PDA

View Full Version : How to deal with Octane's unsupported features?



Sebasvideo
04-11-2014, 04:18 PM
I downloaded the Octane demo and started playing with it, seeing how useful it would be since I have a GTX 770 with 4 GB of RAM and my few experiments with the Blender Cycles GPU engine were a lot faster than CPU. However it seems to have a lot of shortcomings compared to using native rendering: http://render.otoy.com/manuals/Lightwave3D/?page_id=61

I mean, polygons with holes is not something that a proper scene should have, but many scenes have more than 4 vertices. Some other things only mean an adjustment in the way you do surfacing, and I figured out how to do a basic surfacing with a texture without reading the manual, so it's not too hard to learn.

While it doesn't support multiple UV maps at least it allows scaling the texture, which Lightwave native doesn't and requires scaling the UV map itself and making multiple UV maps if the scaling is not the same for all the textures. But perhaps for some things multiple UV maps are needed regardless of scaling.

However, the two things that worry me the most are that it can't render particle systems and hypervoxels, so unless I'm missing something, it can't do clouds, correct? So what do you do when you need clouds? Just composite them in the layer behind it?

I read some of the official Octane thread in this forum, but it's 122 pages long, and while some things you can search for, like hypervoxels, I would have to spend hours reading all 1819 posts or at least the ones from the last year or so to get some kind of idea about this. So if you don't mind me asking, I don't need all the details, but when you need Hypervoxels, do you just do compositing, or is there some trick?

Also, the Lightwave backdrop doesn't work, right? You have to set the backdrop using nodes somewhere? At least in the demo my backdrop appears all white, and I have it set to a gradient like a sky in Layout.

Thanks

04-11-2014, 05:38 PM
It does have limitations.
Know them.
The software, though, is extremely stable and functional for anyone that recognizes the first line of this post.

You will not find hypervoxels as a useable item in that thread as it is a limitation, and thus, unusable with Octane. Comps work wonders.


I think the lw backdrop works. Something like that does. I haven't used it but I think it does. Join us in the thread.

prometheus
04-11-2014, 05:48 PM
Thatīs a downside to it currently, I donīt see it as an vfx shot tool, more like a product shot tool for products, interiors and archviz..maybe some motion graphics..so thus using on a vfx shot with interactive volumetrics isnīt what to be expected or
what to use it for, you still got Native lightwave that could take care of some of that at least...and using backdrop images is what is most appropiate to use....you could probably do a scene match, and render out hypervoxels cloud if you have set
the lighting in octane, use the same scene setup and ligthing angles in native lightwave to affect hv clouds, render out, add as backdrop in octane so you get the same lighting affecting the clouds just as it affect the octane scene overall.
Some stuff wonīt pass over nicely though, such as sunlight shadows from clouds actually affecting scene in octane, since it is only a backdrop, but the best would be to find a proper photage instead and match that if possible.
or Get vue and do skydomes at designed light angles.

fire and smoke commercials are prohibited:)

Michael

Sebasvideo
04-11-2014, 06:08 PM
Thatīs a downside to it currently, I donīt see it as an vfx shot tool, more like a product shot tool for products, interiors and archviz..maybe some motion graphics..so thus using on a vfx shot with interactive volumetrics isnīt what to be expected or
what to use it for, you still got Native lightwave that could take care of some of that at least...and using backdrop images is what is most appropiate to use....you could probably do a scene match, and render out hypervoxels cloud if you have set
the lighting in octane, use the same scene setup and ligthing angles in native lightwave to affect hv clouds, render out, add as backdrop in octane so you get the same lighting affecting the clouds just as it affect the octane scene overall.
Some stuff wonīt pass over nicely though, such as sunlight shadows from clouds actually affecting scene in octane, since it is only a backdrop, but the best would be to find a proper photage instead and match that if possible.
or Get vue and do skydomes at designed light angles.



Right, it seems like very specific, kind of like the Cycles engine in Blender, which has a completely different shading method, but it can really accelerate renders, and that IPR window is great, it gives me what I have in Modo with the separate preview window, but faster. Then again, not supporting HVs is a big problem...

I downloaded the Vue learning edition and it seems great, but I'd rather spend time learning Modo and Lightwave, which are difficult enough, adding even one more completely different program to the mix will only delay learning the other two. However Vue is useful in that it has a bunch of presets for trees, plants, mountains, etc, that can be exported to lwos, but you have to set the Ray Recursion Limit to 32 or some of their trees will show the backdrop through their leaves.

prometheus
04-11-2014, 06:19 PM
Right, it seems like very specific, kind of like the Cycles engine in Blender, which has a completely different shading method, but it can really accelerate renders, and that IPR window is great, it gives me what I have in Modo with the separate preview window, but faster. Then again, not supporting HVs is a big problem...

I downloaded the Vue learning edition and it seems great, but I'd rather spend time learning Modo and Lightwave, which are difficult enough, adding even one more completely different program to the mix will only delay learning the other two. However Vue is useful in that it has a bunch of presets for trees, plants, mountains, etc, that can be exported to lwos, but you have to set the Ray Recursion Limit to 32 or some of their trees will show the backdrop through their leaves.

Im all with you on that, feels like Iīm in some ways at the same confusing road as you are....though I do not have the pressure to decide anything now about what to choose of focus on, I would like to continue with lightwave preferably, but I also canīt help to recognize that it takes a little too long time, or has been taking a quite a long time for the lightwave team to recognize the "landscape. world generation" tools that is so needed.

so vue, and modo is what Ivé sneaked peaked on, a choice to invest in either modo or cinema might be in some future here, but only until I see what lightwave 12 brings...if those tools are still neglected, I think it will definitly be a new addition of something, with the recent helios released and showcased, it looks more that it now has something that eliminates some things of vue native stuff I didnīt like for a long time, so it might very well be the next tool for me..depending on what the price of both helios and vue will land on.

houdini might just be too expensive(for a full version that is) and to much time costly in terms of learning, cinema4d has been on the carpet too, but I would have to test it first...and trying to test all these stuff takes a little to much time unfortunatly.
Vue..I had a lot of time to play around with ...thanks to ple versions...and not some time limited demos.

If you need a product shot of glasses, pencils, clocks, jewelery, jump in to octane...if not stay away and do vfx shots in native lightwave.
I was by the way under the false impression that lightwave materials was now supported in octane..but I guess I was wrong.

Michael

04-11-2014, 06:40 PM
That's a pass-thru for materials.
Perhaps with 2.x we will get our lw texturing back.

Sebasvideo
04-11-2014, 06:55 PM
That's a pass-thru for materials.
Perhaps with 2.x we will get our lw texturing back.

Well, the Octane nodes have a connector called "Lightwave material". You say that's a passthrough, what does that mean in this context? For example, if I want to use the Lightwave Car Paint node, can I use it with Octane or not? For example, I had this machine that I had used the car paint node for, with Arsenic gray, and it looks amazing. But changing the same surface to the Octane Glossy material and choosing Arsenic gray gives me a completely washed out very light gray, and even if I select black from the color palette it's better, but nowhere near the look of the car paint node. I set the color space to linear like it says in the instructions and I could see it getting darker, but still not the same.

04-11-2014, 07:47 PM
Pass-thru means that you connect that material to the one you speak of and the lw render will render it. That is to say, the renderers know what they do or don't work with.


You don't seem to be acknowledging that Octane is a gpu-based renderer. No nodal/shader/textures coded (scalars work...some of the others, too) for the cpu/LW work with the gpu/Octane.

Though you may want it real bad, it ain't so.

Yeah, but what is your specular and roughness? Specular is your reflection and roughness is your blurry reflections. Sounds like your specular needs to be reduced. Or so that is what I have based on your description.

Oh, and you would have to build a car paint material; glossy alone won't get it done.

Sebasvideo
04-11-2014, 08:07 PM
Pass-thru means that you connect that material to the one you speak of and the lw render will render it. That is to say, the renderers know what they do or don't work with.


You don't seem to be acknowledging that Octane is a gpu-based renderer. No nodal/shader/textures coded (scalars work...some of the others, too) for the cpu/LW work with the gpu/Octane.

So does that mean that you can have both the Octane renderer and the LW renderer working at the same time contributing to the final render?

I know that Octane is a GPU based renderer, but I don't know the internals of it. I know GPU rendering has certain restrictions because GPUs are optimized for a specific set of instructions as opposed to the CPU which can process a much wider set, but other than the very basics, I don't know what goes on inside. And to be honest, I don't want to spend time learning the internals, I just want to see how can I take advantage of the faster rendering and how to deal with its limitations. I wonder, is Hypervoxels not implemented because the Octane team still hasn't worked on it, because it's not in the LW SDK, or because GPUs just can't render HVs?

Sebasvideo
04-11-2014, 10:32 PM
One thing that I just realized that is a problem with Octane besides unsupported features is that once you start a project with it or convert a project to use with it you are screwed if you want to use a render farm service to render the final version. I mean, an online render farm is not going to have Octane. And having read the whole manual on Octane for Lightwave I can tell that it pretty much takes over Lightwave. You have to change all the materials, lights, and many other things. You will set and adjust materials based on what you see in the IPR and Octane renders, and later you will have to change it all back to LW's native materials and lights to send it to the render farm.

So I guess Octane is great if you have several thousands dollars and some spare machines to buy a bunch of high end graphics cards, because if you have a project that is too heavy even if you have a beast of a graphics card, just one won't cut it, or even two on the same machine.

Unless your project is really small, changing all the surfaces back to LW's native is not a great option.

spherical
04-11-2014, 10:38 PM
Yep. Helps to plan ahead and know where you are going before you start, then choose the tools best suited to the tasks at hand.

04-12-2014, 05:28 AM
Sebasvideo,
You might be right. You might not be. Looking forward to what more research shows you.

Sebasvideo
04-12-2014, 09:08 AM
Well so far I'm not too encouraged by Octane, at least when it comes to large scenes. I spent hours converting most of the scene to Octane surfacing and kernels and all that, and I even left out some things like the trees. I read the whole manual to make sure I was doing it right (just the Lightwave Octane one, not the full Octane manual). I bumped up the render settings a little bit, but nothing crazy. It's not like I set it to 120 samples or something like that. 4 minimum 32 max, shading and light samples to 4.

Well, this morning it had rendered only two frames! Looking at the times of the files it took about three hours per frame to render. And we're talking a 1000x563, which is the demo size restriction for 16:9 AR. So a very small frame takes three hours to render while it takes me about 20 minutes to render it on my 3930k CPU, and I don't have to worry about not being able to render on a render farm.

Instances are problematic as well. Without Octane, you can hide the original object and display only the instances. For example, I have about 100000 trees instanced on random using a texture map (which makes it a lot less than 100,000 because the texture weight map only hides the trees in the areas in black). There are three tree models in this instancer, and obviously the originals are right at the origin, otherwise the instancing gets messed up. But in regular mode I can hide these originals and display just the instances. With Octane you can't. I confirmed this by searching in that giant thread, it has to be that way. Juanjo says the solution to that is to move the original out of sight, but if you do that, it doesn't work right. I sent the patch of grass 10 meters below ground. Adjusted the offset to compensate for that, and the grass shows in the air.

The other thing is that while this is a very heavy scene (that in theory I should be doing in Vue but I only have the free version which doesn't allow commercial projects, and I still would have to spend days learning how to use it) it takes about 10 seconds to load and then about 30 to display VPR in its regular version. The Octane version takes about 7 minutes to load the scene into the GPU to start rendering, or to start displaying the IPR. Sure, it's a heavy scene with the ground being a circle 3km in diameter, and a 1x1 meter patch of grass instanced 2 million times. But Lightwave by itself can handle it. It's a bit slow, but can handle it.

Sebasvideo
04-12-2014, 09:48 AM
I just opened my regular scene, set the samples to the same I had in the Octane scene, changed to 1000x563, put the camera in the same place, and rendered with the CPU. It took 11 minutes and 19 seconds. So I guess GPU rendering is not for everything.

On the other hand, I really liked some of the features Octane had, and it looks way more realistic than Lightwave's native renderer. But it's obviously not for everything.

tyrot
04-12-2014, 12:13 PM
octane eats heavy scenes like peanut butter...we do nothing but heavy scenes... Thousands of trees, many humans, cars...

You are a bit fast - and also - not qualified to judge octane. Because you are acting like it is a native renderer. It is a very well integrated GPU renderer.

Your instancing problem with octane. Well do you really know how instancing works? You can always select " L " local instead of "W" for any object in instancing panel.

Try and come back here again and complain...


Actually stop complaining in general - you do not know how to use Octane yet.

If you really wanna use it - we are here to help but from your general posting style you are a bit bashing Octane which is something not cool.. Because it is the COOLEST thing since Lightwave.. Enjoy it, read it, learn it.

Sebasvideo
04-12-2014, 12:41 PM
You are a bit fast - and also - not qualified to judge octane. Because you are acting like it is a native renderer. It is a very well integrated GPU renderer.

Your instancing problem with octane. Well do you really know how instancing works? You can always select " L " local instead of "W" for any object in instancing panel.

Try and come back here again and complain...


Actually stop complaining in general - you do not know how to use Octane yet.

If you really wanna use it - we are here to help but from your general posting style you are a bit bashing Octane which is something not cool.. Because it is the COOLEST thing since Lightwave.. Enjoy it, read it, learn it.

Dude, what did I ever do to you? A little bitchy, don't you think? I wasn't bashing anything, I was just reporting my experience with the demo. It never ceases to amaze me how people take it personally when you say anything remotely critical of their favorite software.

Regardless of the particular origin issue (and thanks for reminding me of the coordinates), the truth is that the demo simply took hours to render a frame that I rendered in 11 minutes with the CPU at the same size and samples. If I disable the instances it gets much faster, but I need the instances.

Like I said, I read the whole Lightwave Octane manual so I learned how to use it, I took hours in setting up everything according to the manual, and although I found the nodal surfacing system rather awkward, I liked some things about it like the IPR window which gives me the preview window I have in Modo. I also liked that the quality seems a lot more photorealistic than LW's own renderer. Some day I might buy it for other things, it's just not convenient for what I'm working on right now.

alexos
04-12-2014, 01:35 PM
Well so far I'm not too encouraged by Octane, at least when it comes to large scenes. I spent hours converting most of the scene to Octane surfacing and kernels and all that, and I even left out some things like the trees. I read the whole manual to make sure I was doing it right (just the Lightwave Octane one, not the full Octane manual). I bumped up the render settings a little bit, but nothing crazy. It's not like I set it to 120 samples or something like that. 4 minimum 32 max, shading and light samples to 4...

Eh, what? Perhaps I'm missing something here, but shading and light samples and AA level have nothing to do with Octane. I mean, you might as well use 1/1 and 1/1 and it wouldn't change a thing, as Octane uses its own camera, AA and sampling method. Also, there might be something fishy going on with your setup if the IPR takes seven minutes to pop up - I've been working with pretty polygon and imagemap-intensive scenes and the most I've ever had to wait was a handful of seconds (barring the occasional, but rare, insta-crash, of course.) Admittedly I've never tried 2 million instances, though, so perhaps there's something there - have you tried turning them off to see what happens? Also, which kernel are you using? And with one or two video cards?

ADP.

tyrot
04-12-2014, 01:42 PM
share the scene file ... please

Sebasvideo
04-12-2014, 01:50 PM
share the scene file ... please

I'm not allowed to do that.

Boris Goreta
04-12-2014, 02:47 PM
I'd just like to add here that in my LW+Octane experience I am super satisfied with the workflow, speed and results. It is just amazing. I believe you just need to learn it a bit more and adjust to it. I do pretty complex renders and animations and it just ripps through it. My interior renders are better 10 times then before. I don't spend any time surfacing objects becuase by default Octane materials look amazing and they are very simple to use.

If you have any problems please extract them and make a simple scene to share so somebody can help.

tyrot
04-12-2014, 03:09 PM
share any other file that you have the same problem ... slow and octane cannot be use in same sentence:)

m.d.
04-12-2014, 04:56 PM
Well so far I'm not too encouraged by Octane, at least when it comes to large scenes. I spent hours converting most of the scene to Octane surfacing and kernels and all that, and I even left out some things like the trees. I read the whole manual to make sure I was doing it right (just the Lightwave Octane one, not the full Octane manual). I bumped up the render settings a little bit, but nothing crazy. It's not like I set it to 120 samples or something like that. 4 minimum 32 max, shading and light samples to 4.



these settings have zero effect on octane FYI

that makes me think you were using the kernal default of 1000 samples probably PMC?

Octane is way faster then lightwave with instances and heavy GEO....
Try rendering the scene with monte carlo radiosity non interpolated with 16 bounces and AA set at 1000 in LW render and see how long it takes.....cause rendering with default PMC or Pathtracing would be the same...I am betting a lot more then 3 hours

To get it the fastest....set it to direct lighting 'none' or ambient' with 1 bounce on all the diffuse gloss and spec settings and samples to about 400...this would just be to test speed of render

m.d.
04-12-2014, 04:59 PM
also if your grass and trees are not animated....set the render target to only update transformations....not full scene reload

Then it will only load your scene once and just move the camera

Sebasvideo
04-12-2014, 05:33 PM
these settings have zero effect on octane FYI

that makes me think you were using the kernal default of 1000 samples probably PMC?

Octane is way faster then lightwave with instances and heavy GEO....
Try rendering the scene with monte carlo radiosity non interpolated with 16 bounces and AA set at 1000 in LW render and see how long it takes.....cause rendering with default PMC or Pathtracing would be the same...I am betting a lot more then 3 hours

To get it the fastest....set it to direct lighting 'none' or ambient' with 1 bounce on all the diffuse gloss and spec settings and samples to about 400...this would just be to test speed of render

Sounds good, I'll try all that and report back.

gristle
04-12-2014, 05:35 PM
share any other file that you have the same problem ... slow and octane cannot be use in same sentence:)

I have to disagree with you on that point to some extent. Over the last week I've been testing Octane with some old 777 plane interior's from a few years back. The lighting is fairly complex and needs several bounces to do it justice. There are some patches of noise that take for ever to resolve out. Sure, I could use neat image to reduce it in post, however the noise levels are inconsistent across the scene.

The original renderings were completed in Kray which chewed through them, but had other issues (arghhh moire effect on textures!!)

My learning from all this is the following. If you are using an engine like Kray, you can tweak the settings as much as you like to speed things up. The downside is potential splotching, light leaks etc. If you are using Octane, you cannot do much, apart from throw more graphics cards at it. There are no ways yet to limit recursion per surface and other tricks LW/Kray users may be use to.

As the others have said, you need to come at this from a different perspective. Good luck!

Sebasvideo
04-12-2014, 06:59 PM
also if your grass and trees are not animated....set the render target to only update transformations....not full scene reload

Then it will only load your scene once and just move the camera

Actually it was set to only update when I left it last night.

Sebasvideo
04-12-2014, 10:25 PM
I'm trying to leave another test render overnight but one thing I can't figure out is the equivalent for "Ray Recursion Limit" from the LW renderer. I need this because I have a tree that I exported from Vue and it opens fine but using the LW renderer I have to set that setting to 32 or the leaves show the background gradient. With Octane, even after I import the texture with the Octane method, the leaves look weird.

I thought maybe it would be the diffuse depth setting in the direct lighting kernel, but that doesn't do anything. I searched for "recursion" in the Octane thread but one result came up and it was about something else, and Google didn't give me anything useful either. Do any of you know what's the equivalent?

m.d.
04-13-2014, 12:19 AM
It is exactly that....
In direct lighting it is the diffuse spec and glossy controls

In the other kernels it is a single control....

It controls how many times the rays bounce....


32 is a very excessive number.....on the 32nd bounce virtually no diffuse light will be contributed to the scene....

I have had to go to 20+ when rendering a pile of diamonds because of the complex retractions....and when octane hits it's final recursion it will return black.....so I would get little black spots in a brightly lit scene...
But that was for reflections and refractions

If you know how Vue works, you'll know it gets it's speed by rendering its ecosystems with usually as little as 2 bounces of radiosity (it has some smart auto settings for ecosystems).....that means effectively a ray recursion of 2 as far as diffuse is concerned

I really doubt you need the 32 bounces....but if you really want them...run the path tracing kernel and you can set it as high as you want....

Just to be clear though....Vue would never run that high....and you don't really need anywhere near that without transparent or reflective surfaces in any style of render....

The issue may be the leaves cut outs are not completely transparent.....and each time a ray passes through the leaf, it uses up a "bounce" or "recursion" or hit ect.....but this is not a diffuse issue....that would be a specular issue....

Try the path tracing and see if that solves it

m.d.
04-13-2014, 01:08 AM
One more thing....

Make sure to check your leaf texture....
Not sure if your looking close up at the tree or not, but if you had the alpha backwards that could be leading to some of the issues...

I got stuck on this myself with some imported plants....LW render needed the alpha inverted compared to octane render, but it was hard to tell what was going on because of the mass of leaves...

Also if the texture is semi transparent...then it could cause some of the issues you describe...it's alpha should be full white or full black

m.d.
04-13-2014, 01:14 AM
I have to disagree with you on that point to some extent. Over the last week I've been testing Octane with some old 777 plane interior's from a few years back. The lighting is fairly complex and needs several bounces to do it justice. There are some patches of noise that take for ever to resolve out. Sure, I could use neat image to reduce it in post, however the noise levels are inconsistent across the scene.

The original renderings were completed in Kray which chewed through them, but had other issues (arghhh moire effect on textures!!)

My learning from all this is the following. If you are using an engine like Kray, you can tweak the settings as much as you like to speed things up. The downside is potential splotching, light leaks etc. If you are using Octane, you cannot do much, apart from throw more graphics cards at it. There are no ways yet to limit recursion per surface and other tricks LW/Kray users may be use to.

As the others have said, you need to come at this from a different perspective. Good luck!

True...
In my experience, some of the things you think octane would excel at it is very slow....some caustics and complex interiors take longer to solve then other renders....then other thinks you would think it wouldn't be good at it flies through

Outdoor scenes it seems to really shine....especially with lots of instanced objects....plants ect
The sheer number of samples does away with the flickering that lower AA passes of other renders are plagued with.

JBT27
04-13-2014, 03:17 AM
This is all interesting - I saw this thread, got to thinking and downloaded the demo; only recently bought hardware that makes it viable to think of such things. Outdoor scenes with tons of instances is a mainstay of what we do, and we could do with some more efficient (and better looking) rendering - I get that the look is also down to the artist, so not kicking LW or Octane or anything on that. I spent some of last night looking round for a blow-by-blow tutorial on getting a scene up and running with it. Used the demo scenes in the standalone, and the demo scenes in LW, but was getting something like 2m40s per frame, for the 1000x600 or whatever, and thinking that was pretty slow ... before anyone jumps on my head, I clearly am also not thinking like Octane, and keen to find out more and learn, and buy if I see the speed being what people are praising all the time.

Are there any tutorials out there, rather than the more more focussed ones on certain elements? I'm persuaded that clearly I am not seeing its real potential at the moment, and obviously that's my fault and my noobness.

Julian.

tyrot
04-13-2014, 04:34 AM
JBT ... actually there must be more LW - octane related videos. you are right on that.

What is your GPU?

JBT27
04-13-2014, 05:51 AM
JBT ... actually there must be more LW - octane related videos. you are right on that.

What is your GPU?

I've got a Titan Black.

Julian.

04-13-2014, 06:25 AM
Julian,
You've got a nice card.

And with that card, you will have to throw a whole lot in a scene to see the worth of it. A scene with few items will often render faster with the native renderer; you see Octanes power when you throw in glass, reflections, and lots of geometry.


As to tutes, look for the ones on the standalone. Really, the standalone can't be ignored -- ya gotta look at what their paradigm is for how to use it. I have learned more about mixing materials than I would ever have bothered with in LW. I think my lw shading is better because of this. Lighting has undergone changes, too, because of the camera-like settings in the imager: so many ways to mix your lighting to get a look you like.

Go over to otoy and have a read there, too. It seems much of the technical questions have been discussed there.


Octane feels very raw, indeed. Feels like there's nothing there. Very deceptive.

One of the things I see in comparisons to speed vs LW is a lack of acknowledgement that Octane is in Monte Carlo, brute force mode all the time. (May not be Monte Carlo-code but you get the gist.)

Sebasvideo
04-13-2014, 08:46 AM
I really doubt you need the 32 bounces....but if you really want them...run the path tracing kernel and you can set it as high as you want....


Actually you need it, and this is the only case where I had to raise the default of 6 for anything. And the only reason I did it was because I searched for this problem and somebody had posted that that was the way to take care of it, and it did. Now, if it's trees in the distance, about 16 takes care of it, but the more you get close to it, at 16 you still have holes that show the background. 32 was the first number that showed no holes. But the truth is, it doesn't really add a lot of time to the render. In my case the big difference is the instanced grass, when I disable it the 32 recursion limit really adds only a few seconds. I can't remember exactly how many, but I tested for it and it was a small difference.

As far as inverting the alpha, that makes the whole tree invisible. I attached a screenshot of the nodal setup and a render of the tree so you can see, without the real grass and just a grass texture on the ground, because the instanced grass still takes Octane 6 to 7 minutes to load the scene into the GPU and show anything, which tells me that Octane doesn't work that well for scenes with too many instances.

121419

121420

m.d.
04-13-2014, 08:55 AM
One of the things I see in comparisons to speed vs LW is a lack of acknowledgement that Octane is in Monte Carlo, brute force mode all the time. (May not be Monte Carlo-code but you get the gist.)

It is very similar to non interpolated Monte Carlo in the path tracing and PMC kernels (with a MC bounces set to 16 instead of 2) but unlike even brute force lightwave....the radiosity is recalculated on every sample...usually hundreds in octanes case.

When lightwave shoots a ray at a pixel it is in a sense divided computationally.....you will have radiosity rays, which are governed by the number of rays, primary and secondary ect, and the rays contribution is then interpolated together (usually).
The radiosity is precomputed before the regular ray trace render starts in every case....interpolated or not.

Then you have your normal ray tracing that is controlled by the ray recursion limit.

Then your AA...which when it detects a contrast difference will fire out additional rays directed at that pixel to solve it....but it can only sample the precomputed radiosity....that is why radiosity is calculated first.

So additional AA cannot add samples to the precomputed finite radiosity solution....it can only sample ray tracing and further refine the pixels contribution to the scene.

With octane it is simple.....it fires all radiosity rays all the time.
For each sample it is recalculating that pixels contribution to radiosity all the time.
Each sample it fires will randomly alter the path the ray takes after the first bounce to refine the solution.

In the PMC kernel....it mutates the path in a random but slightly controlled manner more towards what it thinks is important to sample.

The random path alteration is where Monte Carlo gets its name....
So lightwave will alter the path....only during radiosity....

After the radiosity prepass it will then fire rays at corners and edges of high contrast pixels. But doing so cannot alter the radiosity solution. This is the bias in biased rendering. Lightwave has a biased pre conceived opinion on what the radiosity should be..... :)

Octane will keep firing rays, and keep altering the path of those rays until eventually, the full radiosity solution is reached in a physically accurate way....this is why it is unbiased. It is always refining its opinion of the radiosity solution.

The closest thing to compare octane to for LW users is actually FPrime in MC mode. As FPrime didn't run a radiosity prepass.....so it too would contribute every ray fired to the radiosity solution.

JBT27
04-13-2014, 08:58 AM
Julian,
You've got a nice card.

And with that card, you will have to throw a whole lot in a scene to see the worth of it. A scene with few items will often render faster with the native renderer; you see Octanes power when you throw in glass, reflections, and lots of geometry.


As to tutes, look for the ones on the standalone. Really, the standalone can't be ignored -- ya gotta look at what their paradigm is for how to use it. I have learned more about mixing materials than I would ever have bothered with in LW. I think my lw shading is better because of this. Lighting has undergone changes, too, because of the camera-like settings in the imager: so many ways to mix your lighting to get a look you like.

Go over to otoy and have a read there, too. It seems much of the technical questions have been discussed there.

Thanks! I need to spend more time with the demo and get more familiar with Octane - the CPU renders I'm doing are really fast, so I get what you say about needing to throw very big scenes at Octane to see the advantage. But, I'm going to work with the demo and see what 2.0 has to offer, if I don't buy into it before that.

Julian.

m.d.
04-13-2014, 09:07 AM
Actually you need it, and this is the only case where I had to raise the default of 6 for anything. And the only reason I did it was because I searched for this problem and somebody had posted that that was the way to take care of it, and it did. Now, if it's trees in the distance, about 16 takes care of it, but the more you get close to it, at 16 you still have holes that show the background. 32 was the first number that showed no holes. But the truth is, it doesn't really add a lot of time to the render. In my case the big difference is the instanced grass, when I disable it the 32 recursion limit really adds only a few seconds. I can't remember exactly how many, but I tested for it and it was a small difference.

As far as inverting the alpha, that makes the whole tree invisible. I attached a screenshot of the nodal setup and a render of the tree so you can see, without the real grass and just a grass texture on the ground, because the instanced grass still takes Octane 6 to 7 minutes to load the scene into the GPU and show anything, which tells me that Octane doesn't work that well for scenes with too many instances.

121419

121420

There's something funny with your tree texture.....can you post the tree....not the scene? So I can take a look..

Those GPU loading times are crazy....and I won't argue that....
I did a scene a week ago with 500,000 blades of grass and a few hundred trees and it was taking less then 20 seconds to load...

Sebasvideo
04-13-2014, 09:25 AM
There's something funny with your tree texture.....can you post the tree....not the scene? So I can take a look..

Those GPU loading times are crazy....and I won't argue that....
I did a scene a week ago with 500,000 blades of grass and a few hundred trees and it was taking less then 20 seconds to load...

Sure, I can't attach a zip file here, so I uploaded it to my Mediafire account: https://www.mediafire.com/?h639o03m65xwck3 That zip has the object, the regular texture and the alpha one.

500,000 is not too much, so it wouldn't take long to load that. My grass patch is 5.70x5.70 meters and has 502948 polygons, instanced 2 million times in random mode over a terrain that is like a circle about 3 km in diameter, and hills and mountains sculpted in Modo. Now, I don't know if this is the right way to calculate it, but the number of polygons if multiplying the object by the # of instances would be 1,005,896,000,000. My guess is that it is too much for the 4 GB of VRAM in the card, but the 32 GB of RAM can handle them better. Still, the thing is that the scene doesn't take forever to load, and pressing F9 without Octane starts rendering after a few seconds. VPR takes a few seconds too, sometimes more, sometimes less.

Before any of you goes into snob know-it-all mode and tells me what do you need so many for, blah, blah, blah, I already had that discussion with other people and I'm not going to have it again. My project is what it is for a reason, and I don't plan on changing it. I'm clarifying this because I already had one of you in this thread talking down on me like I was an idiot for not knowing Octane as much as he did, and I'm not going to put up with that again. On the other hand m.d. gave much more constructive input, so I really appreciate that.

m.d.
04-13-2014, 09:28 AM
EDIT: Didnt see above post first
There is something wrong with your texture I beleive

This is a tree of mine with DL set to ray recursion of 1
121421

here bumped up to 5 for quality
121422

Sebasvideo
04-13-2014, 09:33 AM
EDIT: Didnt see above post first
There is something wrong with your texture I beleive

It could be, I made a tree with DP tree and it didn't have all their weirdness, certainly didn't need the recursion limit raised. But the Vue tree was exported to lwo and once I set the rec. limit to 32 it looks fine in a normal render. But I know that trees exported from those types of apps are problematic. Xfrog has a lot of plants and trees that they give out for free, that I downloaded in lwo format and they also have this problem but worse, because in their case, raising the rec. limit doesn't fix it.

m.d.
04-13-2014, 09:36 AM
121423

your alpha was screwed....

I think the PNG for alpha has transparency and color...

change the alpha texture to color image, and invert

EDIT: Like I said ray recursion not the issue

Sebasvideo
04-13-2014, 09:48 AM
I think the PNG for alpha has transparency and color...

change the alpha texture to color image, and invert

Yep, that worked perfect. Thanks!

BTW, your advice to lower the number of samples from the default of 1000 worked as far as rendering times go... When setting it to 400 it takes about 3 minutes to render. Makes me wonder why they set it so high by default, when they want to tout the speed of the product.... Like the LW renderer, the default settings are too low and render fast but not great quality. If Octane gets decent quality at 400 samples why do they set the default to 1,000?

m.d.
04-13-2014, 09:50 AM
121424

Just instanced that tree 250,000 times, about 50 billion polys

took about 15 seconds to load

- - - Updated - - -


Yep, that worked perfect. Thanks!

BTW, your advice to lower the number of samples from the default of 1000 worked as far as rendering times go... When setting it to 400 it takes about 3 minutes to render. Makes me wonder why they set it so high by default, when they want to tout the speed of the product.... Like the LW renderer, the default settings are too low and render fast but not great quality. If Octane gets decent quality at 400 samples why do they set the default to 1,000?

1000 is very clean, if you can afford the time

Sebasvideo
04-13-2014, 09:51 AM
121424

Just instanced that tree 250,000 times, about 50 billion polys

took about 15 seconds to load

Did you use the LW native instancer, or the Octane one that reads the clones? Because I need to use the native one in surface mode and random, with a texture map to control where the instances show.

m.d.
04-13-2014, 09:54 AM
Native

Maybe the weight maps are the issue

Sebasvideo
04-13-2014, 10:00 AM
Native

Maybe the weight maps are the issue

I bet in your case it loads faster because there's no randomization. I tried that but it looks too much like a grid, even though I did the patch in Modeler using the point clone plus and I set random heading and some offset, and it's three different grass objects.

The weight map is a texture with either black or white, some patches of white in a mostly black background.

m.d.
04-13-2014, 10:12 AM
121425
121426

Massively random....rotations, scale strecth texture applied to weight map....50,000 instances loads in less then 5 seconds

you must have a bottleneck somewhere...it may be in the demo

Sebasvideo
04-13-2014, 10:17 AM
Right, but that tree, if it's still the one I sent you, or similar, has 139916 polygons. Multiplied by 50,000 it's 6,995,800,000. Still far from the 1,005,896,000,000 I'm using for the grass.

m.d.
04-13-2014, 10:39 AM
True, now I was running 50 billion without an issue....
I tried a million instances of your tree and shut it down after 1.5 minutes

So 1 trillion polys may be a little high.

Now as far as your needs go....not to be rude, but very few hollywood productions have rendered trillion polygon scenes. I think Elysium has the record at 3 trillion and needed arnold for it....but their trees were up to 20 million poly's each.
http://www.fxguide.com/featured/elysium-a-practical-miniature-and-digital-fx-odyssey/

It may make sense to use some other tricks...cards ect for the distant vegetation....or if doing a flyover load it into like 5 separate passes

Battlestar Galactica even had to load the lead ship in chunks and render it in chunks in lightwave native render...and it was only a billion polys


In a 1920 by 1080 scene there are only 2 million pixels....in a stationary scene that translates into 500,000 polys per pixel

Have you tried this scene in LW native? if so how does it handle it...what are the render times

Sebasvideo
04-13-2014, 11:06 AM
Now as far as your needs go....not to be rude, but very few hollywood productions have rendered trillion polygon scenes. I think Elysium has the record at 3 trillion and needed arnold for it....but their trees were up to 20 million poly's each.
http://www.fxguide.com/featured/elysium-a-practical-miniature-and-digital-fx-odyssey/

Have you tried this scene in LW native? if so how does it handle it...what are the render times

Yes, I'm guessing you definitely would need Arnold Schwarzenegger to lift that many trees :ohmy:

The scene was made in LW natively and it adds quite some time to the render, however, LW loads the scene in seconds and when I press F9 it starts rendering after a few seconds as well. It makes the viewport slow to navigate, but what I do is uncheck the grass instance when working in the viewport and I enable it when I need to see it in VPR.

Granted, this is something that would be best done in Vue, but I don't have a budget for that now and I don't have the time to learn it while I'm still learning Lightwave and Modo.

I still have to run some tests, but even if it's fast I'm not sure Octane it's what I need. For one thing, everything needs the nodal system. I never understood why LW users are so in love with nodes, I hate them. I understand that it offers some huge advantages for very complicated setups for some things, but for simple surfacing it's annoying, it's another huge window taking up space on the screen for no good reason, added to the several other windows that Lightwave has. Being used to the shader tree in Modo which is far more convenient, or even the simple surface editor in Lightwave which is not as good but still faster to interact with and takes less space, I find the Octane nodal system annoying.

But that's just an annoyance that I can get used to, if it's fast enough that will let me render everything at home. The problem is that if for example one frame takes 5 minutes to render, one second would take two hours. Which means one minute would take 7200 minutes, or 120 hours, or 5 days. So I would need to send it to a render farm, but there seems to be only a handful of render farms that work with Octane, while there are many that work with Lightwave. Everybody recommends garagefarm.net as a cheap and good quality service, but as far as I can tell they don't work with Octane.

The only way Octane would be a good alternative for me right now is if it gave me a good render at one minute per frame, but I'm guessing the quality is not that great at one minute.

Sebasvideo
04-13-2014, 11:44 AM
Another thing that I think it's a big mistake by the developers is restricting the frame size. I mean, you already have those waves and the logo in every render, are you going to use the demo for any final render if you can render it at 1080p? Because the truth is, I can't really judge a product in a handicapped state, I know how long it takes to render a frame at 960x540, but how long will it take to render at 1920x1080? Is it going to be twice as long? because most times, that is not the case. They should simply make the demo fully functional for 30 days so you can test it properly, not frame size restrictions, and no watermarks. Otherwise I'm testing something that is not what the real product offers, and $472 may be small change for some people, but a lot of money for others.

If they worry that some people might use the demo for a commercial render, they will only be able to do that for 30 days, and as far as I've read, Octane uses the network card number for activation, so it's not like you can re-install Windows to keep using it after the demo period is over.

tyrot
04-13-2014, 12:28 PM
sebas - buy it.... come on ... :) dont deal with demo ... it will change your renders ... you will earn more money... your girlfriend will love you more (you will have more time for her).... you will start watching NVIDIA conferences... all the time... TITANs are waiting for you... come on... stop dealing with -crippled demo :)

I write here publicly - "IF you do not want it after sometime.... i will buy that license from you...." (but before doing that - check that - license transfer is OK)

Sebasvideo
04-13-2014, 12:42 PM
sebas - buy it.... come on ... :) dont deal with demo ... it will change your renders ... you will earn more money... your girlfriend will love you more (you will have more time for her).... you will start watching NVIDIA conferences... all the time... TITANs are waiting for you... come on... stop dealing with -crippled demo :)

That makes me want to buy it even less.

tyrot
04-13-2014, 01:12 PM
oh boy.. you are a hopeless case

erikals
04-13-2014, 02:23 PM
now, why not use high-res trees for the ones up close to cam and low-res for the other ones...

also it might very well help to use low-quality png files...
https://www.youtube.com/watch?v=iFccG3v5ei8

Lewis
04-13-2014, 02:45 PM
121419


Switch your alpha to "Float image" and invert, then you should see leaves transparency working.

alexos
04-13-2014, 03:27 PM
I still have to run some tests, but even if it's fast I'm not sure Octane it's what I need.

Ah, that's the big question, isn't it. If you were into spaceships and character animation and suchlike, then no, you probably wouldn't need it; in fact it could damage your workflow, because of the limited set of nodes available (more have been promised, but still a fraction of what's available in native LW) AND because, as great as Octane's quality is, it can be significantly slower than LW's renderer in quite a few cases...


For one thing, everything needs the nodal system. I never understood why LW users are so in love with nodes, I hate them. I understand that it offers some huge advantages for very complicated setups for some things, but for simple surfacing it's annoying...

Err... Yeah, so I guess the "not enough nodes" thing isn't going to be a problem for you. Here, however, is where Octane really shines: you mostly won't need those complicated setups. Basic materials such as glass, metal and all those other ones that used to drive me mad, because frankly I never could understand why I'd have to connect seventy-four different bloody nodes just to get some brushed aluminum, are a matter of a single node within Octane; three or four if you want to get fancy and add the usual dirt, rust, blah blah, but generally speaking? Glass is glass. Metal is metal. It's so much simpler it actually feels like cheating, so simple surfacing is indeed better.

Now, of course it is still less straightforward than the old surface editor, but that's the nature of the nodal system in general; there are also other workflow-related issues with projections and whatnot, but Juanjo is working on those and judging from what he's accomplished so far, I'm pretty confident things will get much better soon... Speaking of which, have you seen the displacement system preview? It looks like it could make instanced grass redundant, amongst other things.


But that's just an annoyance that I can get used to, if it's fast enough that will let me render everything at home. The problem is that if for example one frame takes 5 minutes to render, one second would take two hours.

I'm still not convinced about that rendertime - for an exterior scene it really shouldn't be this high, so there might be something wrong with your config/model/video card/I don't know. It might help if you could post a screenshot of your render target network. It also occurs to me that I've had problems with very big (size-wise, disregarding polycount) objects before, so there might be something in there as well...

ADP.

Lewis
04-13-2014, 03:31 PM
Sebasvideo - Here is scene with your tree

https://www.dropbox.com/s/ngbg72at6ktevno/Octane_Summer_TREE.rar

And quick render fo 50 instanced trees

cheers

Sebasvideo
04-13-2014, 04:39 PM
Basic materials such as glass, metal and all those other ones that used to drive me mad, because frankly I never could understand why I'd have to connect seventy-four different bloody nodes just to get some brushed aluminum, are a matter of a single node within Octane; three or four if you want to get fancy and add the usual dirt, rust, blah blah, but generally speaking? Glass is glass. Metal is metal. It's so much simpler it actually feels like cheating, so simple surfacing is indeed better.

You're exaggerating, right? I mean, why use nodes for brushed aluminum in the first place? There might be something you're doing that is a lot more elaborate than just brushed aluminum that needs nodes, would you mind telling me? I'm not doubting you BTW, I'd really like to know because I learn a lot from forum users besides tutorials, paid courses, books, etc.



I'm still not convinced about that rendertime - for an exterior scene it really shouldn't be this high, so there might be something wrong with your config/model/video card/I don't know. It might help if you could post a screenshot of your render target network. It also occurs to me that I've had problems with very big (size-wise, disregarding polycount) objects before, so there might be something in there as well...

ADP.

Well, it depends on the number of samples. I think the reason why my first overnight render was so long is because I hadn't added the direct lighting kernel, actually no kernel at all, because I had added the daylight environment and that was enough, so I'm guessing it was defaulting to some extreme # of samples. After m.d. mentioned the D.L. and sample size it came down to like 2.5 minutes per frame, which is fast, but I'm working on a project that can be 2 minutes long, so that would still take days to render.

Now, earlier I went somewhere for a few hours and I set Octane to a max of 0.5 minutes per frame to see what it looked it. Here's where I was talking about the demo problem. It's not only that they put that stupid watermark with the logo and the lines, but also that it changes every frame. So I come back home, import the rendered frames into After Effects, do a RAM preview, and it seems to look decent, but with some noise, and the damn lines are all over the place moving, and they don't really let me tell how much noise is there. Even more, because of the lines, I can't really run Neatvideo on it to see if I can make it better. Also, because of the size restriction, I don't know how this would look at double the size, which is what I need. Sure, at 960x540 it looks OK, but what happens at 1080p?

So if it takes about 2 minutes to render something decent at half 1080p, then perhaps it needs 4 minutes to render 1080p. Or 3 for 720p. Meaning it would take 72 hours per minute of animation. Still too long for me. And like you said, there are the things that you cannot do with Octane, which was my point in starting this thread. Sure, you can composite clouds in the background. But what if you need particles and hypervoxels in a way that interacts more with the scene? I'm not bashing Octane, I know it's really great for many things, but not so great for others. If, like you say, is not that great for spaceships and character animation (something I don't do but hope to some day), then at least for now it may not make too much sense for me.

spherical
04-13-2014, 05:32 PM
It appears that the alpha map has lost its UV. You're seeing the entire poly of each leaf sub-set. This will add to the recursions significantly, because there are large sections that are now in front of a multitude of other large sections that normally would have been truncated to only the leaves' shapes.

geo_n
04-14-2014, 02:53 AM
I still have to run some tests, but even if it's fast I'm not sure Octane it's what I need. For one thing, everything needs the nodal system. I never understood why LW users are so in love with nodes, I hate them. I understand that it offers some huge advantages for very complicated setups for some things, but for simple surfacing it's annoying, it's another huge window taking up space on the screen for no good reason, added to the several other windows that Lightwave has. Being used to the shader tree in Modo which is far more convenient, or even the simple surface editor in Lightwave which is not as good but still faster to interact with and takes less space, I find the Octane nodal system annoying.

You must have patience OLD friend. Learning is always time and effort and googling wouldn't hurt.

Sebasvideo
04-14-2014, 08:36 AM
You must have patience OLD friend. Learning is always time and effort and googling wouldn't hurt.

I'm not your friend, I don't befriend bullies who chase me across threads to post lies and insulting comments about me. Please stay away from threads I start, and don't reply to my posts on other threads. I already reported you for harassment.

Sebasvideo
04-14-2014, 09:34 AM
It appears that the alpha map has lost its UV. You're seeing the entire poly of each leaf sub-set. This will add to the recursions significantly, because there are large sections that are now in front of a multitude of other large sections that normally would have been truncated to only the leaves' shapes.

I know, overall these trees from Vue have lots of problems. I might end up using DP Tree and make two or three models. I didn't use it because you have to use your own leaves and I figured Vue would save me some time.

Sebasvideo
04-14-2014, 11:00 AM
...and googling wouldn't hurt.

I would like to clarify this, not for this particular individual but for the majority of you who unlike him are very nice and willing to help, because I wouldn't want you to believe what this individual is saying about me, that my first resort is to ask questions in the forum for every single problem I have as opposed to doing Google and forum searches first, and that I'm being "spoon fed" LW knowledge and that I should pay forum users a cut of my "paid" gig.

First of all, my "paid" gig is my first one as it relates to 3D modeling and 3D in general, so I quoted the client a very small price just to get my first gig. It's not like I'm getting rich by asking questions here. In fact, due to being new at this I quoted a very small price thinking it would take me a lot less time and then I realized that the project was far bigger than I thought, but live and learn, right?

Second, it would not make too much sense to post questions here as my first resort, because it takes more time. Time to type the post, which sometimes has to be very detailed so it takes several minutes, and time to wait for someone to answer. So it's not my first choice by far, and most of the challenges I was presented with in this project I solved by searching this and other forums, and Google and YouTube. I watched lots of tutorials, besides having bought Dan Ablan's Signature Courses for both LW 11 and 11.5. I also bought and read William Vaughan's Digital Modeling book. It's just quicker most times to search for a specific problem online because there's a chance somebody else had it before, like for example right now I'm having trouble with DOF because no matter how high I set the samples, the blurred areas still show noise. So I searched here and somebody had the same problem, and Rebel Hill posted a link to one of his always excellent tutorials which shows everything about the render settings and goes beyond the explanations in the manual, which I also read and is also my other go-to solution before posting, and often takes care of the issue, in fact the folder with the four PDFs is in my list of favorite folders on the left pane for easy access.

The implication of this individual that I'm being "spoon fed" is quite ridiculous. "Spoon fed" would apply more to someone who bought Lightwave yesterday, doesn't know anything about it, and his first resort is to post questions about every single thing. While my experience with Lightwave is more or less about six months, and a little over a year with Modo, I still consider myself a newbie because learning 3D is harder than almost anything else when it comes to software. I admire all of you for having so much knowledge about Lightwave and I'm very grateful to those of you who take time to reply to my questions. Which is why I wanted to clarify this to most people here, because the rude and arrogant ones are less than 1% of those who are nice and willing to help.

Hail
04-16-2014, 11:09 AM
I'm not your friend, I don't befriend bullies who chase me across threads to post lies and insulting comments about me. Please stay away from threads I start, and don't reply to my posts on other threads. I already reported you for harassment.

Come onn.. Sebasvideo, I know geo_n can be a pain in the butt sometimes but trust me he is a nice guy to have around:)
Besides we are all lw users(a very rare breed to come across in the CG industry these days)
So what shall we profit if we fight each other to extinction?:D

Boris Goreta
04-16-2014, 11:20 AM
I always first ask here at the forums and then if nobody answers, which almost never happens, I search google, it is much simpler and you get a direct answer. People are happy when they can help you.

andyhow
04-18-2014, 04:58 AM
Sorry to budge in and detract from the original reason of this thread but I bought a used Titan from a guy. He was running Octane with 2 Titans(which he's selling now) and he said it wasn't not enough. So he went on to get Titan Z. Apparently he said Octane doesn't benefit from 2 SLI'ed GPUs and it requires everything in a single processor with all the VRAMs and Cuda cores. Maybe you guys are more experienced with the GPU setups might have better insights. Titan Z is basically 2 Titan Blacks sandwiched into 1 card. So that means 2GPUs with 12GBs.

Point is, I was discouraged and stayed away from Octane after hearing that. It gave me the impression Octane is going to burn a hole in your pocket on GPU muscle acquisition.

tyrot
04-18-2014, 05:26 AM
first of all - SLI is NOT recommended for Octane. SLI actually cripples the performance drastically. But i have two 780 GTX i m very happy with them . Performance is doubled.. (titan Z * 2 i guess heaven on earth)

(and staying away Octane - because of SLI'ed Titans... is not very wise .. because even one Titan can change your render world like no other renderer)

andyhow
04-18-2014, 05:56 AM
So your setup is with 2 displays on 1 780 and another just for octane?

juanjgon
04-18-2014, 09:48 AM
Octane supports all the GPUs available in your system for rendering. The GPUs should not be configured in SLI. The memory you get for rendering is the GPU memory available in the GPU with the smaller memory size ... for example if you have a Titan with 6GB and a 780 with 3GB rendering together you are going to have only 3GB for rendering ... if you enable only the Titan for rendering you are going to have the full 6GB of free RAM.

Octane, as unbiased renderer, linear scale it's rendering performance to all the GPUs used. Two GPUs means double performance than rendering with only one.

-Juanjo

andyhow
04-18-2014, 09:55 AM
Thanks for the info Juanjo!

Sebasvideo
04-18-2014, 10:59 AM
Sorry to budge in and detract from the original reason of this thread but I bought a used Titan from a guy. He was running Octane with 2 Titans(which he's selling now) and he said it wasn't not enough.

Could it be that his scene was extremely large? Yesterday I put in my older card, a GTX 570 with 1.2 GB of RAM, and along with my GTX 770 with 4 GB of RAM, I see quite a speed improvement. The only problem is that it takes the available memory down a lot, not even the full 1.2, it shows much less than that in the IPR, probably because I set the GTX 570 as my main card so I can work on the computer while the GTX 770 is rendering, and if I have other apps open it needs memory for those two. But as far as processing speed, I notice quite an improvement.

andyhow
04-18-2014, 01:51 PM
I've tried Octane demo before when it was in beta stage. I had a GTX570 and a GTX670 running and octane completely pulled my machine down to a stuttering mode(cursor not reacting, things cannot be moved and crashed). That gave me the impression it needs a very good setup and I felt my 2 cards were not up to it. Now, I've got a Titan I can consider Octane but I guess it's still best to have 2.

Anyway, I'm having issues now installing the demo. I get this message(attached) when I launch LW 11.6.2. I put the octane.dat and octane.dll in LightWave/Bin according to instructions.

121513

juanjgon
04-18-2014, 03:05 PM
I've tried Octane demo before when it was in beta stage. I had a GTX570 and a GTX670 running and octane completely pulled my machine down to a stuttering mode(cursor not reacting, things cannot be moved and crashed). That gave me the impression it needs a very good setup and I felt my 2 cards were not up to it. Now, I've got a Titan I can consider Octane but I guess it's still best to have 2.

Anyway, I'm having issues now installing the demo. I get this message(attached) when I launch LW 11.6.2. I put the octane.dat and octane.dll in LightWave/Bin according to instructions.

121513

Are you sure that you are using the plugin file and the Octane.dll and Octane.dat from the same .zip file? ... this error sounds to me like you are mixing a new .p plugin file with old Octane.dll or Octane.dat

Could be possible for you test the x64 demo version? ... I think that you are using the x32, but anyway it should work.

-Juanjo

andyhow
04-18-2014, 03:30 PM
My main unit is in the workshop(motherboard issues). Currently using a standby 32bit machine.

Yes, I'm very sure I used the 32bit version of everything. I just recopied everything and it still happens. Never mind.

I'll try again in a month's time when the main unit is ready.

andyhow
04-18-2014, 03:48 PM
Are you sure that you are using the plugin file and the Octane.dll and Octane.dat from the same .zip file? ... this error sounds to me like you are mixing a new .p plugin file with old Octane.dll or Octane.dat

Could be possible for you test the x64 demo version? ... I think that you are using the x32, but anyway it should work.

-Juanjo


Putting the WRONG files (x32 vs. x64 gives different errors). I am very sure your x32 files are wonky.

Sebasvideo
04-18-2014, 03:57 PM
I'll try again in a month's time when the main unit is ready.

If you don't mind me asking, why are they taking one month to fix your motherboard? Seems a bit long.

juanjgon
04-18-2014, 05:16 PM
Putting the WRONG files (x32 vs. x64 gives different errors). I am very sure your x32 files are wonky.

Thanks, I will check the demo .zip ASAP. Perhaps as you say there is a problem with the x32 demo files.

-Juanjo

juanjgon
04-18-2014, 05:56 PM
Putting the WRONG files (x32 vs. x64 gives different errors). I am very sure your x32 files are wonky.

You are right. The x32 plugin demo version is broken.
Please, download the x32 plugins from this link, replace your .p file with the new one and try to install it again:
http://www.genesisvisual.com/tmp/demo_x32.rar

-Juanjo

spherical
04-18-2014, 06:35 PM
Octane supports all the GPUs available in your system for rendering. The GPUs should not be configured in SLI. The memory you get for rendering is the GPU memory available in the GPU with the smaller memory size ... for example if you have a Titan with 6GB and a 780 with 3GB rendering together you are going to have only 3GB for rendering ... if you enable only the Titan for rendering you are going to have the full 6GB of free RAM.

This one I have to ask about. If I understand the above first scenario, only the memory on the lesser card is used. All of the memory on the greater card goes to waste.

What about Two 6GB Titans? You would not get 12GB to work in, only 6GB? If that isn't the case and you do get 12GB, wouldn't the first example be 6GB (3GB on each card—limited by the lesser card amount, but spanning both)?

juanjgon
04-18-2014, 06:47 PM
This one I have to ask about. If I understand the above first scenario, only the memory on the lesser card is used. All of the memory on the greater card goes to waste.

What about Two 6GB Titans? You would not get 12GB to work in, only 6GB? If that isn't the case and you do get 12GB, wouldn't the first example be 6GB (3GB on each card—limited by the lesser card amount, but spanning both)?

With two 6GB Titans you get 6GB of available RAM to work with Octane. Each GPU store it's own copy of the Octane scene, and it doesn't share it's memory with the other GPUs.

With two Titans you get twice rendering performance, but not twice memory.

-Juanjo

andyhow
04-18-2014, 07:10 PM
If you don't mind me asking, why are they taking one month to fix your motherboard? Seems a bit long.

Usually it takes a month or two for Asus to replace unit here.

JBT27
04-19-2014, 07:31 AM
Well, I bought Octane this past week, after testing it for a few days, and running it on a single Titan Black - impressive, most impressive.

Julian.

tyrot
04-19-2014, 02:29 PM
you start using Octane with TITAN :) you are a lucky man - my friend:) ...

JBT27
04-19-2014, 03:48 PM
you start using Octane with TITAN :) you are a lucky man - my friend:) ...

Well, yes ... and somewhat broke after buying new kit :) But, Octane is a revelation, it really is. Sure, not good for everything, useless for some things even, but that's understood; it actually is a game changer. Although I intend to keep this machine for some years, this is clearly a major consideration in future machine builds. I'm still getting to grips with it, but it's a bit like those first days using CG gear over 20 years ago - it is that exciting ... I hope that doesn't sound too lame or fan-boy :)

Julian.