PDA

View Full Version : new rederning engine?



ufo3d
06-07-2005, 04:01 AM
I heard a rumor that newtek is rewriting the render engine, is it real? any further infomation?

Captain Obvious
06-07-2005, 04:31 AM
I don't know if they're rewriting it from scratch, but they're definitely working hard on it. NewTek recently hired the main developer of ElectricImage's render engine. I doubt he'll be working on the UI or the file I/O. ;)

ufo3d
06-11-2005, 11:42 AM
It sounds good, but after vewing the gallery pages of EI offical web site (http://www.eitechnologygroup.com/community/gallery_characters.html) , I am a bit worry about it.

Para
06-11-2005, 12:02 PM
Rewriting the render engine is inevitable, let it be EI-enhanced or not. I hope they find a way to get "best of both worlds" to the semi-new renderer.

Darksuit
06-11-2005, 12:17 PM
a new rendering engine should be quite intresting to say the least.

However I am rather surprised that I have not heard or seen anyone do a plugin for ATI's render monkey...

http://www.ati.com/developer/rendermonkey/news.html

has some rather cool special effects and such... would be kinda neat to see Newtek incorperate some of the things in Rendermonkey with a new render

Captain Obvious
06-11-2005, 12:51 PM
It sounds good, but after vewing the gallery pages of EI offical web site (http://www.eitechnologygroup.com/community/gallery_characters.html) , I am a bit worry about it.
The quality of a rendering is more dependant on the skill of the creator than the quality of the renderer, I'd say. At least with "fake everything" renderers like EI's Camera or Lightwave's.

-NG-
06-11-2005, 02:52 PM
Lw's render engine is still strong IMO. I think what makes a good render is the way it is lit, lighting can bring out the most subtle and realistic detail.

Captain Obvious
06-11-2005, 02:57 PM
Lw's render engine is still strong IMO. I think what makes a good render is the way it is lit, lighting can bring out the most subtle and realistic detail.
Definitely, and I'd say Lightwave's renderer is quite good at lighting, for the most part. The radiosity needs work, though, but mostly because it's slow.

ufo3d
06-11-2005, 03:25 PM
Compare with many default renderer, I agree LW's renderer is still good and stable. But , it is also a bit behind other non- default renderer. Current LW renderer , only Spotlight supports shadow map, no subpixel displacement, photonmapping , BSSRDF support, the noise of Area light and radiosity, as well as the speed of radiosity.,etc.
Proton mentioned he was excited about what the team has been doing and can't wait for Siggraph 2005, let's see what rock stuff we will see next month, it will be great if that is render engine improvement. ;)

-NG-
06-11-2005, 03:35 PM
Definitely, and I'd say Lightwave's renderer is quite good at lighting, for the most part. The radiosity needs work, though, but mostly because it's slow.

If you study the way light works and travels, you can fake radiosity excellent in LW, setting it up takes a bit longer, but render times are decreased.:)

toby
06-11-2005, 04:18 PM
It sounds good, but after vewing the gallery pages of EI offical web site (http://www.eitechnologygroup.com/community/gallery_characters.html) , I am a bit worry about it.
:p funny!

But EI was used to render all of the non-organic stuff in SW Episode 1, so it is capable, just more difficult - EI's render pipeline is like 10 years old, 32bit images only, so I'm looking forward to seeing what these guys can do with Newtek -

Lude
06-12-2005, 05:53 AM
ummm
But EI was used to render all of the non-organic stuff in SW Episode 1 - really?

Is there a detail about this somewhere?

toby
06-12-2005, 01:23 PM
uhhh...not sure - but I had an internship at Alex Lindsey's school, he was responsible for the Queen ship and worked with the 'Rebel Unit', a group of artists at ILM who used EI on the Mac. I'll do some googling.

ufo3d
06-12-2005, 01:46 PM
What is the main reason and advantage that ILM chose EI rathan than other renderer, such as renderman or Mental Ray.,etc?
I thought ILM only use Renderman and Mnetal Ray mainly.

toby
06-12-2005, 02:27 PM
EI on the Mac was a big-dog in the late 80's - early 90's, especially when the only other option was Alias on a $100,000 sgi machine, and Alex Lindsey still hails it - god knows why. It's only good for rendering static geometry with a texture on it, and only one texture per object, which was enough back in the day but not now!

Other films done with EI that I know of are Titan A.E., most notably the ice crystals at the end done by Mr. Lindsey, and a jet-fighter movie Behind Enemy Lines.

somnambulance
06-12-2005, 02:37 PM
Hopefully the new render engine will keep up with Nexus. It looks like they have it so each processor renders a square, moving in a spiral out. So dual processors actually get used.

hrgiger
06-12-2005, 02:44 PM
I heard a rumor that newtek is rewriting the render engine, is it real? any further infomation?

God I hope not. They should be working on things that are important. Like openGL, and real-time animation and subpixel displacement to name a few.

Para
06-12-2005, 03:41 PM
God I hope not. They should be working on things that are important. Like openGL, and real-time animation and subpixel displacement to name a few.

Anything that makes the rendering slower isn't "important". PLD was important because it made rendering faster but I really doubt subpixel displacement would make it render faster.

somnambulance
06-12-2005, 04:02 PM
Anything that makes the rendering slower isn't "important". PLD was important because it made rendering faster but I really doubt subpixel displacement would make it render faster.

Um... important to who? If I could have subpixel displacment and and a better real time render, I could live with the render times that LW kicks out now.

toby
06-12-2005, 07:01 PM
Faster Radiosity is pretty f'ing important. As we've seen, radiosity can be much faster than LW's, and I know of studios that are trying other renderers because of that.

somnambulance
06-12-2005, 08:43 PM
Faster radiosity would be nice. I would rather have other things first though. You have different needs than I do. Having a better real time render would be better for me. It would cut back on the test renders I do. I have a video card in my computer, Lightwave doesnt use much of it.

Para -maybe using up some of the video card would releave the processor and increase render time? Subpixel displacement does speed up render times, not to mention modeling time, where it is used.

toby
06-12-2005, 08:55 PM
Faster radiosity would be nice. I would rather have other things first though. You have different needs than I do.

Agreed - I'm just trying to think in terms of what Newtek thinks they should do and why they might have hired the EI guys.

Thanks for the compliment a few weeks ago, btw.

somnambulance
06-12-2005, 09:50 PM
I am more impressed by what is in NewTek's gallery for the quality of renders compared to EI's gallery. Who know's, maybe they will make a good team together.

Toby, I went to check out your reel again, I am keeping it on my desktop now for when I need some insperation. I like the idea of pulling footage too. ;)

toby
06-12-2005, 11:58 PM
I am more impressed by what is in NewTek's gallery for the quality of renders compared to EI's gallery. Who know's, maybe they will make a good team together.

Toby, I went to check out your reel again, I am keeping it on my desktop now for when I need some insperation. I like the idea of pulling footage too. ;)

I just posted a new reel, did you see that one?

Putting 3D into footage you find on the web or in dvds is a really good way to get into the industry.

I too am really hoping that whatever they add to LW, it doesn't come out too EI-ish.

Gui Lo
06-13-2005, 04:47 AM
The issue of radiosity times is a bit mute for me since they can be sped up dramatically by Excluding certain objects from the Radiosity calculations. I've been able to use it in some TVC's by only using Radiosity on the objects that really need it.

A scene of a family fishing in a boat rendered in about 30-40 sec per frame on my Amd64 3800. That's with Shadows, reflections and of course radiosity.

Maybe this is a common tip, but I use radiosity for bounces and colour bleeding only. I leave Shadows to Spotlights or Area Lights.

Until we get realtime rendering it will always be our enemy. It can be reduced significantly by optimising what objects has Radiosity or Shadows or the resolution of the Area lights, ect.

I think this is common knowledge but for new users the LW radiosity renderer is very variable.

Para
06-13-2005, 06:14 AM
Para -maybe using up some of the video card would releave the processor and increase render time? Subpixel displacement does speed up render times, not to mention modeling time, where it is used.

Possibly although the modern GPU:s can't yet create new vertices on the fly very efficiently so the speed-ups wouldn't be that big and most of the new features are tied to DirectX/Direct3D so I'm not sure if LW which uses OpenGL heavily would get the benefits of modern GPU:s that easily (I'm not saying it wouldn't work!). I'm a big fan of hardware rendering so that'd make me a very happy camper if I could utilize my GPU to render something :)

ericsmith
06-13-2005, 09:56 AM
The issue of radiosity times is a bit mute for me since they can be sped up dramatically by Excluding certain objects from the Radiosity calculations. I've been able to use it in some TVC's by only using Radiosity on the objects that really need it.

How do you exclude radiosity from some objects in a scene? Do you have to render multiple passes, or is there some setting that you can turn off for individual objects?

Eric

edit- nevermind, I just found it. I never noticed that radiosity was one of the options that you can exclude along with individual lights. I've got to give that a try.

ufo3d
06-13-2005, 10:27 AM
How do you exclude radiosity from some objects in a scene? Do you have to render multiple passes, or is there some setting that you can turn off for individual objects?

Eric

edit- nevermind, I just found it. I never noticed that radiosity was one of the options that you can exclude along with individual lights. I've got to give that a try.


In layout, Object Properties panel->Lights Tab->Exclude Radiosity

wacom
06-13-2005, 10:46 AM
I'm all for them working on the render engine. They should improve the OpenGL though, but I doubt they'll be able to compete with this one:

http://www.cgtalk.com/showthread.php?t=248382

Very impressive indeed. I was a little disapointed though that there was no Mac support, but I'm not sure of the reasons (maybe the new announcement?). Anway some game makers are using the beta version of it and speak mounds of good things about it.

I'm interested in EI's AA solutions- that would be great for print and HD from what I've heard.

ericsmith
06-13-2005, 10:49 AM
I just tried this, but it appears that excluding radiosity from an object also excludes it from receiving radiosity illumination. Is there some way to only exlude an object from being calculated in the light bounce, while still receiving radiosity illumination bounced from other objects?

eric

Captain Obvious
06-13-2005, 11:01 AM
I just tried this, but it appears that excluding radiosity from an object also excludes it from receiving radiosity illumination. Is there some way to only exlude an object from being calculated in the light bounce, while still receiving radiosity illumination bounced from other objects?

eric
You'll have to render it in a separate pass, I think.

Gui Lo
06-13-2005, 12:01 PM
Radiosity on objects will always make a difference but on certain objects the radiosity effect can be 'faked' using a much quicker technque.

For instance I had radiosity applied to the water in a TVC and it took ages. I excluded it and it was much quicker but the water looked darker. The solution was to increase the objects Dessolve level. Also I excluded trees from a forest scene. Of course the trees did not look as good but the render was much, much faster and the characters still had radiosity applied so in general the scene looked good compared with using no radiosity.

I tend to render a frame with Radiosity fully applied and refer to that image to mimick major shadows and bounces in complicated objects. I try to leave the main characters with radiosity fully applied.

Hope this helps

Gui Lo

ericsmith
06-13-2005, 02:24 PM
Yeah, I get it.

It's kind of a bummer, because at Siggraph last year the guys that did shrek 2 were showing this idea of a "bounce light shader" where you could make individual surfaces bounce light. It made radiosity solutions much quicker. I was hoping that there was something like this hidden in LW all along.

Eric

toby
06-13-2005, 03:01 PM
The Modo radiosity demo showed something like 2 million polys render with full AA in less than a minute. I just tested LW with 200k polys and with no AA it was 4 minutes. New renderers are several times faster than LW, there's no amount of optimizing that will save you that much time. I think Newtek knows this and are working on it, because they don't want to lose major studios as customers.

Captain Obvious
06-13-2005, 03:37 PM
It wasn't Modo, it was just a technical demonstration. Anyway, they obviously chose a scene that gave as flattering a result as possible. Their renderer will be dog slow at some scenes too, I assume. Maybe just not as slow. ;) I refuse to believe that they came up with some magical algorithm that is capable of rendering anything and everything in a fraction of the time every other renderer takes.

Para
06-13-2005, 03:45 PM
One way to speed up the renderer is to cache and pre-calculate as much as possible. For example one of the simplest optimizations in any program is that when sin function is used it is not calculated every time when the program hits the part where it should do it but instead the program just reads from a predefined table the correct value for the sin (... :rolleyes: :D).

-NG-
06-13-2005, 04:03 PM
Yeah, I get it.

It's kind of a bummer, because at Siggraph last year the guys that did shrek 2 were showing this idea of a "bounce light shader" where you could make individual surfaces bounce light. It made radiosity solutions much quicker. I was hoping that there was something like this hidden in LW all along.

Eric

If i remember correct there has been a post at the cgtalk forum regarding faking this effect with the use of gradients in LW.

somnambulance
06-13-2005, 04:13 PM
It wasn't Modo, it was just a technical demonstration. ... I refuse to believe that they came up with some magical algorithm that is capable of rendering anything and everything in a fraction of the time every other renderer takes.

It was their development something or other called Nexus. From what I understand it is magic. :rolleyes: From what I remember there werent any colors, so there probably werent any shaders or surface materials involved either. The first scene they render is one from Christophe Desse, you can see it here (http://www.spinquad.com/forums/showthread.php?t=6081). I didnt see any animation demos from them yet either but rumors in the wind say that it is in the works.

sysrpl
06-13-2005, 04:15 PM
Possibly although the modern GPU:s can't yet create new vertices on the fly very efficiently so the speed-ups wouldn't be that big and most of the new features are tied to DirectX/Direct3D so I'm not sure if LW which uses OpenGL heavily would get the benefits of modern GPU:s that easily (I'm not saying it wouldn't work!). I'm a big fan of hardware rendering so that'd make me a very happy camper if I could utilize my GPU to render something :)

Uhm, I write sofware for both OpenGL and DirectX, and I can tell you that anything you can do with modern GPU's using D3D can also be done with OpenGL.

The differences are:

D3D uses API versions to add new features, while OpenGL uses extensions. D3D versions simplify what can be done by providing a specification from a single entity, while OpenGL's extensions allow the driver to add features anytime they are updated. This means that OpenGL's functionality may be greater that D3D's at times, but at the expense of developer effort.

Both APIs allow full access to using programable rendering on the GPU. D3D's primary interface for this is HLSL, while OpenGL's is GLSL. Currently GLSL works (it is supported by RenderMonkey), but it has some 'querks' as it is up to each individual driver to implement their own compiler.

Both API's can fully use hardware based programable rendering with the Cg language giving programmings a cross API choice of higher level shading languages.

And of course the biggest difference between the two APIs is tthat one is tied to a single entity and runs only on Windows (xbox is not a platform used for 'computing'), and the other is open and runs on Windows, Linux, Mac, BeOS, SGI ect.

Captain Obvious
06-13-2005, 04:28 PM
It was their development something or other called Nexus. From what I understand it is magic. :rolleyes: From what I remember there werent any colors, so there probably werent any shaders or surface materials involved either. The first scene they render is one from Christophe Desse, you can see it here (http://www.spinquad.com/forums/showthread.php?t=6081). I didnt see any animation demos from them yet either but rumors in the wind say that it is in the works.
As far as I've understood it, Nexus is just their sandbox. They can add modules to it in order to try them out. The renderer was one such module.

In the scene they demoed in Paris last fall, they used both instancing and texturing, rendering well over a billion polygons with radiosity in about half a minute. Sure, it was instancing, but still fairly impressive.

somnambulance
06-13-2005, 04:35 PM
Careful, this thread is starting to lean towards mutiny!

Captain Obvious
06-13-2005, 05:05 PM
Careful, this thread is starting to lean towards mutiny!
How so? Pointing out that the competition is ahead of NewTek in some areas is not mutiny. Pointing out that the competition will tweak their benchmarks to appear more ahead than they actually are is not mutiny, either. ;)

somnambulance
06-13-2005, 06:00 PM
I say it is getting close, and Chuck doesnt like that. Once upon a time I pointed out a few things about Modo that LW doesnt have and the thread was removed.

Nu Visual Scien
06-13-2005, 06:15 PM
EIs renderer is ridiculously fast and does nice renders too, The ret of the package is a bit errrrm well it's great for solid object stuff but crap for characters, the render engine is cool though, I think if they can get the speed of that combined with quality renders and stability then we could be on to a supreme winner here, I have seen a lot of hype about this renderer and that renderer and i still like the output from LW.

tektonik
06-13-2005, 08:12 PM
mmm fprime lovin' here

just wish everything was 64 bit so i could render away a few billion polys...

also wish it could bake light into textures fith hdri floating point info for tweaking purposes

Gui Lo
06-13-2005, 09:46 PM
New renderers are several times faster than LW, there's no amount of optimizing that will save you that much time.

Sorry I disagee, I find optimizing can save a lot of time and can easily reduce rendering to at least 30% with just a few minutes work. I have litterally saved hours on some animations!

It is no longer a matter of setting up the scene, hit f10 and hope. The rendering is now as much of the design process as the scene setup and usually determines how objects are modelled and scenes are setup.

Some methods I use include rendering bump maps(with all other objects rendered Unseen by Camera) and then re-applying as Front Projection. This works really well with large flat areas, and can really speed up renders. Another is to reduce the Area Light resolution and use Noise Reduction. Oldie but goldie is the reduction of Ray Recursion Limit. If there is no reflection then this can be set to 0 and Area Lights still give a shadow, even with Raytraced shadows switched off.

Again I say that selectively using radiosity by Excluding some objects can have a huge effect on the rendering times. transparent or highly reflective objects do not benefit from radiosity. Also highly textured areas sometimes do not benefit from radiosity. In short I use radiosity simply for surfaces that show a marked benefit from bounced light or colour bleed.

Gui Lo

Hervé
06-13-2005, 11:58 PM
... he he... pretty soon, even Maxwell is going to be faster... :D

toby
06-14-2005, 02:01 AM
You disagree that 5 years of software development can speed up render times? ok :confused:

LW's radiosity came out with 6.0 in 1999, one of the first if not the first. There's been signifigant improvement in radiosity rendering since then, as you should imagine. Things like saving radiosity solutions, and re-calculating only areas that are needed.

Please read the test I quoted again. LW took 4 times longer, with no AA. I did the same test with medium AA, turned down the rays to 3x9, and it took 29 minutes - compared to 1 minute. And still much grainier. You really think that's irrelevant?

I know about optimizing, I do it more than anyone I work with, because I hate waste. On a recent show I cut down a co-workers' render times from 1 1/2 hours per frame to 20 minutes. And he had already excluded objects from radiosity, at my suggestion. What I can do with LW doesn't really matter. If the studios are looking to other solutions then Newtek can either improve or be left behind, and I'm betting they're working on the former. And now this thread really may get locked -

ufo3d
06-14-2005, 03:18 AM
You can save a lot of time by optimizing , but what if a renderer can render the same scene as fast as yours, but no optimizing. And what will happen after "turn-on" optimizng? What we can do after optiziming? It all depends the renderer.
As you said, the optimizing only can be used, but only in some situations. I also use those methods you said. But sometimes you cannot use it, or u may only use it until everthing such as lighting , motion , shading is confimed by the clients, sometimes.
I agree we can speed up the rendering time by optimzing, but I don't agree we can save at 30% by optimizing, always.
I think Newtek also knows the problem and that is why they hired a guy from EI. ;)

Para
06-14-2005, 03:34 AM
Uhm, I write sofware for both OpenGL and DirectX, and I can tell you that anything you can do with modern GPU's using D3D can also be done with OpenGL.

The differences are:

D3D uses API versions to add new features, while OpenGL uses extensions. D3D versions simplify what can be done by providing a specification from a single entity, while OpenGL's extensions allow the driver to add features anytime they are updated. This means that OpenGL's functionality may be greater that D3D's at times, but at the expense of developer effort.

Both APIs allow full access to using programable rendering on the GPU. D3D's primary interface for this is HLSL, while OpenGL's is GLSL. Currently GLSL works (it is supported by RenderMonkey), but it has some 'querks' as it is up to each individual driver to implement their own compiler.

Both API's can fully use hardware based programable rendering with the Cg language giving programmings a cross API choice of higher level shading languages.

And of course the biggest difference between the two APIs is tthat one is tied to a single entity and runs only on Windows (xbox is not a platform used for 'computing'), and the other is open and runs on Windows, Linux, Mac, BeOS, SGI ect.

To be short, that's what I meant. D3D is "easier" and by tying I mean that GPU:s are generally developed with the next version of DX/D3D in mind because it's the mainstream API for games which brings big bucks to the IHV:s.

Gui Lo
06-14-2005, 04:27 AM
Hi Toby,
I disagree that "there's no amount of optimizing that will save you that much time."

I think that optimizing can help to improve render times a lot.

As you say in your example "On a recent show I cut down a co-workers' render times from 1 1/2 hours per frame to 20 minutes".

That is a huge reduction in render time through optimizing.


I am not saying that LW has the fastest renderer but it does have one of the best looking. Also I am trying to give a balanced view to this thread. New users may want to use Radiosity but are disuaded because of the seemingly endless render times. There are ways around this and I want to give some suggestions.

Gui Lo

toby
06-14-2005, 04:56 AM
When I said "won't save you that much time"
I didn't mean "won't save you very much time",
I meant that it won't save as much time as the newer renderers will.

Gui Lo
06-14-2005, 05:04 AM
Oh, my mistake then.

Well, I think were not that far from a realtime renderer and it looks like LW may be one of the first.

Para
06-14-2005, 09:31 AM
Oh, my mistake then.

Well, I think were not that far from a realtime renderer and it looks like LW may be one of the first.

"LW may be one of the first"...what, do you know something we normal mortals don't? :eek:

ufo3d
06-14-2005, 10:23 AM
will we see a big impact in next update, perhaps?

Chuck
06-14-2005, 11:11 AM
I say it is getting close, and Chuck doesnt like that. Once upon a time I pointed out a few things about Modo that LW doesnt have and the thread was removed.

It all just depends on the context, intent and apparent result, Som. Please also understand that moderation policies are not just my personal fiat, so portraying things as though that's the case is really not reflective of reality.

Apologies for the off-topic interruption, folks!

Lude
06-14-2005, 11:29 AM
hehehehehehe :p

kjl
06-14-2005, 11:43 AM
Well, I think were not that far from a realtime renderer and it looks like LW may be one of the first.
Heh, no such thing as a realtime renderer ;) Appetite for complexity will always increase render times back up to whatever your pain threshold is. We can render scenes from Tron in realtime, but we don't, preferring to make the images look better with a gazillion polygons, subsurface scattering, radiosity, and whatever the latest very expensive technique is (e.g. Maxwell's wavelength model of light propagation).

Gui Lo
06-14-2005, 08:00 PM
Yes, at the moment it does depend on the complexity of the scene, models, rendering, etc .

Although, with the advance of Fprime and stuff like Pimp I think LW will soon be working at a similiar speed to say Photoshop.

Also as I understand it, when we get the multi cores we may be able to do the rendering as a background task so in most cases it may not be needed to render when the scene is done.

Gui Lo

Para
06-15-2005, 01:10 AM
Yes, at the moment it does depend on the complexity of the scene, models, rendering, etc .

Although, with the advance of Fprime and stuff like Pimp I think LW will soon be working at a similiar speed to say Photoshop.

Some consider Photoshop as a slow 2D program. Besides the complexity of scene will always cause a hit to render time.


Also as I understand it, when we get the multi cores we may be able to do the rendering as a background task so in most cases it may not be needed to render when the scene is done.

But if you change something, the rendering must be started again (just think of FPrime) so rendering as a background task isn't very feasible, multicore or not.

Gui Lo
06-15-2005, 02:07 AM
Hi Para,
Photoshop is fast enough for all the users I have known.




But if you change something, the rendering must be started again (just think of FPrime) so rendering as a background task isn't very feasible, multicore or not.

Again, that is taking rendering as a sequencial process.
IMHO the GPU technology will be used more in future, so it may be possible for 3d apps to break out of the frame by frame nature of current rendering. For instance each element of the render may be independent i.e the texture, models, shadows, reflections, radiosity, etc may be kept seperately. These are then combined to give the image. So maybe as the app gets used only relatively small changes are made to the code used to make the image.


I'm just guessing but rendering is so costly and there are so many smart people trying to get realtime reality. Also the advances made recently in hardware and software it feels like it will be soon.

Gui Lo

Bigboy
06-15-2005, 10:50 AM
Just to throw in my $0.02....

They need to address the motion blur. Its a bit pants, and yes... with multi-core CPU's being the future of all PC's, they need to support multi-core much better.

AMD+Intel have released server chips with 2 cores on die, and they will only get bigger. in "x" years, you might well have 8,16 or 256 etc..

If they start to support these now, then it'll mean less work in the future when new "features" are added.

Darksuit
06-17-2005, 11:40 AM
Has anyone had a chance to see what the render times are like with these guys...
http://www.maxwellrender.com/

Some of the renders in the gallery look fairly nice...

Captain Obvious
06-18-2005, 05:41 AM
with multi-core CPU's being the future of all PC's, they need to support multi-core much better.
Lightwave has full support for multiple cores, at least in the renderer department. No real need for fixes there, since nothing is broken. You can render up to eight threads at once, or more even with SN. Not many people have quad dual-core HyperThreading Xeons. Eight threads is enough, at least for now. It shouldn't be too hard to change this once necessary, but it won't be for another couple of years.

Pavlov
06-28-2005, 04:57 AM
Regard optimization: yes, it's useful, and knowing how to tweak a scene is always a must. BUT it cannot be anywhere a substitute of better rendering solutions. I'm doing heavy viz works, like 5-10 min movies of arch stuff, with interior, exterior and hibrid situations. In this situation, there's no trick or technique which could barely compare to adding Kraytracing to my arsenal. With it, i can render a 100 bounce radiosity solution and bake just indirect illumination on texture in a matter of hours, max in a day. Then i add direct lighting in LW and i get something like 4 min/frame rendertime.
I could have faked, you say ? Nope at all, hundred of lights were not enough for this, including interior, exteriors, underlevels, open covered spaces, artificial lights and so on. LW's radiosity with microwave, maybe, but it would have needed almost a week for setup and bake with 2 bounces, and i'd get a much inferior result.

Paolo Zambrini

Hervé
06-28-2005, 05:41 AM
in the next version, Maxwell will have some baking features.... ;)

Pavlov
06-28-2005, 06:09 AM
in the next version, Maxwell will have some baking features.... ;)

Good to know, having multiple choice is always better.
Maxwell gove good results, but it's WAY too slow and badly integrated.
I'm focused on Fprime and Kray; If Fprime could have a baker feature, we could have rough texures in a few minutes so we could begin animating with these applied, while Fprime refines them... the best.
Kray is incredibly fast - where "incredible" means 10, 20, 50 times faster than LW. Having 100 bounces in PAL D1 in 2-3 minutes has not possible comparison with LW. Its baker is great, it features direct light exclusion (previously seen only in Lightscape, regarding LW-related engines). By now, there's nothing which can compare with Kray in terms of productivity in Viz movies.
Maxwell has still a *long* way to go, imho. When it will output a print rez interior in 10-20 hours like fprime i'll give a look at it again. Even then there will be better solutions - i'm fairly photon-mapping biased - you have to test Kray in depth to understand the difference. I'm not using F9 for a while ;)

Paolo Zambrini

Hervé
06-28-2005, 06:17 AM
I'm not using F9 for a while ;)

Paolo Zambrini


I have to say me either...

as for Kray, I hear you Paolo... I hear you, we're just not on the same market... no movies here.... ;)

JML
06-28-2005, 06:27 AM
with the beta, maxwellrender got a lot faster, but the main thing is that when you work with it, every surfaces has to have maxwell settings..

so if you work on a project and decide to use maxwellrenderer, you have to stay with it until the end of the project.
you won't be able to change at the last minute and render it with Fprime or LW...
(if your boss tells you it's too slow and need to render it with fprime or LW at the last minute, you won't be able too unless you re-texture everthing and light..)

still though, I can't wait to see the final version..

Hervé
06-28-2005, 06:30 AM
I hear you JML... :D ;)

Pavlov
06-28-2005, 06:43 AM
I have to say me either...

as for Kray, I hear you Paolo... I hear you, we're just not on the same market... no movies here.... ;)

Undestood ;)
BUT... Kray is a killer even for static images. It has still a lot to catch (no procedurals, no shaders, no MB and therefore no soft shadow from spinning lights), but i try this: bake diffuse with kray in your interior, then render in LW with all the whistles. Unbeatable, quality wise and time-wise.
I'm a bit concerned about this - Maxwell seems to be at the centre of attention now, while other much more productive tools are almost unseen.

**** marketing issues, i guess...

Paolo Zambrini