PDA

View Full Version : Radiosity: Fix it, forget everything else



laddehlingerjr
11-16-2005, 04:38 PM
I just bought a brand-spanking new Alienware 64 bit machine. Pertinent specs: Two Dual-Core Opteron 2.4 GHz procs, 6 gigs of ram (yes, 6, s-i-x), 64 bit LW running on it, 64 bit Windows running.

So, okay, this is no slouch of a machine. I bought it primarily to see if radiosity in 8.5 64 bit was improved to the point of useability over 7.5 or 8.

From a purely subjective viewpoint, the answer is a definitive NOPE. Not yet. No way. It still takes too long.

So instead, I will have to bake out (or actually, Microwave) stuff into illumination maps and apply them to a new object. Or else fuss around with the crummy built-in "computer" lights to "fake" radiosity. From a man-power perspective, this seriously sucks. Sucks as in it costs me a lot of money. Sucks as in, it means I can't put as much money into buying new copies of Lightwave. And I'm sure it impacts other people's abilities to purchase. Don't get me wrong, I'm rooting for the good guys - I've spent a lot of years & lots of capital expenditures in LW & don't want to dump it for Max / Mental Ray. And I don't want to see LW end up in the dustbin of history. I'd lose a lot of capital investment that way.

So I'd like to hear everyone's viewpoints on the following. Please, architectural folks, chime in. And hey, Deuce, Chuck, et. al. - can you let us know what Newtek's position on its future radiosity plans are?

To my mind, there's a very simple way for LW to get back to the forefront of the render packages. It requires just three things:

1) The ability to save out the radiosity solution in the scene file or in a separate file, like .mdd files or .pfx files, and re-use them. Even let us blend two or three together, so you could do three or four 4x12 monte-carlos and blend them together for additional quality. This could be a quick hack for us users out here in the real world to network render radiosity solutions.

2) Multi-threaded radiosity calculations. (from what i can tell using task manager, radiosity is NOT multi-threaded right now. Is this true? If so, bad Lightwave, bad! What's up with that??? 8 threaded radiosity render = single threaded dud)

3) The ability to use multiple computers on a network to generate a radiosity solution (and of course, render out a full still image without the hacks required by tequila, or screamer net, etc.). REALLY IMPORTANT. REALLY. REALLY. REALLY. THIS WOULD BE THE LYNCHPIN TO GETTING LW OUT OF THE BACK OF THE CLASS.

To my mind you could toss everything else you guys have planned out the window. Nothing else matters but making radiosity viable. The world expects realistic lighting from the 3D artist, and really, the easiest way for us to accomplish that is with a radiosity system. IF it was fast enough. But it ain't!

Anyone else agree with me here?

jeremyhardin
11-16-2005, 05:15 PM
agree that these are good ideas.

disagree that all other plans should go out the window. Character animation in LW is needing attention as well as many other things. I can see your point from an arch-viz perspective, but not all of us share that perspective ;)

laddehlingerjr
11-16-2005, 05:58 PM
Character animation in LW is needing attention as well as many other things. I can see your point from an arch-viz perspective, but not all of us share that perspective ;)

Actually I'm coming from a military-animation view point (battles, tanks, soldiers, more), which requires everything LW can do and more: character animation, scripting, UV mapping, and more. The kind of work I do runs the full spectrum. To my mind, better character animation tools don't do much good if, when rendered, the image either:

1. Takes forever (because of radiosity) or
2. Looks terrible (because you had to fake radiosity) or
3. Looks terrible (because you had to fake radiosity) AND takes forever (because you've had to fake it with tons of lights) AND is a huge drain on your pocket book (because you've had to fake it with tons of lights and lots tons of time on manpower).

Also, surfacing becomes a nightmare when you don't know how the surfaces are going to look in a given lighting situation and you have to render to tweak them. That holds true for character work as well as architectural or what I do. And FPrime is not a solution for faster renders - it just lets you guess what the final output MIGHT look like, if you can get it to play nicely with your flavor of LW.

Has anyone at Newtek ever done a serious marketing study as to how many of their customers actually USE character animation on a constant basis? Has anyone done a marketing study as to which features are actually important to their user base? If not, why not? And not a silly internet poll, I mean a real marketing focus group & the full works.

Anyway, :lightwave

Hoopti
11-16-2005, 06:02 PM
While Arch/previz is a large income producer for us, I don't believe that Radiosity is the one thing I'd put to the top of the list. We make sure we build in enough time for the final render into our bid.

There are many other things I'd like to see fixed before this, so while the ideas are outstanding, I think there are other things that need consideration.

Part of the problem with discussions with this tone is the fact that lightwave is a multi-disciplined application, its kinda like the German Sheperd of the 3d world. The GS is not known as the best scent hound, the best site hound, the best rescue dog, the best protection dog, etc. If you were looking for one breed to do one of those things, there's a breed better than a sheperd. But if you're looking for a breed to do many of those things at once, then it's the breed to go with. Hence the police in the US using so many, they're great in alot of those areas, although not the best.

LW is the same way, it's great in alot of areas, but not the best in a specific area. That's what's allowed the flurry of plugins, both commercial and free, to allow us users to fine-tune our setups to meet our specific needs. For example, if you want LW to run with Renderman, then go buy the plugin. I don't so I don't want to have to pay for the production costs associated with making it that way. It's an outstanding foundation, and I rue the day that it decideds it wants to become the best out of the box 3d solution for any one discipline. That means it'll have to forego the other areas to meet that need. If they do, I hope it's something I need it for.

Hoop

jeremyhardin
11-16-2005, 06:39 PM
To my mind, better character animation tools don't do much good if, when rendered, the image either:

1. Takes forever (because of radiosity) or
2. Looks terrible (because you had to fake radiosity) or
3. Looks terrible (because you had to fake radiosity) AND takes forever (because you've had to fake it with tons of lights) AND is a huge drain on your pocket book (because you've had to fake it with tons of lights and lots tons of time on manpower).

Also, surfacing becomes a nightmare when you don't know how the surfaces are going to look in a given lighting situation and you have to render to tweak them. That holds true for character work as well as architectural or what I do. And FPrime is not a solution for faster renders - it just lets you guess what the final output MIGHT look like, if you can get it to play nicely with your flavor of LW.


Still, that's your opinion of the situation. And others who use LW to it's full potential for the full spectrum...might still vote for something other than a better radiosity solution while forgetting everthing else :D .
Because others seem to get on just fine with things as they are radiosity-wise, and they AREN'T getting on just fine with character tools, lack of pass management, memory management problems, etc.

I think your calculation and multithreading ideas are brilliant. But there are a million work-arounds for no or poor radiosity. There aren't a million workaround to the rigging problems you encounter in LW. Or many of the other things that need attention.

Exception
11-16-2005, 07:31 PM
I'm a full time architectural visualiser, and I get by allright with lightwave's radiosity, although it could use some improvement. For stills, I can pretty much beat anyone with any software package, including fprime, on radiosity speed with the interpolated setting and using the motion blur trick. That works great to the point of being useable for animations, although the jitter is difficult to get out.

Things would be indeed improved a lot by being able to save out radiosity solutions... however, since we have adaptive mesh subdivision coming in [9], I hope that from then on a LightScape radiosity render method will be possible, which will blow out everyone of the water for arch. vis.... if they do it right that is...

I would actually really like to see viewport to vector image dumping like rhino and FormZ do, and ORTHOGONAL rendering! That is for me priority no.1.
(and JJ's special projection engine doesn;t work because it ignores line rendering and a whole bunch of other stuff).

jeremyhardin
11-16-2005, 07:38 PM
I would actually really like to see viewport to vector image dumping like rhino and FormZ do, and ORTHOGONAL rendering! That is for me priority no.1.
(and JJ's special projection engine doesn;t work because it ignores line rendering and a whole bunch of other stuff).

orthogonal and viwport rendering will be in 9.
but i dunno if it will be possible to viewport dump to a vector though.

Captain Obvious
11-17-2005, 12:58 AM
Try Kray or Fprime. Either way, I'm inclined to agree with Exception. Lightwave's interpolated mode is pretty darn good when you use it with motion blur. It also gives you excellent control over the render times. Turn down the settings for a really quick preview, turn up the settings for film-quality.

laddehlingerjr
11-17-2005, 04:38 AM
Again, FPrime is a decent solution for determining what a final still render will eventually look like (if you can get it to play nice with your flavor of LW). Not for speeding up the final output.



I think your calculation and multithreading ideas are brilliant. But there are a million work-arounds for no or poor radiosity. There aren't a million workaround to the rigging problems you encounter in LW. Or many of the other things that need attention.

There are plenty of workarounds for poor radiosity, you're right. The problem is that those workarounds cost just as much money in labor as sitting there waiting for stuff to get squeezed out of a loooong rad render. Ergo, they are not really true workarounds. To the hobbyist or do-it-yourselfer, a workaround that takes more time is a decent workaround; to the businessman, in which time = money, a workaround that costs just as much or more = less money for more lightwave seats, period.

And that's a raw deal for everyone, 'cause less Lightwave seats bought = less $ for Newtek to spend on improvements like, I don't know, character animation tools. Salaries for Newtek programmers. And so on.

I know for a fact that a lot of LW businessmen across the country are very fearful that if the basics of economics continue to be ignored by NT (like doing real surveys on the economics of what customers want & need as opposed to what looks cool on a box or a website), their capital expenditures in learning LW, buying LW, training employees in LW, etc., will go up in smoke. That's coming up and coming up fast...

Last time I had to hire several LW modelers & animators, I tried to hire experienced 'wavers. What an excercise in futility; they didn't exist. Had to train kids out of college who were all users of the competition. 3-5 years earlier that wasn't the case.

LW needs an exponential leap forward against the competition. 2 or 3 times faster rendering doesn't hack it when you're talking the difference between 72 hours vs 24 hours. BOTH times are unnacceptable. The ability to take 24 hours down to 1-5 minutes IS an exponential leap forward and something that might push LW back to where it belongs, at the front of the pack.

Even if taking 24 hours down costs me lots of $ in capital expenditures (computers, networking, even $100-$200 extra per computer for an LW "radiosity" software add-on) you bet your rear I'll spend it! Where do I send the freakin' check? A lot safer for me to spend that money on new computers & infrastructure than spend it on employees to train them on how to use workarounds (and optimize models, and microwave, and bake, and so on). Employees can quit, move on, get sick. You spend 6 months teaching them LW and they almost get semi-productive, then "poof" they are gone. Or the project is over and you have to let them go. And they spent the majority of their time while working on the project doing workarounds.

NOT making the most awesome looking image on the planet.

Which is where an LW user used to be in the Babylon days.

Take a notice of the stats on all of those cool photorealistic looking images in the galleries and works-in-progress areas. 26 hours here, 18 hours there, sometimes 72 or more hours. What an enormous waste of time. Think of these alternatives:

1) 24 hours for a top-of-the-line render, photorealistic.
2) 1 hour for a top-of-the-line render on 24 computers, photorealistic.
3) 20 man-hours to generate a radiosity work around, with a non-photorealistic result that takes 4 hours to render because of the number of lights involved.

All choices take 24 hours. But I'll take number 2. And I'd be a fool not to. The $$ spent on the computers improves my speed forever. The $$ spent on 1 is lost, it doesn't give me added benefit down the road. The $$ spent on 3 is doubly worse, not just because $$ spent on labor is lost, but the because the result isn't as good as 1 or 2. Which means I might lose the next job I'm bidding on to the guy who's using a competing package...

And hey, an important point: the other packages have the same radiosity issues!! Not just the cheap ones, but the expensive ones. Which is why it is so crucial, so ultimately crucial, for LW to get an advantage in this area. You strike the competition at their weakest point first, NOT where they are strongest. Other packages have great character animation tools. Other packages have great UV editing tools. Etc. Other packages do NOT have photorealistic rad rendering that can be exponentially faster than all the others.

Get the high ground in one area, then conquer the other hill tops. Don't keep attacking everything at once from the bottom of the hill. That's futile, and the current position in the marketplace PROVES it.

To Newtek, I say, strike them where it will do you some good! And earn a nice big Katrina-oil-company-sized profit. You guys are in Texas after all.

:I_Love_Ne

jeremyhardin
11-17-2005, 08:57 AM
Again, FPrime is a decent solution for determining what a final still render will eventually look like (if you can get it to play nice with your flavor of LW). Not for speeding up the final output.
Have you used FPrime? You can't go around saying that it doesn't speed up the final output. It's faster, ergo it does speed up the final output.




The problem is that those workarounds cost just as much money in labor as sitting there waiting for stuff to get squeezed out of a loooong rad render. Ergo, they are not really true workarounds. To the hobbyist or do-it-yourselfer, a workaround that takes more time is a decent workaround;
...
3) 20 man-hours to generate a radiosity work around, with a non-photorealistic result that takes 4 hours to render because of the number of lights involved.
I'm sorry then. You must be hiring the wrong people. I and many others can get good lighting that looks just as good as a radiosity in FAR less time.


I know for a fact that a lot of LW businessmen across the country are very fearful...
Thats great. I know a lot around the world that aren't. It seems we've different takes on this too.


Last time I had to hire several LW modelers & animators, I tried to hire experienced 'wavers. What an excercise in futility; they didn't exist. Had to train kids out of college who were all users of the competition.
Hmm. We exist methinks. Did you post here? on Spinquad? Perhaps its a matter of failure to draw them, not so much a matter of them 'not existing.'

Pavlov
11-17-2005, 09:56 AM
i fully agree: NT should really put hands on GI engine and put into 9 something up to industry standards.
Max, Maya and XSI have both Mental ray, Cinema integrates AR 2.5 which can almost compete with Vray.
I'd be kinda knocked down if 9.0 had not something up to these standards, for quality, depth of control and speed.
Much faster GI needed, newer methods. Photon mapping, irradiance caching, final gathering, MLT - all are new GI techs out there, we're still at an old and unefficent montecarlo method. I wonder if anyone had ever used LW's montecarlo in production (people with hundred of machines does not count).
Again, caching is needed, Auto-baking too; XSI can bake to object in a few minutes, in LW you've AT LEAST to have microwave since Baker sucks, baking will take 10x time and hassles, and you will never get even close to a professional GI engine's output.
Beside this, you can use Fprime for static images, it's very fast and looks good; for heavy GI works, difficult setups and professional output needings, you must go for Kray (Maxwell is awesome but very far from being a productuion tool).
Kray is in an early stage but it's several years ahead LW both quality and speed-wise. Its baker is truly fast; you can setup GI, test and bake GI maps in a 1-2 days for a mid-complexity project, and you're ready to animate.
You cant get even close with LW alone, even if you have a lot of time to setup, fake, by-pass problems, find workarounds, and so on.
Waiting for Kray update in december, and for LW 9.
Hope NT took in account that many users (at least all Viz people) are here because of Fprime and Kray, when they sorted their priority list.

Paolo Zambrini

Nitisara
11-18-2005, 07:12 AM
Once I made a scene and rendered it with radiosity. It took 1 min 30 sec to render. I was very satisfied. I showed this image to one guy and proudly told him my render time. He asked me to give him that scene. I gave him, and he imported it in Max and rendered with VRay. Render time was... 13 sec. He smiled and told me: "Welcome to 21st century!" I was quite hurt. Although I can play with render settings and maybe I can reduce render time twice, or even thrice, sacrificing quality a little bit and using tricks, it will never be 13 sec in current version of LW. Pity...

Pavlov
11-18-2005, 07:23 AM
I was quite hurt. Although I can play with render settings and maybe I can reduce render time twice, or even thrice, sacrificing quality a little bit and using tricks, it will never be 13 sec in current version of LW. Pity...

You can; but first you have to work an hour setting up gradients and fill lights, and quality wont be the same anyway.

Paolo

kopperdrake
11-18-2005, 10:42 AM
Try FPrime. For stills I pretty much use nothing else in architecture. For animation, well, I've had to resort back to baking luminosity - I'd like FPrime have the ability to bake out textures, that would make me happy.

You're right about the competetion though, I've just had a client/colleague tellign me to switch to max as it's what they use, then in the following breath he was cursing Vray for its complexity. I couldn't believe how nice and forward FPrime was - I've no time to work out best photon bouncing thingies or anything else for that matter. The client wants a picture tomorrow and doesn't care how you do it.

Dunk

spirit_of_stars
11-18-2005, 05:50 PM
NT said it will be possible to have different renderer in LW9.x. Maybe in 2 or 3 years It will be possible to have Vray in LW. For the moment VRay still for me one of the best rendering software. His displacementis is great and terribly fast. For GI rendering (under production time) LW is outdate. But for standard rendering it do a greate job.

laddehlingerjr
11-19-2005, 06:00 PM
Have you used FPrime? You can't go around saying that it doesn't speed up the final output. It's faster, ergo it does speed up the final output.

You're right. It is faster. Faster like a turtle is faster than a snail.

A speed increment of 2 or 3 times is about the best it does. Its primary purpose is not so much to speed up rendering / radiosity, but to allow one to work and render simultaneously, hence, reducing the number of errors. Which is absolutely valuable, don't get me wrong.

I quote their website at http://www.worley.com:


FPrime is a rendering engine with a unique design focus. Unlike traditional renderers, FPrime is not just a final output step. Instead it is your companion during the entire process of creation from initial setup, through texturing and lighting, keyframing, and of course final rendering. It revolutionizes workflow by giving you constant feedback about your current scene and the effect of changes. You can always just see your render.

And yes, I have used fPrime; I own it and yes, it's an invaluable tool. Same with all of Worley's products, and the guys over there are great.

But again, reducing render times by 1/2 or 1/3 is not very exciting. If the original render takes 48 hours, then a 1/2 decrease is still 24 hours. No, what I'm talking about is the ability to generate near instantaneous renders with full radiosity solutions for still frames.


I'm sorry then. You must be hiring the wrong people. I and many others can get good lighting that looks just as good as a radiosity in FAR less time. Hmm. We exist methinks. Did you post here? on Spinquad? Perhaps its a matter of failure to draw them, not so much a matter of them 'not existing.'

I haven't seen your work, but I will be more charitable then you've been to me, and I will grant for the sake of argument that you are an expert that can get great lighting that looks just as good as radiosity in a reasonable period of time. You are no less than a full master at digital work-around chiaroscuro.

But if I had a system that allowed me to generate radiosity solutions in, say, under one minute using only surface attributes within the model as the sole lighting ("practical lighting" is what they call it in cinematography), I don't think any tricks would be able to beat that in terms of quality or speed. This would reduce such tricks and work-arounds to the Paul Bunyan dust bin of history.

And I am convinced that this speed is entirely possibly with today's hardware technology; the only thing stopping us from doing it is software.

As to your point that I am hiring the wrong people, they may take exception with that. Last time I was hiring was 2 - 2 1/2 years ago. I did not post on Spinquad (really wasn't a big name then). But I did post on this site, Flay, Monster, Dice, the various CG neworks, the LW User's group on Yahoo, and High End 3D, in addition to contacting some folks at Newtek directly. I received probably 15 resumes, but they were all freelance / telecommute / weekenders, hundreds of miles away; unfortunately, due to the nature of the project, I needed on-site full-timers.

For the record, I ended up with about 100 resumes from people with other 3D package skills.

In my general four-state area resume search for people with Lightwave skills on Monster, I got a total of three hits: mine and two of my employees, whom I'd trained five years previous.

At the present time in my general metro area, as far as I can tell, there are a total of 15 competent 'wavers, many of whom I've trained. Of those, only four use LW where they are currently employed. Students at the various colleges are taught M-x and M-ya.

Much as I hate that it's happening, the number of full-time 'wavers across the country is dwindling; this is quite simply a fact, much as I hate that it's happening and sincerely want the trend reversed. I'm sorry if my saying it gets your back up; I certainly don't like it either. I'm particularly angry to watch a near 10-year investment of time & money in LW & Newtek slowly go down the drain.

Going into it the why's and how's of the dwindling too deeply would be outside the scope of this thread; suffice to say, no sense in looking back. Point is to stop it immediately, to fix the problem ASAP.

And yes, it's fixable. If there was a pathway for the user to achieve near-instantaneous distributed still-frame radiosity rendering, i.e., allowing the user to plug in additional computers to work on a combined solution. Why would this fix LW's marketplace woes?

Because no one else has it.

While character animation tools & the like are important, they exist elsewhere. "Me too" products and products that try to play "catch up" always, always lose - even when they are cheaper.

But nearly all renderers are weak in the area of radiosity, and there is no technical reason why LW couldn't bust out on the marketplace with an exponential leap in speed, gain some marketshare, and THEN be able to afford all of the others features LW needs - new character animation tools, etc.

Time for LW to stop being a "me too" "catch up" product and get back to where it belongs; it needs to start earning some decent money, get a bigger programming staff, and give big bonuses to everyone currently employed there.

And how do I have the time to type all this? You guessed it, I'm waiting on a render. Even with fPrime.

Weepul
11-20-2005, 05:43 AM
Once I made a scene and rendered it with radiosity. It took 1 min 30 sec to render. I was very satisfied. I showed this image to one guy and proudly told him my render time. He asked me to give him that scene. I gave him, and he imported it in Max and rendered with VRay. Render time was... 13 sec. He smiled and told me: "Welcome to 21st century!" I was quite hurt. Although I can play with render settings and maybe I can reduce render time twice, or even thrice, sacrificing quality a little bit and using tricks, it will never be 13 sec in current version of LW. Pity...

Ouch... :(

Captain Obvious
11-20-2005, 10:16 AM
Much faster GI needed, newer methods. Photon mapping, irradiance caching, final gathering, MLT - all are new GI techs out there, we're still at an old and unefficent montecarlo method. I wonder if anyone had ever used LW's montecarlo in production (people with hundred of machines does not count).
Not to be nitpicking, but Lightwave already has irradiance caching. That's what the interpolated mode is. It just... well... needs some improvement. ;)

Pavlov
11-20-2005, 11:22 AM
Not to be nitpicking, but Lightwave already has irradiance caching. That's what the interpolated mode is. It just... well... needs some improvement. ;)

yes, but it really doesnt work.
try Kray to understand what i mean: it cahces irradiances, and if first frame is 20min, all other are 2-3 min.
In Lw you get almost no difference between frames.

Paolo

Captain Obvious
11-20-2005, 12:07 PM
Well, that's radiosity caching, not irradiance caching. ;) Irradiance caching basically means that you use the same method as for a per-pixel sampler ("monte carlo" radiosity in LW), but instead of calculating using completely new samples for each ray, it stores the irradiance at each hit. If another ray hits right next to a place where you've already taken a sample, you reuse that. It saves a bunch of time, and it makes it possible to store the radiosity solution, but the real purpose of it isn't as such to store it between frames in an animation or something such.

</nitpick>

gaushell
11-20-2005, 12:09 PM
I'll throw my two cents (or 2dimes.com) in. I've been using LW since 1993 and have a staff of LW users churning out arch viz, medical, product, aviation, etc illustrations and animations daily.

LW needs to focus on the render engine big time. That helps everyone. Antialiasing is pretty awful for animation and I concur on many of the radiosity issues.

We have learned how to fake it pretty well and can setup and render quickly, but it would be even faster on the production end if the calculation methods were changed for true radiosity and it would look that much better.

fprime rocks! Love Worley. But until all of the SDK issues are resolved and it can easily render using our render farm, it has limitations.

I too would love to know of all of the professional users (full time making a living off of LW) who uses it for what, and what is most improtant to them. I'm not aware of this being asked in a scientific way - Chuck Baker - can you please do this? Seems like the best way to determine needs for Lightwave.

I for one, have volunteered many times over the years to give feedback and recommendations as a professional user only to be ignored. That is really frustrating. Like Ladd, I have a lot of money and time invested in LW and want it to be successful and am more than willing to help.

Let us Newtek!

I am excited about some of the proposed changes to v9 but as of yet, am unclear about many of the rendering quality related issues. We are keeping our fingers crossed here at Paradigm Productions.

And as far as finding quality LW users to hire - they are few and far between in the US. We are growing and need good people and there are few to select from. We are currently looking for 2 and there isn't much to choose from. That scares me more about the future of LW than anything - but I believe it is a user response due to base issues with the program (and marketing).

Having said that - LW is a good program, but it needs some focus. I would really love to see NewTek reach out to the professional users (and not just Hollyweird) to discuss issues.

laddehlingerjr
11-20-2005, 01:59 PM
I for one, have volunteered many times over the years to give feedback and recommendations as a professional user only to be ignored. That is really frustrating. Like Ladd, I have a lot of money and time invested in LW and want it to be successful and am more than willing to help.

In regards to the whole "ignoring" thing - I've run across the same thing. And I've heard this complaint not only from users but from dealers who wanted to sell LW. One dealer in particular, a good friend of mine, tried to set up a meeting before, during and after SIGGRAPH 2000 and just gave up after that. This dealer is now selling a lot of seats of competing products to the military market - STRICOM in particular. Which is frustrating because LW has a lot to offer in the way of real-time modeling for the simulation world. Especially now that it has 64 bit capabilities.

A scientific market survey would do a world of good. Perhaps it would help them break out of the hollyweird group-think. Softimage fell into the whole "hollywood loves us" syndrome, too, and spent enormous resources catering to them. It's pretty much at the bottom of the heap now despite drastic price cuts.

Stooch
11-20-2005, 05:22 PM
I totally agree. Right now I am at a point where there are superior tools that do things faster and more efficiently then LW, modeling, animation, even texturing. What shines for me is the render engine, and you know im not bound by a huge investment, the studio is just starting up, if the rendering doesnt improve, you can be assured that i will part with my long time partner in crime. There comes a point when love alone is not enough. This is a great opportunity for newtek, and it would be extremely closed minded to ignore the people who justify its existence.

Pavlov
11-20-2005, 05:59 PM
Well, that's radiosity caching, not irradiance caching. ;) Irradiance caching basically means that you use the same method as for a per-pixel sampler ("monte carlo" radiosity in LW), but instead of calculating using completely new samples for each ray, it stores the irradiance at each hit. If another ray hits right next to a place where you've already taken a sample, you reuse that. It saves a bunch of time, and it makes it possible to store the radiosity solution, but the real purpose of it isn't as such to store it between frames in an animation or something such.

</nitpick>

Hi, thanks for the infos. I intended irradiance caching as a pre-solution stored in a file etcetc, exactly what Kray and other engines do.
Sorry for misunderstanding ;)

Paolo

Captain Obvious
11-20-2005, 06:43 PM
Hi, thanks for the infos. I intended irradiance caching as a pre-solution stored in a file etcetc, exactly what Kray and other engines do.
Sorry for misunderstanding ;)
Oh, that's okay. I didn't know any of this just a few months ago! ;)

Ztreem
11-22-2005, 03:51 AM
I agree that the render engine needs some heavy updating in alot of areas, but I don't think Newtek should concentrate all their efforts on the render engine. It's alot in LW that need to be updated.

Pavlov
11-22-2005, 04:09 AM
I agree that the render engine needs some heavy updating in alot of areas, but I don't think Newtek should concentrate all their efforts on the render engine. It's alot in LW that need to be updated.

I partially agree, because in LW almost nothing *doesnt* need a deep update.
Render engine was a strongpoint of LW; it has a pretty good quality, but now it's objectively the weakest in the hi end area.
The best thing is they're getting SDK open (i pray they do this in a complete and good way) so we'll benefit of 3rd part engines.
But i think LW's base engine should be revamped in depth too. Give a first look at *tools*: there are just a few, spare option and controls for everything, while other software allow a deep control of every aspect. Control on Lighting, Camera and Rendering (MB, DoF, Netrender, texture filtering, displacement, etc) is very weak.. *too* weak for 2005.
It seems Camera got some attention in 9, something "NT way" has been shown regarding displacement so lets wait for these; nothing has been said about Lighting and Rendering. After many years of stasis on these, i think we're allowed to ask for some - credible - answers. Render engine is a crucial component of a 3D tool, not all users are modeler or char animators only.

Paolo Zambrini

Nitisara
11-22-2005, 04:27 AM
nothing has been said about Lighting and Rendering
It was announced that render will be faster x2 times.
We are waiting for this. I hope it will include radiosity calculations also. If it will be 2 times faster in practise, then it could compete with other renders in speed.
:newtek:

Pavlov
11-22-2005, 06:43 AM
It was announced that render will be faster x2 times.
We are waiting for this. I hope it will include radiosity calculations also. If it will be 2 times faster in practise, then it could compete with other renders in speed.
:newtek:

hi, i dont think the same way; 2x wont be enough for many tasks.
If you need high quality in GI calculations LW can crawl at times 10x higher than other engines. Apologize for the example, but it's indicative af a situation: recently Cinema, which had already a faster and better GI than LW, got a new advanced render module which gives Cinema a 2-4 x speed, so i guess it's now several times faster than LW.

Again, control issue: we need updated, deeper, more powerful control tools in lighting/rendering area, and nothing has been mentioned until now.
A simple "2x speed" is not an answer to a lot of long, overdue user's questions.

Paolo Zambrini

Matt
11-22-2005, 08:39 AM
Since I bought FPrime, I've stopped using the built-in LW renderer altogether.

Pavlov
11-22-2005, 09:07 AM
Since I bought FPrime, I've stopped using the built-in LW renderer altogether.

me too. Never touched F9 in months.
Now that i've Kray too, i didnt even remember of it ;)

Paolo

Panikos
11-22-2005, 09:59 AM
me three.

I even stop rendering using my 20node farm.
FPrime seems almost faster than the farm.