PDA

View Full Version : LW 2019.0.1 native render



Cageman
03-02-2019, 04:29 PM
Just a doodle. LW2019.0.1 with db&w Tri-planar mapping node + Patina node and Bulled Physics. :)

https://www.youtube.com/watch?v=qR_n5pGE5zc

Enjoy! :)

//Cageman

hrgiger
03-02-2019, 04:34 PM
Nice work.

Ztreem
03-02-2019, 05:38 PM
Rendering wise it looks really good. I don't like so much the rubbery feel as it breaks and that you can see the pre-fractured parts.

Cageman
03-02-2019, 06:34 PM
Rendering wise it looks really good. I don't like so much the rubbery feel as it breaks and that you can see the pre-fractured parts.

Maybe you are mixing up the pre-factured parts with what the Tri-planar mapping node shows?

EDIT: I wanted to have a felling of that Copper doesn't break easily... especially at the first shot from one of those chrome balls. Also, that it is in small scale. The whole statue reacts to the initial impact, but holds together, and breaks appart when the second shot hits it.

I might be totally wrong about it though, so you might have a point. :)

Ztreem
03-03-2019, 03:49 AM
Maybe you are mixing up the pre-factured parts with what the Tri-planar mapping node shows?

Maybe? We can call it artifacts instead. :D




EDIT: I wanted to have a felling of that Copper doesn't break easily... especially at the first shot from one of those chrome balls. Also, that it is in small scale. The whole statue reacts to the initial impact, but holds together, and breaks appart when the second shot hits it.

I might be totally wrong about it though, so you might have a point. :)

I understand what you’re after and also that this is a personal test project with time constrains. But if you would be able to see a copper statue wiggle like that it would only be if you film with a high speed camera, but I doubt a small statue like that would wiggle that much. I would buy the wiggle more if you did it in slow motion.

tischbein3
03-03-2019, 05:29 AM
nicely done

thomascheng
03-03-2019, 06:41 AM
Your broke the bird, what a waste ;)

Cageman
03-03-2019, 07:42 AM
Maybe? We can call it artifacts instead. :D




I understand what you’re after and also that this is a personal test project with time constrains. But if you would be able to see a copper statue wiggle like that it would only be if you film with a high speed camera, but I doubt a small statue like that would wiggle that much. I would buy the wiggle more if you did it in slow motion.

Yep. You are absolutely right... what would happen if the impact is hard enough, it would just fall over... probably not break at all, or just a very small peice from the impact area would fall off. But that is not as fun as making it go into pieces. :D

ianr
03-03-2019, 08:33 AM
is that falcon Maltese?

Ztreem
03-03-2019, 10:03 AM
Yep. You are absolutely right... what would happen if the impact is hard enough, it would just fall over... probably not break at all, or just a very small peice from the impact area would fall off. But that is not as fun as making it go into pieces. :D

True! :)

Ma3rk
03-03-2019, 03:40 PM
Just a doodle. LW2019.0.1 with db&w Tri-planar mapping node + Patina node and Bulled Physics. :)

https://www.youtube.com/watch?v=qR_n5pGE5zc

Enjoy! :)

//Cageman

Very clean looking. What sort of rendering times were you getting?

jwiede
03-03-2019, 04:11 PM
Yep. You are absolutely right... what would happen if the impact is hard enough, it would just fall over... probably not break at all, or just a very small peice from the impact area would fall off. But that is not as fun as making it go into pieces. :D

Yeah, if you really hit it hard enough to break off significant pieces, it'd probably just shatter altogether (and it'd take a lot more force than those balls moving at those rates could manage) -- and even if so, better chance the balls themselves would shatter or embed. Much more likely no damage/minor denting, or perhaps at most small cracking/deformation with that scale of projectile (even moving at barely eye-visible rates).

Physics aside, the surfacing did look nice, though, no question.

prometheus
03-04-2019, 12:46 PM
Thatīs a Griffin.
Nice copper look and patina, I agree with it being wiggling like a rubbery look though.
and some smaller parts are sliding too much in the background, and also the head is taking a bit of a slide in the end that doesnīt look right.

Cageman
03-04-2019, 02:20 PM
Very clean looking. What sort of rendering times were you getting?

Using a renderfarm of about 60 machines, it clocked at around 40s / frame. :)

Cageman
03-04-2019, 02:26 PM
and also the head is taking a bit of a slide in the end that doesnīt look right.

That is because the head landed on a piece that later turned on its side due to preasure.

As for "realism", who really cares if it looks cool? Artistic freedoom and all that. :D

EDIT: I just did a doodle on this... didn't try to make it look real, I just imagined the way I wanted it to look regarding the physics.

As I stated earlier, this would probably not happen to a copper statue of the size of 20 centimeters. :)

Cageman
03-04-2019, 02:32 PM
Also... someone wrote a reply saying that I nailed the physics...

I mean, depending on how you view a thing, it looks right, or it doesn't.

To me, it looks cool and fun; but not right. Because, reality is, kind of, boring. :) ESPECIALLY for a sunday afternoon doodle. :D

Ztreem
03-04-2019, 03:08 PM
For a doodle it is really good.
I think, just because it looks so good in regards to lights, surfaces and rendering you start to see the small things that doesn’t look as good. I was only nitpicking about the rubbery thing. It still looks cool and all but I still think that if you spend another sunday afternoon on this you can make it even better. Keep up the good stuff.

Cageman
03-04-2019, 03:33 PM
For a doodle it is really good.
I think, just because it looks so good in regards to lights, surfaces and rendering you start to see the small things that doesn’t look as good. I was only nitpicking about the rubbery thing. It still looks cool and all but I still think that if you spend another sunday afternoon on this you can make it even better. Keep up the good stuff.

Of course. But at the same time, it feels good doing a Doodle, and then just have it done. No matter if it is "wrong" or "right". It is just fun, you know. At least for me to do something quick and dirty. :)

Cageman
03-04-2019, 03:43 PM
is that falcon Maltese?

It is called a Griffin. :)

Ztreem
03-04-2019, 04:50 PM
Of course. But at the same time, it feels good doing a Doodle, and then just have it done. No matter if it is "wrong" or "right". It is just fun, you know. At least for me to do something quick and dirty. :)

Absolutly. I do the same thing. When you work with 3d, you do so much of long projects at work so its nice to do some quick and dirty stuff from time to time.

prometheus
03-04-2019, 04:51 PM
That is because the head landed on a piece that later turned on its side due to preasure.




artistic freedom yep, as long as you can choose from realistic and non realistic, sliding pieces as artistic choice..I am not sure about, itīs your call though I think it may be just that you are "doodling" and not
investing too much time on some areas.

the head landed on a piece ..yes I noticed that, there are still pieces that are sliding too much I reckon, especially piece you can see in the middle between the gap of the main part.
If itīs a question of friction, mass or simulation rate I do not know...I do know that it has always been hard to get it to stop jittering.


Been a while since I was doodling with bullet though.

Rayek
03-04-2019, 05:09 PM
While it looks nice for a doodle, the physics of the material seem to change properties drastically after the griffin statue falls over. The vibrating wing feels as if the material is made of a rubbery plastic, but when it falls over, the body suddenly inexplicably explodes into many small pieces, which looks as if the material is made out of extremely brittle sugar glass.

As for copper and metals responding to being hit by bullets: it depends. In any case, copper has both good ductility (it can be stretched into wires) and good malleability (it can also be rolled into sheets). It would never spontaneously act all brittle like in this physics doodle, and hence, the realism factor (IF this was meant to simulate how a copper statue responds to being hit by high velocity metal projectiles) is zero. It should respond more like a highly viscous fluid where it is hit. It certainly shouldn't break up. Perhaps one or two parts being torn off. Mostly just deformations, dents, and stuff. (Copper can be brittle if certain impurities are present, such as bismuth exceeding 0.005%, but even then would not explode into tiny bits).

After being hit by the first projectile, a copper statue would probably just fly off the table due to the first impact.


Then again, as was said, reality is often over-rated ;-P One merely has to watch those Marvel movies. Or the flying spontaneously exploding cars in 80s series and movies.

PS here is a nice video showing how copper bullets deform when they hit metals and other materials:

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

Cageman
03-05-2019, 02:53 AM
If itīs a question of friction, mass or simulation rate I do not know...I do know that it has always been hard to get it to stop jittering.


Been a while since I was doodling with bullet though.

If this would have been something that I got payed to do, I would have taken the sim into Chronosculpt and stabilized it. I just don't think for a doodle like this that it is important enough.

prometheus
03-05-2019, 06:53 AM
If this would have been something that I got payed to do, I would have taken the sim into Chronosculpt and stabilized it. I just don't think for a doodle like this that it is important enough.

absolutely...
only so much time when doodling..
and not trying to push it to the end.

just a bit of too much hazzle to go through chronosculp...at least for doodling...but for me..a certain free software yields much better result without
jittering and sliding...i was surprised and impressed by that...that leaves an impression of bullet not being good enough for ground collision at least.

Cageman
03-05-2019, 07:24 AM
absolutely...
that leaves an impression of bullet not being good enough for ground collision at least.

Bullet is standard in many packages. I mean, the sim was in realtime. I can probably up the settings and let it take a couple of seconds per frame instead of instant calculation.

Ztreem
03-05-2019, 08:04 AM
just a bit of too much hazzle to go through chronosculp...at least for doodling...but for me..a certain free software yields much better result without
jittering and sliding...i was surprised and impressed by that...that leaves an impression of bullet not being good enough for ground collision at least.

That software is also using Bullet as a majority of all DCC apps do. I think it's more about the implementation than the library itself.

prometheus
03-05-2019, 11:52 AM
That software is also using Bullet as a majority of all DCC apps do. I think it's more about the implementation than the library itself.

Probably, I just know that for basic cracks of pieces and let them fall and settle, I donīt see a single jittering or sliding anywhere near in terms of what Lightwave produces as jittering and sliding, and even if you do raise quality or simulation rate, there are still pieces not settling most of the case in Lightwave, and you may have to use chronosculpt and what that means in terms of time in exporting, adjust sending back etc, but maybe chronosculpt is a saviour for it, though it is manual eyeballing correctio.
With that other software it just seem to work splendid, with default sim settings, so maybe they could disect what it is if possible..since both use bullet.

Cageman
03-05-2019, 07:17 PM
Been a while since I was doodling with bullet though.

So... is Bullet much better in Blender compared to LightWave regarding jittering? I mean... you say it has been a while since you doodled with Bullet, after all.

prometheus
03-06-2019, 01:56 AM
So... is Bullet much better in Blender compared to LightWave regarding jittering? I mean... you say it has been a while since you doodled with Bullet, after all.

I think I tried bullet in 2018, tried it in 2015.3 at least...and as I can recall there where no changes in how bullet performs in that regards, 2019 and eventual improvements in collision accuracy..I donīt know yet.

As for much better less jittering in bullet in blender, in a scenecase of a simple cracked box falling and cracking with scattered pieces...yes in my opinion it is, no jittering at all in my experience...but that is for defaults, if you change properties it may behave differently, I havenīt done extensive test on it.

But the case is basic cracking of a box falling to the ground, and almost no matter what you do in Lightwave..the jittering is hard to stop and settle, while in blender..no problems.

I love blender fracture modifier as well, where you can crack pieces in realtime, without being destructive, and also apply subdiv modifier ( subdiv standard first, then subdiv catmull clark to acheive patches of broken stone) with smooth edges...there is nothing in lightwave that can do similar, you would have to use the crack it plugin to break pieces in such shapes (native wonīt do) and then perform modeling subdivision on it..and all of this is destructive and canīt be changed once comitted to, unlike blenders fracture modifier which can be changed and tweaked to get more fracture pieces or less, or change where it breaks non destructively.

Apart from that we have dust and debri and fluid paint directly from fracture collision parts.

Lightwave bullet has a lot of other features that are really sweet with bullet motors and bone dynamics..that blender doesnīt have.

Marander
03-06-2019, 02:58 AM
But the case is basic cracking of a box falling to the ground, and almost no matter what you do in Lightwave..the jittering is hard to stop and settle, while in blender..no problems.

I love blender fracture modifier as well, where you can crack pieces in realtime, without being destructive, and also apply subdiv modifier ( subdiv standard first, then subdiv catmull clark to acheive patches of broken stone) with smooth edges...there is nothing in lightwave that can do similar, you would have to use the crack it plugin to break pieces in such shapes (native wonīt do) and then perform modeling subdivision on it..and all of this is destructive and canīt be changed once comitted to, unlike blenders fracture modifier which can be changed and tweaked to get more fracture pieces or less, or change where it breaks non destructively.

Yes and that's how other 3D applications work as well except LightWave (as far as I know). Other apps allow to art direct excactly what you're after in real time (until a certain complexity of course). Simulations and Fracturing is no fun in LW and leads to issues seen in the video posted.

I agree, nice render quality but the rest no so much.

Ztreem
03-06-2019, 04:33 AM
Lightwave bullet has a lot of other features that are really sweet with bullet motors and bone dynamics..that blender doesnīt have.

Bone dynamics is not as easy to setup in Blender as in Lightwave but I think its possible, remember seeing something about that. Motors exist in Blender, look at this for an example.
https://www.youtube.com/watch?v=klLI5skuwj4

prometheus
03-06-2019, 06:23 AM
Bone dynamics is not as easy to setup in Blender as in Lightwave but I think its possible, remember seeing something about that. Motors exist in Blender, look at this for an example.
https://www.youtube.com/watch?v=klLI5skuwj4

ah..I wish the new bullet bone was even easier, just as easy as with IK booster, you just activate it, no manual setup.

Cageman
03-14-2019, 06:36 AM
Just a heads up about Bullet...

When you feed it very small scale objects (meassured in milimeters) it will crap out and the result will not be predictable at all (sliding pieces and wonky looking stuff). So... if you want to explode a small object like that, make it 100 or even 1000 times bigger, simulate it like that. Then, for rendering, scale it back, and speed up the playback of the alembic/mdd.

This can easily be reproduced in LW as well as Houdini... so I think Blender would behave the same as well.

prometheus
03-15-2019, 03:11 PM
Basic employ of a crack action...1m cube, see movie samples attached at the bottom of this post.
Basic settings by default from the two programs, no change in dynamics settings or parts/body settings.

Fracture done in blender fracture build, then just make the plane passive rigid and the cube a rigid body, then just play...runs smoothly and without any sliding and jittering and it yields a crack scattering that looks realistic.

Exported the cracked box and plane as obj and imported to Lightwave..made the plane static body and the box a part body..then just play, and it doesnīt scatter as naturally by default, and it of course shows unatural sliding behavior by default and it would jitter if I had the animation longer.

Also, the images showing subdiv modifier, is Realtime when changing fracturing parameters, and as a modifier non destructive, this kind of smoothing/ rounding the pieces to form like rock or building blocks isnīt possible with lightwave ..Unless using the old crackit plugin, which creates better cracked pieces that works with a subdivision operation (bullet fracture doesnīt hold up)...and the crackIt pluginis a bit slow, and once you have done that in Lightwave..it is finishied and canīt be changed, while the blender modifier allows for turning it on or off
in a blink of an eye, add multiply subdiv modifiers such as mixing catmull and simple and also various levels of the subdivs to acheive sharper or smoother splits on the pieces.

The middle subdiv modifier set to simple and raising subdiv levels will determine how sharp the split will be, while the last subdiv modifier set to catmull and raising subdiv level will only smooth out the pieces with denser mesh without really changing the main shape.

These subdiv modifiers to round/smooth the broken pieces, also works when running the simulation, though at this stage you may experience some jittering.

https://forums.newtek.com/attachment.php?attachmentid=144431&d=1552683801

I think this type of workflow needs to find itīs way in to the Layout section ASAP to keep Lightwave in the competition for VFX dynamics and fracture stuff, with a close connection to Particle driven controls( itīs there in the movies you know:) )

Movie samples attached, I donīt have the time to polish my voice today, and I am a bit tired..and I would have to configure some recording settings in my nvidia hardware game ready updates to record it, so only small sample vids here.

I forgot, you can of course enter Lightwave bullet properties ..increase collision margin a bit if using mesh mode, and change some other settings, such as changing friction with distance to object and much more, but it is
much more work to get decent results than what you do get with basic default in blender.

Greenlaw
03-15-2019, 03:21 PM
That's a fun test, Cageman! Nice render and very nice render times.

Glad to see you participating in the forums again. I always learn a lot from your works. :)

prometheus
03-15-2019, 03:50 PM
and a bit unfairly from my side, the default mode in blender Is Convex hull, and as such it yields less jitter, switching to mesh and blenders collision by default will explode it badly unless changing the settings, while Lightwave seem to be set as mesh by default.

Cageman
03-16-2019, 04:08 PM
Basic 1m cube split into 100 segments in LW works like a charm without any issues. That said, for the groundplane, make sure you have thicknes and just not a flat polygon. Makes a HUGE difference.

- - - Updated - - -


That's a fun test, Cageman! Nice render and very nice render times.

Glad to see you participating in the forums again. I always learn a lot from your works. :)

Thanks! What are you up to nowdays?

Greenlaw
03-17-2019, 01:52 AM
Thanks! What are you up to nowdays?

Thanks for asking!

I've been at DreamWorks Animation for the past four years, animating mainly for 2D cartoons and occasionally creating 3D animation and effects (using LightWave, AE and other programs.) You can check out my '2017' reel (http://littlegreendog.com/2017/02/20/d-r-greenlaw-dreamworks-animation-demo-reel-2017/) below to see a few examples. That reel is kind of 'old' by now but I'll have a new reel to post this year.

For personal projects, I always have something in the works...actually, too many somethings according to my wife. The Brudders project you advised me on years ago is never-ending, but we've completed other shorts since starting that one. I still intend to finish it one of these days (doggone it!) :)

When it comes to animated film projects, my daughter is way more prolific than I am. Here's her latest short film, created last summer when she was 11 years old: 'Class Act' by Sienna Kai Greenlaw (https://vimeo.com/292533023/2362886848)

prometheus
03-17-2019, 05:22 AM
Basic 1m cube split into 100 segments in LW works like a charm without any issues. That said, for the groundplane, make sure you have thicknes and just not a flat polygon. Makes a HUGE difference.

- - - Updated - - -



Thanks! What are you up to nowdays?


What do you mean without any issues? do you mean no jittering and sliding?

Yes..I know about the thickness, though as I recall...that did not make much of a difference in my test, but I may have to revisit again though to verify.

Anyway..the blender fracture build with realtime fracturing and adding smoothness in realtime.. is a bit more of what lightwave can do currently, so right now I prefer to crack the stuff in blender.
I may start to do more comprehensive test between the programs soon.

But the dynamics of parts excluded, blender offers interesting particle debri and fluid emission derived from parts impact and debri, for the fire and smoke engine, we donīt have those tools
in Lightwave and particle setup and fluid emission of smoke in those areas would involve so much more work if you want to acheive similar.

Missing in Lightwave..

TFD fluid weight emission/weigth paint emission.
TFD fluid PBR rendering of smoke and fire.
Non destructive fracturing (preferably almost realtime) in Layout.

Cageman
03-17-2019, 08:26 PM
Thanks for asking!

I've been at DreamWorks Animation for the past four years, animating mainly for 2D cartoons and occasionally creating 3D animation and effects (using LightWave, AE and other programs.) You can check out my '2017' reel (http://littlegreendog.com/2017/02/20/d-r-greenlaw-dreamworks-animation-demo-reel-2017/) below to see a few examples. That reel is kind of 'old' by now but I'll have a new reel to post this year.

For personal projects, I always have something in the works...actually, too many somethings according to my wife. The Brudders project you advised me on years ago is never-ending, but we've completed other shorts since starting that one. I still intend to finish it one of these days (doggone it!) :)

When it comes to animated film projects, my daughter is way more prolific than I am. Here's her latest short film, created last summer when she was 11 years old: 'Class Act' by Sienna Kai Greenlaw (https://vimeo.com/292533023/2362886848)

I remember you went to Dreamworks after RH had to close the doors. Nice to hear you are still there doning stuff with LW. :)

And wow... your daughter really has a talent! Thanks for sharing! It put a smile on my face! :)

- - - Updated - - -


What do you mean without any issues? do you mean no jittering and sliding?

Yes... works like a charm. No jitter, no sliding. Try make a 10mm box in Blender and fracture it and apply Bullet to it.

vncnt
03-18-2019, 12:43 PM
Thanks for asking!

I've been at DreamWorks Animation for the past four years, animating mainly for 2D cartoons and occasionally creating 3D animation and effects (using LightWave, AE and other programs.) You can check out my '2017' reel (http://littlegreendog.com/2017/02/20/d-r-greenlaw-dreamworks-animation-demo-reel-2017/) below to see a few examples. That reel is kind of 'old' by now but I'll have a new reel to post this year.

For personal projects, I always have something in the works...actually, too many somethings according to my wife. The Brudders project you advised me on years ago is never-ending, but we've completed other shorts since starting that one. I still intend to finish it one of these days (doggone it!) :)

When it comes to animated film projects, my daughter is way more prolific than I am. Here's her latest short film, created last summer when she was 11 years old: 'Class Act' by Sienna Kai Greenlaw (https://vimeo.com/292533023/2362886848)
Did you use LW for character animation? If the answer is yes, any thoughts on this?

Greenlaw
03-18-2019, 01:30 PM
I can't go into details about the studio's pipeline but it's probably okay to say I've used LightWave for some character work there. (I get asked to do a lot of different things and have a lot of different tools to work with.)

Personally speaking, I think LightWave is fine for character animation. Since the late nineties, I've used it to animate characters, creatures, and many other things for feature films, tv shows, commercials and video games. I use it almost exclusively for 3D in my personal projects.

LightWave does what I need, it's affordable, and I like using it. :)

vncnt
03-18-2019, 01:39 PM
Thanks Greenlaw.

Cageman
03-26-2019, 06:36 PM
Still waiting for the 10mm Cube with 300+ fragments simulated in Blender without any issues, prometheus.

prometheus
03-27-2019, 12:41 PM
Still waiting for the 10mm Cube with 300+ fragments simulated in Blender without any issues, prometheus.

Oh.. I donīt have the time currently..not the upcoming days, maybe..just maybe this weekend, but not for some kind of proof of how that turns out..I want to showcase
the realtime fracturing and subdiv modifiers on it within a simulation...maybe...just maybe compared to how fracture does it in Lightwave and at the same time we may be able
to compare how something around 300+ fragments are working.

Have to reboot the computer since I record with nvidia game ready options..and drivers needs to be updated otherwise the recording doesnīt work..but I am also testing the cloud stuff with the new volumetrics, and since I
can not save in discovery mode, I have locked my self up for some days exclusively to test that without rebooting the computer, thus no recording option.

prometheus
04-01-2019, 12:31 PM
Still waiting for the 10mm Cube with 300+ fragments simulated in Blender without any issues, prometheus.

Just upgraded to Lightwave 2019 after skipping 2018, I spend so much time with the volumetric Clouds and some changes in there along with VDB and Dpont upgrades sort of eliminated some frustration.

So just installed yesterday and havenīt had much fiddling with it, I created a stair step in Lightwave( still prefer to model hard stuff in there till I learn blender modeling more) I then sent it to blender (fracture build) and used a fracture of 1500 shards.
run the simulation with convex piece mode wich is default in blender, runs decently fast, then I sent the object to Lightwave 2019, made it a part body and changed to convex piece mode sinc that is not default and the result...

Blender Is simulating the smashing pieces much faster, and it doesnīt have to decompose, but after the initial decompose in Lightwave, having it falling to ground is quite a bit slower than blender, and hitting the floor and spreading the pieces is also slower than blender.

Now..I used a 5 meter stair, if you urge for trying it with 10mm to get faster speed? I might look in to that.
But..this is all with the default settings in each program, it may be or may not be that if really going through all settings, the simulation time may be more equal?

prometheus
04-01-2019, 12:43 PM
Tried 5000 pieces, and it doesnīt really affect the simulation speed that much in blender, Lightwave seem to be affected more in simulation speed than blender.

Cageman
04-07-2019, 07:23 PM
You are missing the point.

Scale has important results on the simulation.

10mm Cube. It has NOTHING to do with speed, all about Scale. Dynamic simulations have SCALE as a big input as in how well it will solve it.

So, can you simply do what I am asking?

I see this in Houdini as well as LW... when the scale of objects are small, the result of the Bullet simulation will have trouble resolving properly. No matter how much time it takes.

Cageman
04-07-2019, 07:25 PM
Tried 5000 pieces, and it doesnīt really affect the simulation speed that much in blender, Lightwave seem to be affected more in simulation speed than blender.

With a 1m cube, LW Bullet is realtime and doesn't cause any sliding...

What happens in Blender, when you do the same thing with a 10mm cube? That is the SIMPLE question I am asking. :)

Ztreem
04-08-2019, 12:11 AM
You are missing the point.

Scale has important results on the simulation.

10mm Cube. It has NOTHING to do with speed, all about Scale. Dynamic simulations have SCALE as a big input as in how well it will solve it.

So, can you simply do what I am asking?

I see this in Houdini as well as LW... when the scale of objects are small, the result of the Bullet simulation will have trouble resolving properly. No matter how much time it takes.

I would say without doing any tests, that the result will be similar in Blender as it’s bound to the Bullet library as all the others. Size has a major effect on physics calculations and especially Bullet where you should be in the 1-10m range to yield good results.

prometheus
04-08-2019, 12:04 PM
With a 1m cube, LW Bullet is realtime and doesn't cause any sliding...

What happens in Blender, when you do the same thing with a 10mm cube? That is the SIMPLE question I am asking. :)

As I said to quote myself...
"Now..I used a 5 meter stair, if you urge for trying it with 10mm to get faster speed? I might look in to that."

Itīs not a priority for me, Iīll do that later...or if you have the time, you do it and show the results here, and you will probably get the answer faster...or have you already tested and it all got wonky?

I am having 3D free time (no work at all with 3d) for some days now....doing other things.

prometheus
04-08-2019, 12:16 PM
With a 1m cube, LW Bullet is realtime and doesn't cause any sliding...

What happens in Blender, when you do the same thing with a 10mm cube? That is the SIMPLE question I am asking. :)

Also ..I am not getting what you are trying to get at?...if we disregard the sliding,
with 1 m cube..as I refered to with 5000 fractured shards, Lightwave isnīt realtime, blender is much faster, default settings and at 1 meter scale of the cube..so what gave you that idea? not on my Asus G20CB nvidia GTX
32 gig ram i7 6700cpu

Couldnīt resist..
You screwed up my 3D free day :)
In blender
10 mm cube...well..it falls right throught the plane, so adjustment is needed in collision margin, changed collision margin and friction as well for the cube.
Best results is using the standard 1 meter cube it seems, seems the default settings for it is based on 1 m, and changing to smaller size requires other settings.
And yes..getting pieces to settle and not jitter with a 10mm cube in blender is hard, compared to a 1m cube, that is with convex hull.

Cageman
04-08-2019, 03:42 PM
Also ..I am not getting what you are trying to get at?...if we disregard the sliding,
with 1 m cube..as I refered to with 5000 fractured shards, Lightwave isnīt realtime, blender is much faster, default settings and at 1 meter scale of the cube..so what gave you that idea? not on my Asus G20CB nvidia GTX
32 gig ram i7 6700cpu

Couldnīt resist..
You screwed up my 3D free day :)
In blender
10 mm cube...well..it falls right throught the plane, so adjustment is needed in collision margin, changed collision margin and friction as well for the cube.
Best results is using the standard 1 meter cube it seems, seems the default settings for it is based on 1 m, and changing to smaller size requires other settings.
And yes..getting pieces to settle and not jitter with a 10mm cube in blender is hard, compared to a 1m cube, that is with convex hull.

Thank you! :) That was all I asked.

erikals
04-08-2019, 03:54 PM
Very clean looking. What sort of rendering times were you getting?

Using a renderfarm of about 60 machines, it clocked at around 40s / frame. :)
dang, no criticism of your work, but isn't that extremely slow ?
i mean... 60 machines > 40s/frame ? for a very simple scene.

https://i.imgur.com/zHYIj7K.gif

Ztreem
04-08-2019, 04:24 PM
If you have 60 machines rendering at 40 sec a frame, that is equal to about 0,6 sec per frame on a single machine. That is quite fast.

erikals
04-08-2019, 04:27 PM
oh, was it a total?

then yeah, that changes the game completely.

https://i.imgur.com/xKAj3vM.gif

Greenlaw
04-08-2019, 04:33 PM
dang, no criticism of your work, but isn't that extremely slow ?
i mean... 60 machines > 40s/frame ? for a very simple scene.

Wait, 40 seconds per frame for a photoreal rendered HD scene with lots of little fragments to anti-alias is...slow????

Render farm or not, I thought that sounded pretty good.

Maybe I need a better computer. :eek:

erikals
04-08-2019, 04:57 PM
Wait, 40 seconds per frame for a photoreal rendered HD scene with lots of little fragments to anti-alias is...slow????
yes, that is slow, for 60 machines, if that is the total, and not per machine.

Octane could close to pull that off using 1 (expensive) graphics card.

Greenlaw
04-08-2019, 05:10 PM
Ok, maybe I'm misunderstanding something here. Are we talking about 60 computers working on a single distributed frame, or 60 computers each rendering a single frame in the sequence?

If it's the first then, yeah, that sounds slow, but it also sounds like an unusual way to render animation on a farm. Distributed rendering like this is typically used for rendering very large still images for print production.

The latter is what I'm used to seeing in animation production and 40 secs per frame is pretty darn good. Also, consider some of the time is spent just loading the data across the network to each render node, at least for the first batch of frames.

Cageman
04-08-2019, 05:33 PM
Hold your horses. :D

I need to re-render this project, because I lost ALL my data and actually also why I said "40s/frame".

It is about CPU vs GPU and the cost, not just in electricity but also Green Power and licensing. CPUs have gone a long way being energy effective, GPUs... not so much. BUT, 40s/f shouldn't be too far from the target. And the reason I am interested in this is because, where I work, I have the _potential_ access of over 400 CPUs.

But... I really need to re-render this because I lost all data due to a re-install of my computer at work. :(

And, instead of me driving a sort of agenda (vouching for CPU rendering because of cost and speed) I will publish all data in this thread and you can do the math yourself.

Also... for a person working at home, freelancing.... GPU or Realtime are the only ways... really that's it, to get your rendertimes down, unless you go for the threadrippers AMD produces. Or, use a renderfarm service.

That said, if you work at a place like Weta... they have over 40k CPUs.... :D

Stay tuned for more accurate data on rendertimes...

erikals
04-08-2019, 07:40 PM
Hold your horses. :D

I need to re-render this project, because I lost ALL my data and actually also why I said "40s/frame".

It is about CPU vs GPU and the cost, not just in electricity but also Green Power and licensing. CPUs have gone a long way being energy effective, GPUs... not so much. BUT, 40s/f shouldn't be too far from the target. And the reason I am interested in this is because, where I work, I have the _potential_ access of over 400 CPUs.

But... I really need to re-render this because I lost all data due to a re-install of my computer at work. :(

And, instead of me driving a sort of agenda (vouching for CPU rendering because of cost and speed) I will publish all data in this thread and you can do the math yourself.

Also... for a person working at home, freelancing.... GPU or Realtime are the only ways... really that's it, to get your rendertimes down, unless you go for the threadrippers AMD produces. Or, use a renderfarm service.

That said, if you work at a place like Weta... they have over 40k CPUs.... :D

Stay tuned for more accurate data on rendertimes...


from what i've seen GPU (for freelancing) is still the way to go, even though Threadrippers are closing in.
the new Nvidia cards save a lot of power, so they are less expensive.

here for example, a comparison in regards to CPU vs GPU
https://www.youtube.com/watch?v=p3I1gsVmyvg

but hey, if you got access to 400 CPUs (no jealousy here, none at all... https://i.imgur.com/PcBTNPH.gif then that is a good way to go. (especially for fluids simulations etc)

---------------------

an older thread, but worth a read >
https://forums.newtek.com/showthread.php?158039-Choose-your-weapon-GPU-or-CPU-rendering

rustythe1
04-09-2019, 05:07 AM
im not sure what 7980xe all these people keep using for these videos, as all these videos are just wrong, the intel has a lower power consumption than the threadripper its only 165 (I can even see it on my smart meter), it renders almost twice as fast, so from what people say in these videos, my intel, and many dual xeons are actually better than having GPU, on top of that intel have just released 56 thread xeons which will be used for the base of the new XE later this year. thread counts are increasing at a crazy rate, but gpu counts are not increasing as fast, not much change between a 1080 and 2080, it was more about the RTX, I don't think there is anything between them when your comparing both high end solutions, so its down to what you gain or lose in terms of features and quality on the renderers rather than speed and performance. I still have never seen a video of the same scene showing a "single" GPU rendering against a CPU side by side to see which finishes first,

gar26lw
04-09-2019, 06:10 AM
what speeds are you getting? whats the price of the cpu system?

rustythe1
04-09-2019, 07:08 AM
speeds are just relative, best example is marble scene is down to about 11 mins, I built the system for around Ģ4000, that includes 2x 1070ti cards as I run 3, 4k screens, water cooling, nice glass case, hefty psu, WD black drives, solid state drive, at the moment it would cost me around Ģ3000 just to get 2x top 2080ti cards as we get stung for GPU in uk (its actually cheaper to order in US and pay import duties through E-bay), a single Titan RTX card is currently Ģ2500,

erikals
04-09-2019, 10:57 AM
7980xe is $2000 over here, just for the cpu
1950x is $600 over here, just for the cpu

Qexit
04-09-2019, 11:21 AM
7980xe is $2000 over here, just for the cpu
1950x is $600 over here, just for the cpuQuick search here gives:

7980xe is Ģ1800 , just for the cpu
1950x is Ģ580 , just for the cpu

So neither will be on my shopping list in the near future :D

erikals
04-09-2019, 11:30 AM
if you go the Threadripper route, go with 1920x
close to same speed, but cheaper.

likewise, if you go the Ryzen route, choose the 1700x. (great speed, low cost)

Rayek
04-09-2019, 11:47 AM
Quick search here gives:

7980xe is Ģ1800 , just for the cpu
1950x is Ģ580 , just for the cpu

So neither will be on my shopping list in the near future :D

You Brits have it tough! Even here in Canada the 1950 is ~800 can at the moment.

Aren't Rzyen 3 and zen 2 around the corner? I intend to wait for the next gen, and then build a new rig around Christmas time (hoping for a good deal).