LW B747 Benchmark (LW 2020+)

Cageman

Almost newbie
Very interesting, Cageman.

So I tried changing Polygon Intersection Mode from Watertight to Fastest to see the impact of the setting in a fairly slower machine.
Here it is:

PIM set to Fastest: LW_B747_Bench.lws 20m 36s
PIM set to Watertight: LW_B747_Bench.lws 22m 11s

So indeed that was quite substantial.

Renders done with LW 2020.0.1 on an i7-5960X 8C/16T OC to 4.2GHz

View attachment 148296
View attachment 148297

Wow... that is way more than my machine shows... Which scene did you render? The regular or the Volumetric one? Almost 2 minutes difference with that simple setting... wow. :)
 

Lewis

Member # 777
Thanks for test guys, I've not wanted to mess with polygon intersection 'coz as Cageman mentioned my settings are default one and i was not sure does LW save that with Scene (LWS) or Layout configs so it would be "unfair" advantage if i change it on my layout and then user loads this in vanilla Layout settings :). But anyway the point was not to absolutely make fastest scene but benchmark usable on fast and slower machines. also scene is not up to final quality, there is still way too much noise to be cleaned but that would make it too long on slower machines so i opted for some middle ground between quality and speed :).
 

Cageman

Almost newbie
Thanks for test guys, I've not wanted to mess with polygon intersection 'coz as Cageman mentioned my settings are default one and i was not sure does LW save that with Scene (LWS) or Layout configs so it would be "unfair" advantage if i change it on my layout and then user loads this in vanilla Layout settings :). But anyway the point was not to absolutely make fastest scene but benchmark usable on fast and slower machines. also scene is not up to final quality, there is still way too much noise to be cleaned but that would make it too long on slower machines so i opted for some middle ground between quality and speed :).

Well... when you start LW with an empty scene, Polygon Intersection Mode is by default Watertight.

We always change this to Fastest, because it has, in the past help shave off 1-2 minutes / frame on the renderfarm... and it is a setting that is saved with the scenefile.

- - - Updated - - -

Thanks for test guys, I've not wanted to mess with polygon intersection 'coz as Cageman mentioned my settings are default one and i was not sure does LW save that with Scene (LWS) or Layout configs so it would be "unfair" advantage if i change it on my layout and then user loads this in vanilla Layout settings :). But anyway the point was not to absolutely make fastest scene but benchmark usable on fast and slower machines. also scene is not up to final quality, there is still way too much noise to be cleaned but that would make it too long on slower machines so i opted for some middle ground between quality and speed :).

Well... when you start LW with an empty scene, Polygon Intersection Mode is by default Watertight.

We always change this to Fastest, because it has, in the past help shave off 1-2 minutes / frame on the renderfarm... and it is a setting that is saved with the scenefile.
 

allabulle

Active member
Well, the scenes aren't certainly about how fast can LightWave be, but how to compare our computers while using a standarized LightWave scene with the new renderer. So yes, I think I am with Lewis here: not about optimizing performance and or quality. It's not about that, here.

What transpired, though, due to Cageman's experimentation, is that a significant gain was to be obtained by changing the Polygon Intersection Mode. And his machine crunches scenes like the one at hand like nothing. So I got tickled and wanted to try out how far that would go on an already slower machine, like mine. I thought it was an interesting find, albeit completely unrelated to the intent of the original proposal.

It's not the point if we can make this particular scene faster or cleaner, as stated. And yet, finding how to make LightWave faster in general in probably most scenes as a by-product of the experimentation of these scenes kindly provided by Lewis was not irrelevant to the LightWave community as a whole. Or so I presume.

Now I'll quit wasting more and more of your time, fellow LightWavers. Promise.

And thanks again, Lewis.
 

Sure... but since this is a benchmark scene, I am fairly sure that those using Blender, will try to optimize the **** out of it, while we, LW users are being handcuffed behind our backs because it is a "benchmark scene" and you are not allowed to touch it in any form, in order to make it render way faster.

What we will get is a banana vs apple comparison in the end, unless we are allowed to optimize the scene, right?

not saying it is fair, but when the other option is 10x faster, on a simple test, then it should curve some eyebrows,
not only that, motion blur is soon to receive an update making it 10x faster also.

agree, difficult to compare, but no, i did make some tweaks to this scene in LW, and it doesn't come anything to close to the speed of App [X]

so more of a watermelon vs an apple.


 
Last edited:

allabulle

Active member
If we find we can do something similar in some other app, that would be fine, but it wouldn't be that relevant to the ones talking about how to use this one. Wouldn't it? And hardly anything at all to do with the topic at hand here in this thread.

What I mean, dear erikals, and with all due respect, is that your derailing of the thread is even worse than mine. At least mine had something to do, tangentially, to the use of LightWave and the topic. More or less. ;-P
 

well, we closed it, but then someone brought it up again.  


At least mine had something to do, tangentially, to the use of LightWave and the topic. More or less. ;-P

my message wasn't about App [X] being better, rather it was about App LightWave being really-really slow. and thus it -absolutely- has something to do with LW.

i'm sure several other Wavers will agree with me that it is somewhat relevant, others, like yourself of course, will not.

anyways, i'll leave this thread/subject alone for some time.
 
Last edited:

gar26lw

New member
What transpired, though, due to Cageman's experimentation, is that a significant gain was to be obtained by changing the Polygon Intersection Mode. A

I already did that and got slapped :p

- - - Updated - - -




not saying it is fair, but when the other option is 10x faster, on a simple test, then it should curve some eyebrows,
not only that, motion blur is soon to receive an update making it 10x faster also.

agree, difficult to compare, but no, i did make some tweaks to this scene in LW, and it doesn't come anything to close to the speed of App [X]

so more of a watermelon vs an apple.



yeah, important comparison.
 

UnCommonGrafx

Wandering about
I applaud the zeal.
Cageman brought up a point, which I think is paramount to perception and performance: how vanilla lightwave is setup is less than ideal.
To see others point to this as the reason to not absorb this into the conversation; to not encourage the ubiquitous 'they' to update the vanilla settings for better performance; to bat away the opportunities to learn, in building a better startup setting and benchmark scene in how better to use the rendering engine, seems incoherent in the moment.

I understand the initial thrust of Lewis' post. Lewis does high-quality work. Others who have commented, as well. 2020, with it's revised engine, continues to throw curves on its learning.

I wonder: why can't the benchmark serve a functional purpose, like a better community startup scene/settings, that perhaps is respected enough to have the devvers make it the de facto startup scene?
 

allabulle

Active member
Let me see: So Lewis can't have a thread to benchmark our computers using his scene because we have to discuss some other important matters, here and now. Nah, come on, gentlemen.
We may be better off starting a thread with that purpose. By all means, let's do that. The ones concerned with such matters the most could easily start that conversation. A conversation worth having, I agree. But it would be nice if this thread was to be kept in topic. As all threads roughly should, I'm tempted to say.

And now I'm probably the one derailing it the most. We can't win, can we? :)
 

Cageman

Almost newbie
make sure to use the nvidia tool that allows for open gl acceleration under remote desktop

I thought this only affected non Quadro-cards? With the new drivers Nvidia released, all cards should be equal regarding RDP? I got a cheap Quadro card for it just because I knew that I will not be abandoning the current workstation anytime soon. :)
 

Cageman

Almost newbie
Well, the scenes aren't certainly about how fast can LightWave be, but how to compare our computers while using a standarized LightWave scene with the new renderer. So yes, I think I am with Lewis here: not about optimizing performance and or quality. It's not about that, here.

What transpired, though, due to Cageman's experimentation, is that a significant gain was to be obtained by changing the Polygon Intersection Mode. And his machine crunches scenes like the one at hand like nothing. So I got tickled and wanted to try out how far that would go on an already slower machine, like mine. I thought it was an interesting find, albeit completely unrelated to the intent of the original proposal.

It's not the point if we can make this particular scene faster or cleaner, as stated. And yet, finding how to make LightWave faster in general in probably most scenes as a by-product of the experimentation of these scenes kindly provided by Lewis was not irrelevant to the LightWave community as a whole. Or so I presume.

Now I'll quit wasting more and more of your time, fellow LightWavers. Promise.

And thanks again, Lewis.

Yeah.... I didn't mean to derail this thread into "how to optimize".
 

jimzombie

Member
I just put my new rig together a couple of days ago and wanted to do a benchmark and thought I'd start here.

old rig I7-4770:
49m 16s

New rig 3900X
standard: 12m 19s
volumetric: 40m 12s

I'm pretty happy with the improvement

lw_benchmark_12m19s.PNGlw_benchmark_vol_40m12s.PNG
 
Top Bottom