Render Tests

slacer

Member
You would need to have the scenes and settings which I could not find available for download. He used some scenes and modified them for these tests.
Even if you were able to recreate the interior scene, you would still need to know more about his hardware used for this "benchmark" test.
Otherwise you would compare your Threadripper against his Pentium Core Duo:ROFLMAO:. Your LW setup would be faster not because of a faster renderer...
 

UnCommonGrafx

Wandering about
Build one of the scenes and give it a go. Have you ever rendered with Blender that this kind of test will make a difference to you, for your work with Lightwave?
You would need mapped fbxs to keep the textures. Or all image-mapped stuff.
The lesson I've learned is that if you are invested in Octane, you got the goods - from there it's about your video card. And really, that's about the artists exploiting the most from the cards. Enterprise use of renderers is not what I think you are speaking to. Meaning that you aren't worried about a render farm.
If a render farm is in the picture, then working the same or similar chip as the farm keeps you in the ballpark of how long your renders will take.

You've got great hardware in your signature. An Octane user, I bet. That's not really LW. The LW renderer is dated. I would be curious as to a cpu render with it compared to a cpu renderer of a higher quality. Like Arnold or c4ds redshift. I would never exercise the patience to do such a test. lol
Most of those renderers are gpu-based. LW can't be included based on the native renderer if that's the criteria. Thus, the comparison would be somewhat unfair.
Robert
 

prometheus

REBORN
Informative for blender users mostly, but says nothing about any comparison with LW native render engine, it would however most likely be so slow that it won´t make it in to the chart at all.
As for octane for Lightwave, different matter.

Personally, it gave me some insight on what to look in to more, and octane for blender is not it for me, considering the horror of installing the license, doesn´t access blenders or lightwaves procedurals fully.
So it´s foremost e-cycles that takes the number one spot currently of interest but that will have to wait till cycleX has implemented volumetrics and a comparison of that against e-cycles.
Vray for blender, perhaps..need to look in to that as well..but I think it´s also a considerably more costly alternative.

There´s too much cycling going around, even though re-cycling is good, but they should just drop the standard cycles and focus on the next generations, perhaps a completely new name to avoid confusion between cyclesX and e-cycles.
 

UnCommonGrafx

Wandering about
That's why my questions. The speed of LW is the same as it has been. And I know of no one bragging about the speed of their cpu rig in rendering.
Like with Mr. Schmidt's hardware. I would be curious as to what it does under e-cycles or k-cycles. If it's all about speed, of course. Glass inside glass in under 2 seconds always makes for cool images, instant gratification and all. That's most gpu renderers, though.
Again, for LW comparisons, you have to do it against another cpu renderer. Fewer are trying to power through on the cpu for rendering tasks nowadays. LW is sorely left behind in these types of comparisons. Octane notwithstanding.
 
Unfortunately, these tests only say something meaningful to a limited extent, as Andrew himself says in the video (even if he then of course wants to put it into more relative terms).
Much less can we say about how Lightwave would compare. This is all just speculation. And by the way the Lightwave renderer is not outdated, in fact it is one of the newer parts of Lightwave.

But what is very interesting in the video is the comparison between GPU and CPU renderer.
If you take a closer look at Vray's render times you can compare the cost of GPU vs CPU. Vray is well suited for this because I assume that the render and material settings here are indeed the same in both tests. Besides, Vray is the overall winner.

The GPU costs about 4 times as much as the CPU (at least the cheapest graphic card with this chip set). I do not calculate that to render with the GPU of course you also need a CPU and vice versa.
In the all-rounder test, the GPU computes only 2.7 times faster with secondary solver and also only 3.86 times faster with brute force, but costs four times as much.

It shows that it can be worthwhile to take a closer look at this topic, even if everyone always sees the GPU as the better solution.

ciao
Thomas
 

UnCommonGrafx

Wandering about
Indeed, Thomas, you make great points. It alludes to my point about render farm usage: cpus are economically better when a lot of images are being rendered.
Threadrippers are getting the closest. The M1 chip might be of further interest once things are optimized for it. I wonder at what point it will be acknowledged the age of LW's renderer compared to those that get upgrades.
 

Rayek

Active member
I saw this video over on Blender Guru and it made me wonder how Lightwave would perform using the same scenes and settings.

Render engine tests on Blender Guru

I've done some tests in the past to compare Lightwave's 2019/2020 renderer with plain Cycles and eCycles. While the CPU render time compared very favourably (LW was faster than CPU Cycles), there was just no comparison using GPU (e)Cycles. It is a crying shame all development on Lightwave was stopped before its render engine was adapted in a GPU version, because I think it would have been F.A.S.T.

But now we will never know.

Andrew's testing was a bit surprising to me: it seems Octane's supposedly faster GPU rendering compared to the competition is just that: hot air bellowed out by its users and the developers.

But what really stands out is how difficult it is to compare all these different render engines. And how an expert may reduce render times dramatically, as well as how dependent results are on the scene contents and materials. Add options like denoisers, and no sane comparison is possible ;-P

As such the saying goes: there is no free lunch...

And with (semi) real-time render engines like Unreal, Marmorset, and Eevee (which will receive raytracing out some point), things become even more confused!

Good times.
 

prometheus

REBORN
I've done some tests in the past to compare Lightwave's 2019/2020 renderer with plain Cycles and eCycles. While the CPU render time compared very favourably (LW was faster than CPU Cycles), there was just no comparison using GPU (e)Cycles. It is a crying shame all development on Lightwave was stopped before its render engine was adapted in a GPU version, because I think it would have been F.A.S.T.

But now we will never know.

Andrew's testing was a bit surprising to me: it seems Octane's supposedly faster GPU rendering compared to the competition is just that: hot air bellowed out by its users and the developers.

But what really stands out is how difficult it is to compare all these different render engines. And how an expert may reduce render times dramatically, as well as how dependent results are on the scene contents and materials. Add options like denoisers, and no sane comparison is possible ;-P

As such the saying goes: there is no free lunch...

And with (semi) real-time render engines like Unreal, Marmorset, and Eevee (which will receive raytracing out some point), things become even more confused!

Good times.

Hard to say.

If I look at cycles hair vs lw hair, even cycles cpu rendering of hair is faster than lw hair cpu rendering as I experienced it with my slow cpu system, but cpu rendering is almost never anything I use with cycles anyway. (lw 2019.1.5)

So I guess you didn´t to a full test of the hair rendering and compared then if you felt lw CPU was faster than cycles CPU?
 

Rayek

Active member
Hard to say.

If I look at cycles hair vs lw hair, even cycles cpu rendering of hair is faster than lw hair cpu rendering as I experienced it with my slow cpu system, but cpu rendering is almost never anything I use with cycles anyway. (lw 2019.1.5)

So I guess you didn´t to a full test of the hair rendering and compared then if you felt lw CPU was faster than cycles CPU?
Actually, you are correct. I just tested it once more on my new system. Not sure if my memory failed me here, but Cycles CPU is much faster than LW CPU in the test scene that I used. Based on that 747 test scene - no hair used.

As for GPU rendering: eCycles with my older GTX 1080 is more than 5 times faster in that scene even on a AMD3900X.

But again, apples and pears.
 

prometheus

REBORN
I´ts a bit tricky to compare volumetric items though with cycles CPU and LW CPU, Usually I try to avoid any GI on volumetrics in lw...as long as it is CPU rendering only available and octane isn´t an option, while I always use GI and multiple volume bounce scatter in cycles, I am still unsure if it even Ever does any multiple volume bounce scattering inside the volumetrics in Lightwave, or just boost the general GI somehow.

Been looking at a cloud setup from some blender user who made some nodal fractal setup for that, and trying to render that, compared to a similar setup in lightwave, but as mentioned it seems to be impossible to make any good comparison, firstly I would need to know if lw really performs a multiple scattering in the volumes or if it´s something else, secondly the fractals themself takes different time to render depending on what fractal you use, third, you would need to set up exact the same size of the volume item and the same step rate, and the system handles it a bit differently.

The sample of the cloud noise volume item provided from a blender user, was a mess of nodes, I removed many of them and that boosted the speed up significantly, that is why I stress on blender needs more true fractals than a huge mess of nodes linked and having to be evaluated slowly, for lightwave it can be enough with one or two three nodes since many of the textures with rman collection as extension does a pretty good job, on the other hand..you are stuck with no true multiple scattering it seems and lack of GPU power.

I reacted to one thing though in the video comparison, blender guru states a check with environment volumetric , but that it´s not the true environment volume scatter he uses for that, but just a box with volumetric items, it would be like if we should compare a box volume item in lightwave and discard the Global scattering.
To me the "global" scattering you access in the world manu in blender actually seems worse to render than lightwaves global scattering, so there is one potentional item that May ..or may not be faster to render in lightwave, and that is regardless if I use GPU or CPU in blender, but I would need further testing of it all.

e-cycles and cyclesX is what seems to be most promising right now, but I need to await for the implementation of volumetric handling in cyclesX before I can decide to complement anything.
Wouldn´t mind seeing a GPU e-cycles, or cyclesX alternative within lightwave though, but depends on cost and also very important, if it can access all the lw textures natively as well as what Denis P is providing.

some fellow seem to be working on it, at least cycles, perhaps e-cycles, but I reckon for it to really come true, there needs to be some sort of team behind it, and not a single person ..at least if it´s not from the authors of it all.
 
Last edited:
Hard to say.

If I look at cycles hair vs lw hair, even cycles cpu rendering of hair is faster than lw hair cpu rendering as I experienced it with my slow cpu system, but cpu rendering is almost never anything I use with cycles anyway. (lw 2019.1.5)

So I guess you didn´t to a full test of the hair rendering and compared then if you felt lw CPU was faster than cycles CPU?

Hi,
have you compared this with the current hair materials (LW 2020)? Especially AFC Hair shader is much faster than the old hair material. Without transmission up to 15 times faster, with transmission still 5 times faster.

ciao
Thomas
 

prometheus

REBORN
Hi,
have you compared this with the current hair materials (LW 2020)? Especially AFC Hair shader is much faster than the old hair material. Without transmission up to 15 times faster, with transmission still 5 times faster.

ciao
Thomas

No..I haven´t, don´t have 2020 and I didn´t find it attractive at all to even bother downloading the demo.
transmission is always slowing down hair render, also in blender.
But there are other things apart from just render speed in the blender hairsystem that I find way more effective and attractive than LW hair.
Didn´t like the marketing images of the monkey they used for showcasing the new hair materials in Lightwave 2020, they should have worked on or picked something better.

But should we compare 2020 speed vs 2019 or how much 2020 has catched up against blender, perhaps 2020 CPUis more unpar with blenders CPU cycles, but GPU?
someone else have to check that though.

I would think if you really are in to messing around with hair in Lightwave, octane seems to be a must though.

Older blender 2.73...asus G20B rog, 1080 GTX GPU..( I need a new machine within 2-5 years I think)

 
Last edited:

RPSchmidt

Member
Build one of the scenes and give it a go. Have you ever rendered with Blender that this kind of test will make a difference to you, for your work with Lightwave?
You would need mapped fbxs to keep the textures. Or all image-mapped stuff.
Actually, in my day job, I ended up having to default to Blender back in September of last year (after using Lightwave over the past eleven years) due to IT purchasing / licensing snafus. I had switched divisions and they didn't have any Lightwave licenses available. So far, budgeting hasn't allowed for the purchase.

At the same time I was also unintentionally stuck with a standard workstation computer. Another IT snafu.

On a heavy scene created in Blender, both Cycles CPU and Cycles GPU rendered the same frame in essentially the same amount of time (within seconds of each other); the single Quadro M4000 GPU just didn't have the oomph to render faster than the dual CPUs in the system.

But also, I was just interested in seeing how Lightwave would compare with the same scenes against the other CPU renderers in his test setup.

The lesson I've learned is that if you are invested in Octane, you got the goods - from there it's about your video card. And really, that's about the artists exploiting the most from the cards. Enterprise use of renderers is not what I think you are speaking to. Meaning that you aren't worried about a render farm.
If a render farm is in the picture, then working the same or similar chip as the farm keeps you in the ballpark of how long your renders will take.

You've got great hardware in your signature. An Octane user, I bet.

I originally bought my personal hardware with an eye on investing in Octane... But then I didn't.

It's a combination of their licensing set-up, online vs. offline dongle (dongle?), and cost that isn't particularly appealing.

It's still a consideration, but honestly, I'm hoping that a more reasonable solution will come along.
 

prometheus

REBORN
I originally bought my personal hardware with an eye on investing in Octane... But then I didn't.

It's a combination of their licensing set-up, online vs. offline dongle (dongle?), and cost that isn't particularly appealing.

It's still a consideration, but honestly, I'm hoping that a more reasonable solution will come along.
Strange that you can have cpu and gpu renderin almost at the same speed, must have been a much higher focus on the cpu power on that machine.

as for the rest, not only licensing setup( which didn´t work for me and I got so tired of bothering with it despite them trying to help me by mail to get it working)
It seems it lacks access to many lw fractals like rman collection, and uses much of it´s own noise textures, so that is a no no for me as well.
 

RPSchmidt

Member
Strange that you can have cpu and gpu renderin almost at the same speed, must have been a much higher focus on the cpu power on that machine.

as for the rest, not only licensing setup( which didn´t work for me and I got so tired of bothering with it despite them trying to help me by mail to get it working)
It seems it lacks access to many lw fractals like rman collection, and uses much of it´s own noise textures, so that is a no no for me as well.

I thought it was strange as well, but that's the way it shook out.

It was a battle between a single Quadro M4000 card (older workstation card) and dual Xeon E5-2680 v4 cards (28 cores @2.4ghz).

If I had a Quadro K or RTX card, I am fairly certain that would have smoked the dual Xeons in a CPU vs GPU render test.
 

prometheus

REBORN
I thought it was strange as well, but that's the way it shook out.

It was a battle between a single Quadro M4000 card (older workstation card) and dual Xeon E5-2680 v4 cards (28 cores @2.4ghz).

If I had a Quadro K or RTX card, I am fairly certain that would have smoked the dual Xeons in a CPU vs GPU render test.
If you had my system, the results wouldn´t be the same..Drasticly different in speed with GPU.
quadro cards of older models I don´t think have much cuda cores, so probably why you don´t get a ratio result other than quite equal.
my system (need a new one soon)
Asus G20CB CPU i7-6700 3.40 GGz
32 RAM
GPU ..
The GTX 1080 comes with 8GB of GDDR5X memory, with a 10Gbps memory speed, 256-bit memory interface and a memory bandwidth of 320GB/sec. The base clock is 1,607 MHz, with a boost of 1,733 MHz, and it comes with 2560 Cuda cores. Overall, the GTX 1080 is capable of 9 teraflops of computing power
 

UnCommonGrafx

Wandering about
That sounds familiar. hehe. I didn't have the licensing snafu but a teaching snafu: couldn't tout LW as a way to earn a living into the future as an independent. Had to move to something else and this was the quickest, fastest, least painless method.
That computer sounds horrible!
Yeah, I couldn't see going Octane once I had both it and cycles at my access. While I think Octane renders beautifully, I've been able to get cycles (e-cycles, k-cycles, cycles-x) to render as beautifully as I think Octane looks. And quite quickly.

How has your blender time gone? Compared with your LW rendering, of course?
Actually, in my day job, I ended up having to default to Blender back in September of last year (after using Lightwave over the past eleven years) due to IT purchasing / licensing snafus. I had switched divisions and they didn't have any Lightwave licenses available. So far, budgeting hasn't allowed for the purchase.

At the same time I was also unintentionally stuck with a standard workstation computer. Another IT snafu.

On a heavy scene created in Blender, both Cycles CPU and Cycles GPU rendered the same frame in essentially the same amount of time (within seconds of each other); the single Quadro M4000 GPU just didn't have the oomph to render faster than the dual CPUs in the system.

But also, I was just interested in seeing how Lightwave would compare with the same scenes against the other CPU renderers in his test setup.



I originally bought my personal hardware with an eye on investing in Octane... But then I didn't.

It's a combination of their licensing set-up, online vs. offline dongle (dongle?), and cost that isn't particularly appealing.

It's still a consideration, but honestly, I'm hoping that a more reasonable solution will come along.
 

prometheus

REBORN
That sounds familiar. hehe. I didn't have the licensing snafu but a teaching snafu: couldn't tout LW as a way to earn a living into the future as an independent. Had to move to something else and this was the quickest, fastest, least painless method.
That computer sounds horrible!
Yeah, I couldn't see going Octane once I had both it and cycles at my access. While I think Octane renders beautifully, I've been able to get cycles (e-cycles, k-cycles, cycles-x) to render as beautifully as I think Octane looks. And quite quickly.

How has your blender time gone? Compared with your LW rendering, of course?

I haven´t had much time for anything 3D this summer, just sporadicly testing things, so not even in blender either, and I can´t judge spent time in blender VS lightwave properly.
Probably blender but it´s mostly fluid smoke sims.

I expect to jump on to more 3D late autumn perhaps, but before that other things to take care of, I am letting the new cycles versions settle in with blender before working to much with rendering there anyway, sculpting is however an area I don´t have to wait on, as well as trying out geometry nodes.

As for Lightwave not much in there that stirs up passion anymore, it´s stuck with dragging a performance heavy chain either in simulating vdb, or rendering and I have no idea If it will be improved due to the none development transparency..so not putting energy on that, but I know what will come for cycles.
 

RPSchmidt

Member
That sounds familiar. hehe. I didn't have the licensing snafu but a teaching snafu: couldn't tout LW as a way to earn a living into the future as an independent. Had to move to something else and this was the quickest, fastest, least painless method.
That computer sounds horrible!
Yeah, I couldn't see going Octane once I had both it and cycles at my access. While I think Octane renders beautifully, I've been able to get cycles (e-cycles, k-cycles, cycles-x) to render as beautifully as I think Octane looks. And quite quickly.

How has your blender time gone? Compared with your LW rendering, of course?
My Blender time has been a lot like many others; the learning curve wasn't too drastic with the newer versions, which I really appreciated. I remember trying Blender some years ago and it was so painful that I just gave up.

Render results are very good, comparable with what I could achieve with Lightwave. Some things are a little more involved, sometimes as a result of differences in the way that settings panels are laid out, sometimes due to slightly illogical placement of settings... sometimes due to the fact that Blender seems to move settings randomly between version releases, which can be frustrating.

Due to my current hardware situation, I haven't really been able to test GPU rendering effectively. I installed it on my personal system, but I haven't really played with it at all.... regular work just saps my will to do freelance work or experiment in my free time.

Modeling-wise, Blender has everything that I need, although it seems like it takes several more steps to get to the same point as I could in Lightwave; so modeling in Blender just feels like it takes waaaayy longer than it did in Lightwave. I could chalk that up to relative experience with each application, but I've been using Blender for about ten months or so now, and it honestly just feels like for regular modeling, Blender takes five clicks to do what Lightwave would do with two.

Also, the way that some tools work in Blender is a little wonky, which, especially early on, led to some frustration. Even now that I understand those issues, it still seems silly that there are additional steps I have to take to to do something that should be intuitive.

Overall, Blender is extremely capable. There are definitely things that I would change if I could, but the same can be said for Lightwave.
 
Top Bottom