly what settings you are using and what the animation is for.

#17

Nevertheless, LW’s interpolated GI modes do not support deformations. That’s not an opinion, it’s official.

So, I presume that you must either be using non interpolated, in which case I am shocked that you can get it faster than Fprime, as I’ve been extensively testing it and found it to be much slower, or you are simply animating objects on SRT only and using the animation cache.

#18

My 2 hour renders (via LW 9.0) might not be all that bad considering it eliminates animation flicker (mostly) on a deforming mesh. I bet 9.5’s interpolation will help, but I’m not expecting miracles.

-Mike

#19

It was an arch-viz walk-thru. We used interpolated Final Gather. There were animated 2D billboard walking humans (from Marlin), a realflow mesh object replacement fountain and a pool and pond with animated mesh deformation. We had no problems and everything looked great. The client liked it VERY much and we’re still making tweaks and alterations because they keep changing things.

FPrime was significantly slower - we have loved FG since it was introduced in 9.2. (At least I think it was 9.2) and use it in animations always. I love FPrime, but for us, LW FG is the way to go for animations. Perhaps if we did character work, that might change. We’ll of course be testing when we begin some character work next year.

But again, FPrime need lots of work to be updated to work WELL with the current version of LW. Though honestly I don’t even know if it will be updated. I’ve been a loyal Worley customer and have all of their plugins - quite a few licenses of FPrime and Sasquatch as well - and we haven’t seen any word from them for a LONG time. Sas hasn’t been updated for YEARS and FPrime is just getting too long in the tooth for our tastes. But again, that’s just my opinion.

#20

There will be new real-time interactive renderer soon… :slight_smile: