PDA

View Full Version : Facebook Lightwiki 2018 tests and bugs



gar26lw
01-08-2018, 06:01 AM
https://www.facebook.com/groups/lightwiki/

there is a lot of good info by a dude called jason on the wiki. tbh, i think it would be more useful if the tests and conversation were held here.

lw devs, it’s worth reading, although i have to say, facebook is not the best place for this kind of communication. there could be some issues in 2018 with bullet dynamics and render settings, aa, gi and such.

something else to note - andrew comb should add some info and knowledge to the documentation regarding rendering settings, shaders and lights.

cheers

DogBoy
01-08-2018, 06:30 AM
I'm not so sure. There is a bunch of spurious stuff getting posted there.
That isn't a jibe at Jason BTW.

gar26lw
01-08-2018, 06:36 AM
i think he’s doing a great job and being extremely tenacious.

there’s also a video there from someone else illustrating ambient and an application thereof in someone’s particular workflow and process. whether you agree with it or not, this is a great example of showing that while a feature may not seem needed (especially to those that only render a certain type of material) or be mathematically correct or in keeping with correct rendering, there is a case to have that feature. tbh, i can’t help but feel that some people’s comments are quite selfish and have no regard for or empathy towards others npr/artistic needs. anyway, i digress but i just wanted to get that off my chest :)

i think someone else pointed out a fear of lw getting too mathy with nodes and such. i would say i can see that as something that need balancing.

raymondtrace
01-08-2018, 07:15 AM
If LW devs are reading this, the advice I would give is to get in front of the confusion. Post some structured training that shows how to use LW 2018. Everybody seems to be stabbing in the dark. I appreciate the stuff Rebel Hill is preparing but the lack of official training from NewTek shows poor preparedness in this release, especially when a lack of education about new processes makes render times seem so much longer. That is poor marketing.

DogBoy
01-08-2018, 07:23 AM
i think he’s doing a great job and being extremely tenacious.

there’s also a video there from someone else illustrating ambient and an application thereof in someone’s particular workflow and process. whether you agree with it or not, this is a great example of showing that while a feature may not seem needed (especially to those that only render a certain type of material) or be mathematically correct or in keeping with correct rendering, there is a case to have that feature. tbh, i can’t help but feel that some people’s comments are quite selfish and have no regard for or empathy towards others npr/artistic needs. anyway, i digress but i just wanted to get that off my chest :)

i think someone else pointed out a fear of lw getting too mathy with nodes and such. i would say i can see that as something that need balancing.

Yeah, I agree that if someones workflow relies on a feature, you shouldn't give them too much hassle: It may not be physically correct, but they use it constantly. That said a PBR based renderer may not be a good fit for them.
Jason is being tenacious, i just hope he is actually reporting these issue and not just grumbling to the choir.

- - - Updated - - -


Post some structured training that shows how to use LW 2018.
That is what Antti is doing. That is what Andrew Comb is doing.

MarcusM
01-08-2018, 07:39 AM
In my opinion most examples from LW 2018 content should be rendered or video captured and put on YouTube. Other software users and older LW's should have easy way to see that. Blog is not well known outside i think.

samurai_x
01-08-2018, 07:56 AM
I'm not so sure. There is a bunch of spurious stuff getting posted there.
That isn't a jibe at Jason BTW.

The problem is people are telling Jason that the new renderer is fast. So Jason keeps going in circles posting tons of examples.
When did a pathtracer become fast? Not unless you have fast processors like gpus.

There are two or three arnold users there which is pretty close to the new lw engine. They'll tell you arnold is slow. But can crunch heavy scenes.
The new lw renderer is more suited for studio production work that have render farms, which is ironic because there are probably more lw users who are freelance with only a few computers for rendering.

rustythe1
01-08-2018, 07:57 AM
yea, he maybe doing a good thing testing the hell out of it, but i wouldn't hold too much weight with everything he says or thinks he finds, with the risk of sounding offensive for someone who is listed as working at an animation studio and has an 88 thread machine rendering, he doesn't seem to get the apples and oranges thing and is constantly trying to make the orange an apple, which it will never be, even today i think he is trying to compare how all the 2015/2018 content performs in vies versa, dynamics and everything and was very quick to go to bed when i took up his challenge of getting the same render out of 2018 from 2015 faster, which again turned into "the settings are not the same!" no the question was can you get the same results, not the same settings?

THIBAULT
01-08-2018, 07:59 AM
The problem is people are telling Jason that the new renderer is fast. So Jason keeps going in circles posting tons of examples.
When did a pathtracer become fast? Not unless you have fast processors like gpus.

There are two or three arnold users there which is pretty close to the new lw engine. They'll tell you arnold is slow. But can crunch heavy scenes.
The new lw renderer is more suited for studio production work that have render farms, which is ironic because there are probably more lw users who are freelance with only a few computers for rendering.

Totaly agree !

rustythe1
01-08-2018, 08:00 AM
The problem is people are telling Jason that the new renderer is fast. So Jason keeps going in circles posting tons of examples.
When did a pathtracer become fast? Not unless you have fast processors like gpus.

There are two or three arnold users there which is pretty close to the new lw engine. They'll tell you arnold is slow. But can crunch heavy scenes.
The new lw renderer is more suited for studio production work that have render farms, which is ironic because there are probably more lw users who are freelance with only a few computers for rendering.

that's not what he was asking, he is trying to get a simple plain render fast, but using the same settings you use in 2015, you cant, but you can get fast renders by doing things differently

DogBoy
01-08-2018, 08:00 AM
There are two or three arnold users there which is pretty close to the new lw engine. They'll tell you arnold is slow. But can crunch heavy scenes.
The new lw renderer is more suited for studio production work that have render farms, which is ironic because there are probably more lw users who are freelance with only a few computers for rendering.

Or have invested in GPUs/Octane.

DogBoy
01-08-2018, 08:04 AM
... who is listed as working at an animation studio and has an 88 thread machine rendering I think you are confusing him with another guy, Abdelkarim Keraressi.

DogBoy
01-08-2018, 08:09 AM
that's not what he was asking, he is trying to get a simple plain render fast, but using the same settings you use in 2015, you cant, but you can get fast renders by doing things differently

I agree, he is missing the apples/oranges analogy.
He posts here a lot, so maybe he'll wade in and explain.

gar26lw
01-08-2018, 08:09 AM
yea, he maybe doing a good thing testing the hell out of it, but i wouldn't hold too much weight with everything he says or thinks he finds, with the risk of sounding offensive for someone who is listed as working at an animation studio and has an 88 thread machine rendering, he doesn't seem to get the apples and oranges thing and is constantly trying to make the orange an apple, which it will never be, even today i think he is trying to compare how all the 2015/2018 content performs in vies versa, dynamics and everything and was very quick to go to bed when i took up his challenge of getting the same render out of 2018 from 2015 faster, which again turned into "the settings are not the same!" no the question was can you get the same results, not the same settings?

there is a good point with some stuff. i tried rendering a scene posted here with a poly with a checker procedural on it. thatís all and 2015 was twice as fast. i do find that puzzling as thereís nothing much going on and iíd expect, in that case, for both renderer to be neck and neck. am i being dumb? :)

samurai_x
01-08-2018, 08:10 AM
that's not what he was asking, he is trying to get a simple plain render fast, but using the same settings you use in 2015, you cant, but you can get fast renders by doing things differently

He has posted a bunch of threads there. I skimmed through it.
But I don't think he's the guy with the 88 thread machine doing interiors with noisy results.
I think they're genuinely surprised how slow and noisy pathtracers are. :D

gar26lw
01-08-2018, 08:16 AM
andrew comb posted some stuff about samples and such but it’s buried in comments. i think he could write a very good explanation of the new settings and processes, for the manual. he was on the beta for all that time, consider it payment :p

rustythe1
01-08-2018, 08:40 AM
He has posted a bunch of threads there. I skimmed through it.
But I don't think he's the guy with the 88 thread machine doing interiors with noisy results.
I think they're genuinely surprised how slow and noisy pathtracers are. :D

139294

ianr
01-08-2018, 09:36 AM
If LW devs are reading this, the advice I would give is to get in front of the confusion. Post some structured training that shows how to use LW 2018. Everybody seems to be stabbing in the dark. I appreciate the stuff Rebel Hill is preparing but the lack of official training from NewTek shows poor preparedness in this release, especially when a lack of education about new processes makes render times seem so much longer. That is poor marketing.



Agreed but maybe Dr.Cross said Enough! Launch!

Hence the LWO3 loader pinged out onto the LW Plug-ins Page

on the 1st sadly unannounced here or wiki?

ElusiveElephant
01-08-2018, 10:53 AM
All credit to Jason, he's definitely making the most out of his Trial :)

I wouldn't say 2018 is faster than 2015 in all areas, and can be noisy under certain conditions, but it sure as hell looks more realistic and is easier to control. Also has a tonne more functionality. Just have to choose what you're after in production. If you're after surfaces and volumes that react realistically to light, and a renderer you can use PBR workflow without any hassle then 2018 is nice. If you want to build your own shaders and materials from scratch using comparatively unrealistic lighting, or do non-photoreal stuff, then maybe stick with 2015.

For the asking price though, you may as well just jump in and learn how the new one works anyway :)

samurai_x
01-09-2018, 12:42 AM
139294

I think he linked the guy with that machine that did the interior render. Look at Jason's previous post in that thread he's just mentioned the 88 threads thing. Or ask him here.

https://s13.postimg.org/9w5xthql3/26696764_1949677281950936_1525880707_n.jpg

His new post there is disable inverse falloff. :twak:

rustythe1
01-09-2018, 01:13 AM
well we found a bad bug when using bullet and bones yesterday, if you open the mech leg in content and recalculate the dynamics with the display subpatch set above 0 it is painfully slow, as also is scrubbing the timeline, set it to 0 and it runs fine, even though the bullet is only attached to the bones, even if you disable the dynamics after the bone deformations are painfully slow, yet if you work on a scene that has one or the other on a dense subpatch mesh its fine.

gar26lw
01-09-2018, 06:26 AM
that’s a nice find.

jboudreau
01-09-2018, 08:11 PM
I think you are confusing him with another guy, Abdelkarim Keraressi.

Yes that certainly was not me. thanks man :)

jboudreau
01-09-2018, 08:22 PM
I think he linked the guy with that machine that did the interior render. Look at Jason's previous post in that thread he's just mentioned the 88 threads thing. Or ask him here.

https://s13.postimg.org/9w5xthql3/26696764_1949677281950936_1525880707_n.jpg

His new post there is disable inverse falloff. :twak:

Yes that is exactly what I was doing. I was referring to other users that were on the site saying the rendering was slow and they had Dual Xeons with 88 threads. I don't have that. I mainly use my laptop which only has 8 thread, My main machine downstairs in my office which is currently being renovated (hense why I am on my laptop) Is a Dual Xeon with 64 threads.

Yes I know and I got tons of flac for that one LOL.

Thanks,
Jason

- - - Updated - - -


well we found a bad bug when using bullet and bones yesterday, if you open the mech leg in content and recalculate the dynamics with the display subpatch set above 0 it is painfully slow, as also is scrubbing the timeline, set it to 0 and it runs fine, even though the bullet is only attached to the bones, even if you disable the dynamics after the bone deformations are painfully slow, yet if you work on a scene that has one or the other on a dense subpatch mesh its fine.

Yeah that one is still a mystery to me, I reported it and gave as much information as I could to the development team. Hopefully they can get to the bottom of it.

jboudreau
01-09-2018, 08:49 PM
yea, he maybe doing a good thing testing the hell out of it, but i wouldn't hold too much weight with everything he says or thinks he finds, with the risk of sounding offensive for someone who is listed as working at an animation studio and has an 88 thread machine rendering, he doesn't seem to get the apples and oranges thing and is constantly trying to make the orange an apple, which it will never be, even today i think he is trying to compare how all the 2015/2018 content performs in vies versa, dynamics and everything and was very quick to go to bed when i took up his challenge of getting the same render out of 2018 from 2015 faster, which again turned into "the settings are not the same!" no the question was can you get the same results, not the same settings?

Sorry man that was not me, but i'm always interested in a challenge haha ;)

This is your guy "Abdelkarim Keraressi "

Thanks,
Jason

samurai_x
01-10-2018, 12:51 AM
I didn't know lw 2015 modeler had a problem. I always used lw 11 32bit to model and only rendered in lw 2015 64bit.
Is this confirmed?

https://image.ibb.co/mN88xR/fps.jpg

gar26lw
01-10-2018, 01:24 AM
I didn't know lw 2015 modeler had a problem. I always used lw 11 32bit to model and only rendered in lw 2015 64bit.
Is this confirmed?

https://image.ibb.co/mN88xR/fps.jpg

yes, same for me. model in x32 and render in x64, linked via HUB

samurai_x
01-10-2018, 01:56 AM
Need to test this. So we gained a bit of speed in lw 2018 modeler but lw 11 is still fastest using multitexture. Multitexture is my default shading mode in modeler. Not good.

rustythe1
01-10-2018, 03:14 AM
another bug, camera selector not working, changes in the viewport but when render scene it dosnt change camera and just renders the first camera and then just the backdrop when it should be on the second camera, if you move the time slider to the second camera it then only renders the second camera view for all frames, if you remove the plugin, the renders become stuck on last selected camera, (tried both versions as there seem to be 2 in there)

BeeVee
01-10-2018, 03:50 AM
There are two in there, which is a bugbear of mine for a long time. One has a nice interface and all, but doesn't work with distributed rendering, the other lacks a nice interface but does. I can never remember which is which until I have already started rendering...

B

rustythe1
01-10-2018, 06:11 AM
they seem to have more or less the same interface now but the both have the same results
139325
"Cameraselector" dosnt have the clear button, "camera selector" does

BeeVee
01-10-2018, 06:17 AM
The one on the right doesn't work with Screamernet if I remember rightly. Submit it as a Feature Request that they either get rid of one or make it clearer. :) (I've done it often enough)

B

rustythe1
01-10-2018, 08:10 AM
but neither of them work at all when doing standard scene renders anyway (have bugged it) i just have a scene with a perspective camera and ortho camera set up to give quarter views then ortho views of a model, but the camera dosnt change with "render scene" but does when you scrub the timeline and use f9

JamesCurtis
01-11-2018, 10:13 AM
I'll have to try this myself as I do use multiple cameras when doing my industrial work for clients.