Show your LW2018 Test Renders

[Svein-Erik]

Another possible reason for long render time?

Laptop, win 10, cpu i7 7700HQ 2.8 GHz, systemdisk: Intel pci express form factor M.2
Rendering a very simple scene I got:
1: LW2015 rendering 35 sec
2: LW2018 rendering 120 sec

Then I used Defraggler from Piriform and defragmented all files under Lightwave 2018 directory. Then the next test showed LW2018 rendering very close to LW2015.
This is very strange because the system disk is a pci express disk.
If I remember correct, many functions has now been written in Python, could a reason be Python memory management or something?

I did these tests twice, but cannot repeat these tests to day because right now I am testing/using Bitlocker encrypting.
Btw. Bitlocker is NOT adding anything to the render time.

-sd-

 

[MichaelT]

...

 

[MichaelT]

It is bucket rendering, the GPU is set to 2 minutes. The extra 11s are for whatever it is wrapping up at the end.
Vray can be compared.. because it is primarily a CPU based renderer.

In any case.. I won't render anything from the internal renderers (there are four of them)

And also.. if I did.. then yes.. the standard renderer is slower to get clean in low light too. But not as slow.
And perhaps more important.. I actually can get the image to be very clear from noise. Something that (at least for me)
is significantly harder to achieve in the new renderer in LW.

Update:

I decided to add one image from the standard renderer in C4D. This is 2m 2sec:
I just changed renderer and did some quick changes. nothing special..
There are two outside light sources.. hence the brightness.. and the weird shadows. But I could not be bothered to change it.

 

[jbrookes]

Nice work! While 2018 does have a few issues I am amazed at how good the renderer looks. The lighting system is so beautiful that for most work radiosity is not needed. And shading is a real joy with the new system. To each there own but for $300 it's a bargain IMO. I'm using it now on a product shot list of 28 hi-res images and I have had 0 crashes. Once they polish the renderer to a greater degree and eliminate some of the firefly issues it will be a top renderer.

I have to agree as well. Lighting is something I wanted to see improvements in, and from what I can tell it looks like they've accomplished that with LW2018. Although I have issues with the rendering times and -- to some degree -- the long-ish workflow for stamping out rendering artifacts, I also think LW2018 offers a lot for the price. Yes, I've found bugs (but only minor ones and nothing near to the degree of what I find in a well-known competitor's products that begin with an 'A' and end in a 'K'). I've yet to have a full-on crash in LW2018.

 

[jwiede]

I've yet to have a full-on crash in LW2018.

I wish I could say that, I've encountered quite a few. Rendering's pretty solid, though.

 

[Schwyhart]

Layout has been open on my 2013 Mac Pro for a week and it hasn't crashed. I've rendered several times too.
My trial period is up, but as long as it stays open, I can continue using it. I know, I'm a horrible Lightwaver, but I'm planning on purchasing it with my tax refund though!

 

[jbrookes]

In my opinion, at this stage, the LW 2018 rendering engine is very weak.
I guess it is even worse than LW 2015.
Maybe NT will improve it in a while, but for the moment I am very disappointed.
3 years of waiting and the release of LW 2018 is dramatic.
I still wish the best for NT and LW, but I can not wait any longer.

I am in the process of changing the software. I'm going to Cinema and Corona or Vray.

sad but true, my patience is over

Based on your portfolio, LightWave 2015 is clearly serving you well. What in your opinion would you like future versions of LightWave to deliver?

If it's the modeling/surfacing side of things, you could always take Chris Jones' approach and supplement LightWave with Sculptris and Krita (both are free). And if you need a compositing app, Digital Fusion is also free now. His animation "Ed" is certainly what I'd consider a good proof of concept. For those few who haven't seen it: https://www.youtube.com/watch?v=HjHiC0mt4Ts

As for the free apps...

https://pixologic.com/sculptris/emails/operating-system.php

https://krita.org/en/

http://www.cartoonbrew.com/tech/fusion-compositing-software-now-free-available-macs-139573.html

 

[artzgo]

You have to write a lot about what is weak in LW ... one day I will describe my point of view.
Sad that not much has been done for the last 3 years.

See the sample render for Corona, I did it in 5 min.

 

[MichaelT]

I wouldn't say that.. much have been done. But perhaps not the areas where it is of most importance to you (or me for that matter) For me two things are of greatest importance.. 1) Speed. The new renderer isn't fast enough. If a scene is well lit.. then yes.. it is fast. But if you have darker areas in the scene (which is for me, nearly always the case) then no.. it isn't fast. And that have more to do with point 2) Too easy to get noise. Especially in darker areas. And to get rid of it, you either have to rely on the interpolation which is woefully susceptible to "splotches", or up the samples to ridiculous levels. Making the scene take hours (or even days) to finish. (Frankly I think the interpolation method must be replaced with something better. Those splotches it creates are truly annoying.)

Now the point (at least for me) is that other tools out there render the same scene, to nearly noise free levels in a fraction of the time. I know (and I've said it before) I could very well be doing things very wrong. I admit that, but then I (or we) need good tutorials on how to achieve both fast renders, with little to no noise. By using more challenging scenes, that have areas where you'd expect a lot of noise. Like shadowy areas etc.. But if I am *not* doing things wrong.. then LWG simply must up their game here. They can look at plenty code out there that does things fast. The early versions of what later became Octane is out there (even on Github last I looked) .. or blender for that matter. So there isn't that there are lacking resources to learn from. I'm sure there are plenty of examples out there on how to deal with low light and get rid of noise as efficiently as possible. And at some point look into GPU.

But these two areas are like I said, at least *my* main issues with the current renderer. Otherwise I am very happy with the new version, there are noticeable improvements. And I am especially happy that things seem to move faster.. and the dialogue is more open. But that also means it is important to listen to critique, and respond to it. In my case.. I can very likely be wrong, and that is totally fine.. but so is the opposite. As long as it means things are moving forward in a constructive way.

That's my two cents anyway.

 

[Ztreem]

you could always take Chris Jones' approach and supplement LightWave with Sculptris and Krita (both are free). And if you need a compositing app, Digital Fusion is also free now. His animation "Ed" is certainly what I'd consider a good proof of concept. For those few who haven't seen it: https://www.youtube.com/watch?v=HjHiC0mt4Ts

As for the free apps...

https://pixologic.com/sculptris/emails/operating-system.php

https://krita.org/en/

http://www.cartoonbrew.com/tech/fusion-compositing-software-now-free-available-macs-139573.html

I did exactly that. I started to use sculptris for sculpting and then moved over to Blender to do smoke simulation, water simulation, sculpting, UV's and some texture painting etc ... and now I have a hard time to see what I should use Lw for. Blender has almost all feauters and more and in most cases much faster workflows, active communnity, open communication from devs, lots of plugins, more rendering engines, do I have to say more. Maybe LW will unlock some of its long lasting limitations in a couple of versions, but I don't feel like waiting that long.

 

[prometheus]

I did exactly that. I started to use sculptris for sculpting and then moved over to Blender to do smoke simulation, water simulation, sculpting, UV's and some texture painting etc ... and now I have a hard time to see what I should use Lw for. Blender has almost all feauters and more and in most cases much faster workflows, active communnity, open communication from devs, lots of plugins, more rendering engines, do I have to say more. Maybe LW will unlock some of its long lasting limitations in a couple of versions, but I don't feel like waiting that long.

I did try to load a skinfaxi spaceship model found on foundation3d, lightwave loads within a couple of seconds, blender takes minutes, and it all ended up like crap as well in blender.
consider trying to use the same amount of subdivision and run displacements on it in Blender 2,79, VS lightwave, in my experience ..Lightwave is Way faster to handle that, (blender sculpting excluded)
So for heavy scenes, I wouldn´t put a dime on working with blender really, unless evee brings much more better handling in displacement on dense polymeshes.

Also, I miss lightwaves wonderful world of large amount of various procedural textures, not blender, not houdini has that much.

Then I also have issues with how blender handles volumetric scattering in terms of rendering speed, quality, VS turbulenceFD...and also the fact that turbulenceFD is easier to work with the shading, then again blender has other smoke and fire features that is better than turbulenceFD, Including ...slightly cheaper

But yeah..I may start to spend more time in blender as well VS Lightwave...not sure though, still have almost two months to decide.

 

[jbrookes]

You have to write a lot about what is weak in LW ... one day I will describe my point of view.
Sad that not much has been done for the last 3 years.

See the sample render for Corona, I did it in 5 min.

There's a lot more to 3D CGI than rendering lifelike images. Otherwise all we'd have would be a bunch of render engines.

I've seen a lot of debate (especially in the last 10 years) over who has the best render engine. Many times the examples held up as 'the best' have serious deficiencies in other areas.

Strategically, now's the time for you to list what you want out of LightWave (while the devs have time to actually implement revision changes).

 

[MichaelT]

I'm happy with most of the updates. Although it is clear this is an interim release, as they are moving things over. Speaking of "wonderful" things.. Houdini is the only renderer I know which.. if you set the renderer to some higher settings for sampling.. where the shadows suddenly can begin warping (notice the window shadows on the floor) So every renderer has its quirks I suppose:

 

[Ztreem]

I did try to load a skinfaxi spaceship model found on foundation3d, lightwave loads within a couple of seconds, blender takes minutes, and it all ended up like crap as well in blender.
consider trying to use the same amount of subdivision and run displacements on it in Blender 2,79, VS lightwave, in my experience ..Lightwave is Way faster to handle that, (blender sculpting excluded)
So for heavy scenes, I wouldn´t put a dime on working with blender really, unless evee brings much more better handling in displacement on dense polymeshes.

Also, I miss lightwaves wonderful world of large amount of various procedural textures, not blender, not houdini has that much.

Then I also have issues with how blender handles volumetric scattering in terms of rendering speed, quality, VS turbulenceFD...and also the fact that turbulenceFD is easier to work with the shading, then again blender has other smoke and fire features that is better than turbulenceFD, Including ...slightly cheaper

But yeah..I may start to spend more time in blender as well VS Lightwave...not sure though, still have almost two months to decide.

No software is perfect. If I would do a list of all + and - of both I think that right now Blender wins. But I’ve just started to use Blender so I need to do more actual projects to know how it performs. I still have to test LW2018 but that has to wait until I’ve used and decided if Blender can do my daily work in a more effective way, which I think it can.(unified environment and undo helps a lot)

 

[prometheus]

No software is perfect. If I would do a list of all + and - of both I think that right now Blender wins. But I’ve just started to use Blender so I need to do more actual projects to know how it performs. I still have to test LW2018 but that has to wait until I’ve used and decided if Blender can do my daily work in a more effective way, which I think it can.(unified environment and undo helps a lot)

What I see problematic with Blender wins, it´s where you are in the work process, and by that for example...huge scenes, and that is at the end of the line where it seems blender can not cope with too much loaded stuff, you can not bypass that...for modeling, you can always bypass a lot with lightwave, though it may not come as easy as working with some modeling stuff in blender, and besides..you can always use use blender together with lightwave, what you can not do it seems, that is to replace the scenery at the end of the line as I intially mentioned, that Is where I think Lighwave winds..and blender falls flat, it´s just that if you do work with huge poly amount and displacements..it seems blender can not deal with it in the same manners like lightwave, and that is what I think very important.

As for smoke and fire, don´t think the native smoke and fire engine is at the same level when simulating higher res simulations and at larger scales, where I believe turbulencen wins...that is my initial impressions, though I need to test those muche more.

It all of course comes down to what you work on, product shots, liquid commercials, games, archviz..space shots and dynamic vfx shots etc, so highly individual I suppose..if blender wins or not.

 

[Bytehawk]

the last one

like the camera view to modeler functionality
don't like the render times and fireflies

undecided if I ll upgrade, trial is over

 

[Rayek]

I did try to load a skinfaxi spaceship model found on foundation3d, lightwave loads within a couple of seconds, blender takes minutes, and it all ended up like crap as well in blender.
consider trying to use the same amount of subdivision and run displacements on it in Blender 2,79, VS lightwave, in my experience ..Lightwave is Way faster to handle that, (blender sculpting excluded)
So for heavy scenes, I wouldn´t put a dime on working with blender really, unless evee brings much more better handling in displacement on dense polymeshes.

Also, I miss lightwaves wonderful world of large amount of various procedural textures, not blender, not houdini has that much.


But yeah..I may start to spend more time in blender as well VS Lightwave...not sure though, still have almost two months to decide.

To be fair, that Skinfaxi Battlestar model is in Lightwave's native format, so I would expect it to load quickly. Lightwave doesn't support *.blend files, after all ;-)

Loading foreign 3d formats is always going to be slower in 3d software. With lwo files it took me some trial and error to come up with a painless import workflow. I first load and resave the lwo files with Accutrans, because that fixes incompatibilities with Blender's lwo importer. The 18mb file loads up in around 11 seconds on my machine (which is pretty old by now). That's not bad for a non-native file format in my book. The full model takes around 40 seconds to load up. UV textures and materials are converted to the old Blender internal renderer, so I use the "Convert All to Cycles" material add-on to automatically convert all the materials to material nodes. UV textures and image maps are retained, but non-UV mapped surfaces and LW native procedural textures need to be manually repaired.

When I save the scene as a native Blender file, loading up the complete scene takes LESS than a second. Click, done. Instant loading. The same native scene in Lightwave takes five seconds to load. Blender will always be faster in this regard, because a blend file is basically a memory dump.

As for viewport performance: avoid texture view mode. Either material or solid mode is just as smooth as Layout. I duplicated the ship twenty times (29 million quads) and the viewport keeps running at 25fps. I use Blender 2.79 (one of the latest builds).

Seems to me that Blender can handle larger scenes quite nicely as well. And scene management in Blender is ahead with its library system, in my opinion.

Anyway, a while ago I tested several very heavy scenes in both 2.79 and the alpha 2.8 release, and my initial tests indicate that 2.8's viewport runs SIGNIFICANTLY faster than 2.79. At least five - ten times as fast depending on the scene.

I also duplicated the Skinfaxi Battlestar model 8 times in Layout. While the viewport remained smooth, object selection became awfully laggy, and it took at least three or four seconds before Layout resolved the click and the object that I had selected. Blender lags selecting object with 20 ships and their components as well, but selection takes around a second and a half to resolve. Faster with the outliner. This scene takes up 30gb of RAM, btw.

Yes, there are viewport performance issues with Blender, but they aren't nearly as bad as some people think. A lot is mitigated by learning about certain weaknesses. And don't forget that the subdivision modifier has that nice OpenSubDiv option to take advantage of the GPU to speed up things tremendously while animating subd's.

I hear you regarding the built-in procedural textures: I miss the ones in both Lightwave and C4D when working in Blender.

PS a question on the side: why are 99 out of a 100 Lightwave models so messy? I always need to clean up the geometry: double vertices in particular are a given when I load up lwo models in Blender (w-->remove doubles). And often a lot of unnecessary tris - alt-j to clean those up. I just wonder what's up with that?

 

[Ztreem]

What I see problematic with Blender wins, it´s where you are in the work process, and by that for example...huge scenes, and that is at the end of the line where it seems blender can not cope with too much loaded stuff, you can not bypass that...for modeling, you can always bypass a lot with lightwave, though it may not come as easy as working with some modeling stuff in blender, and besides..you can always use use blender together with lightwave, what you can not do it seems, that is to replace the scenery at the end of the line as I intially mentioned, that Is where I think Lighwave winds..and blender falls flat, it´s just that if you do work with huge poly amount and displacements..it seems blender can not deal with it in the same manners like lightwave, and that is what I think very important.

As for smoke and fire, don´t think the native smoke and fire engine is at the same level when simulating higher res simulations and at larger scales, where I believe turbulencen wins...that is my initial impressions, though I need to test those muche more.

It all of course comes down to what you work on, product shots, liquid commercials, games, archviz..space shots and dynamic vfx shots etc, so highly individual I suppose..if blender wins or not.

I don't do much clouds or displacements, but if I do then its just a small part of all I do. I do agree that for displacement high poly counts LW is both faster and better with more procedurals.

huge scenes, and that is at the end of the line where it seems blender can not cope with too much loaded stuff, you can not bypass that...for modeling, you can always bypass a lot with lightwave

I don't really follow you here?


For me it's the small everyday things that makes a big difference. I just tested a 3 million poly fbx here at work and LW2018 actually loaded it a few seconds faster, but it missed the most importent thing...VERTEX NORMALS!!! So in the end, fast but useless.

Then its things like targeting an object towards another and in LW you must have your object modeled towards Z+ to get it right, in blender you can just choose which axis you want to use, quick and easy. I can list a lot of these examples and it's all these small things that I think makes the overall difference and gives a smooth user experience.
As always it all depends of what you do and what you need and how you like to work.

 

[mummyman]

Those speckles are called fireflies. It is an artifact when you get a sample ray with extremely high intensity. You can use clamp in the render settings to limit the effects. The downside to that is that you may lower the overall intensity of the scene as a whole (like putting a dampening effect on it) So sometimes it is better to choose a higher limit, and then fix the image in post. Like in PS or AE for instance. But it is a problem with PBR rendering in general unfortunately.

I am thinking of making a tutorial to help people out on how to deal with it. But maybe Craig is already working on that?

I'd like to see a tutorial. I'm getting fireflies without GI and Principled BSDF glassy look. Trying all sorts of settings. Gonna have to stick to 2015 until I can get better results. Dang.

 

[omichon]

I'd like to see a tutorial. I'm getting fireflies without GI and Principled BSDF glassy look. Trying all sorts of settings. Gonna have to stick to 2015 until I can get better results. Dang.

Try not using Normalized lights. In most of my tests, I get rid off fireflies by turning off this option, and cranking up the light intensity to compensate.
More a workaround than a real fix, but there must be something broken with normalized lights at the moment.

Edit : At least, I was able to deliver my first project rendered with 2018 (visible on my website) and without a single firefly ;D