PDA

View Full Version : LW 10 VPR - fastest realtime render



yolao
08-14-2010, 09:58 AM
Hey

I`m not a LW user, but i have been checking all the new tools that LW will offer in the near future and the new real time render VPR looks AMAZING. It seems to be the fastest of its kind...is it?

In this (http://tv.newtek.com/player.php?recordID=70) video the speed and the quality of VPR are amazing.

Here are some questions about VPR.. Can someone please answer them.

-Can VPR render everything..?..i mean all kind of shaders...all kind of lights, etc..?.... most of today realtime renders..(the GPU based ones) can not render things like displacement, SSS, fluids..etc.

-Can i use the final solved VPR image output as a final render like fPrime..?..that would be fantastic, specially for render animation...

-In your opinion, is this VPR the fastest realtime render in the market?... checking some videos it seems that is FAR more faster that any realtime render or interactive render out there.. for example check this (http://www.youtube.com/watch?v=QwoSP6cHLVU) video

Thanks

UnCommonGrafx
08-14-2010, 11:46 AM
- Most everything. It isn't done yet so the capabilities are not completed.
- At the moment, capabilities of this type are not completed.
- Suffice it to say, it's fast enough. Opinions of this nature seem like troll-bait more than anything substantial. By that, I mean,
- Join the Core team and find out all this info firsthand. Then the opinion that will matter will be your own.

inquisitive
08-14-2010, 12:03 PM
If I had the money I would just upgrade to LW10 from LW9.6 and frankly I will as soon as possible ;)

yolao
08-14-2010, 12:42 PM
thanks for the info.I will be checking this as soon as a demo is available.:)

Svenart
08-14-2010, 02:08 PM
the movie from the second link was very nice. The vpr looks very fast.

Does anyone knows if this VPR modus is GPU based or not?

*Pete*
08-14-2010, 03:03 PM
the movie from the second link was very nice. The vpr looks very fast.



ah...THAT video?...Matt made it months ago.

Elmar Moelzer
08-14-2010, 03:20 PM
Does anyone knows if this VPR modus is GPU based or not?

It is currently CPU only.

Kuzey
08-14-2010, 03:21 PM
Speaking of which :D

Will vpr....be feature complete at the time of LW10 release?

You know, save rendered images/animations...render everything the normal render engine can etc. or is that slated for LW11?

Kuzey

Cageman
08-14-2010, 06:40 PM
Speaking of which :D

Will vpr....be feature complete at the time of LW10 release?

You know, save rendered images/animations...render everything the normal render engine can etc. or is that slated for LW11?

Kuzey

I'm quite sure that VPR will not become a renderer per see, since it is the same render as F9, so whats the point? It might be possible that you can do previews with VPR as you now do previews in OGL (you know, make preview, save preview and all that).

As for what it will support...well... there are some things talked about in HC that I can not spill here. :D

wesleycorgi
08-14-2010, 06:49 PM
Not having a tool like FPrime, VPR (warts and all) already has changed my workflow --- no more spot-check (F9) renders.

probiner
08-14-2010, 06:51 PM
I wonder how "make preview" will work with VPR. (did i touch a button cageman? :D)

Oh well comparing to Fprime, i think if VPR would do cache renders like Fprime does, it would really be a plus for animation people.

Matt
08-14-2010, 10:11 PM
I'm quite sure that VPR will not become a renderer per see, since it is the same render as F9, so whats the point?

http://tv.newtek.com/player_siggraph3.php

erikals
08-14-2010, 10:52 PM
yolao, just a note,
they used a Very fast computer, but yes, from what i know VPR is fast.
(but needs more work, as some features are missing)

Cageman
08-15-2010, 03:34 AM
http://tv.newtek.com/player_siggraph3.php

And your point is exactly what?

What in that video suggest that VPR will be able to replace F10 renders?

EDIT: I realize that I made a typo there... in the post quoted I had written F9, it should have been F10. I hope that will make more sense. :)

*lol*

Kuzey
08-15-2010, 04:51 AM
I'm quite sure that VPR will not become a renderer per see, since it is the same render as F9, so whats the point? It might be possible that you can do previews with VPR as you now do previews in OGL (you know, make preview, save preview and all that).

As for what it will support...well... there are some things talked about in HC that I can not spill here. :D


And your point is exactly what?

What in that video suggest that VPR will be able to replace F10 renders?

EDIT: I realize that I made a typo there... in the post quoted I had written F9, it should have been F10. I hope that will make more sense. :)

*lol*

The point is, even though it uses the same render engine...it's much faster and hence better as a previewer...I'm not talking about final renders or animations here. If you can export an image from vpr (which would be great and an obvious feature), then it would be wise to be able to export animations as well.

Now, I haven't used "make preview, save preview" before, but that sounds about right. I presume vpr would be even faster, in which case the current "make preview" should be updated to take advantage of vpr speed and quality.


Kuzey

Kuzey
08-15-2010, 05:04 AM
So vpr will not be finished at LW10 gold, which is understandable...since they moved focus during development at some point.

Along the same lines....will the modelling features of Core be compete at release. Will Core have all the current tools Modeler has, I'm not talking about Nurbs etc which will come later.

Basically, I want to know if you can use Core for all your modelling needs without having to launch Modeler...will that be possible? Again, not counting legacy plugins, Lscripts etc. and similar issues.

BTW.....are the current modelling tools in Core stable...or do they still need work?

Kuzey

Cageman
08-15-2010, 06:28 AM
Why not join HC and find out for yourself? ;)

hrgiger
08-15-2010, 09:08 AM
Along the same lines....will the modelling features of Core be compete at release. Will Core have all the current tools Modeler has, I'm not talking about Nurbs etc which will come later.

Basically, I want to know if you can use Core for all your modelling needs without having to launch Modeler...will that be possible? Again, not counting legacy plugins, Lscripts etc. and similar issues.

Kuzey

No. It was stated to us early on that CORE would not be a complete replacement for modeler or Layout for the first version. That much hasn't changed and is no secret. The plus side is that CORE tools are by their very nature more interactive then modelers and of course can be animated.

The features PDF more or less gives you an idea of what tools will be available for modeling. I think there's a few in CORE that aren't listed and I think there are a few listed that aren't in CORE yet but as you can see from the PDF, it will not replace the wide range of tools available for modeling. However, I think you will see a lot of third party add-ons for this first version.

As far as VPR goes, no not complete but then Fprime has gone through a few different versions adding more functionality as it went along so I gather the same will be true for VPR. But again, this hasn't changed from the beginning either as CORE only promised basic level animation and rendering and VPR started in CORE....

Matt
08-15-2010, 11:05 AM
And your point is exactly what?

Oh nothing major, just that video shows IPR being used interactively, just cool is all!

Kuzey
08-15-2010, 12:43 PM
Why not join HC and find out for yourself? ;)

Haha...wouldn't that be interesting :D


No. It was stated to us early on that CORE would not be a complete replacement for modeler or Layout for the first version. That much hasn't changed and is no secret. The plus side is that CORE tools are by their very nature more interactive then modelers and of course can be animated.

The features PDF more or less gives you an idea of what tools will be available for modeling. I think there's a few in CORE that aren't listed and I think there are a few listed that aren't in CORE yet but as you can see from the PDF, it will not replace the wide range of tools available for modeling. However, I think you will see a lot of third party add-ons for this first version.

As far as VPR goes, no not complete but then Fprime has gone through a few different versions adding more functionality as it went along so I gather the same will be true for VPR. But again, this hasn't changed from the beginning either as CORE only promised basic level animation and rendering and VPR started in CORE....

Sure, I was just trying to find out the level of Core completeness in a practical way. I understand, the focus went from modelling tools to vpr and perhaps even bullet physics, to mention two areas...so parts will be unfinished here and there...I don't have a problem with that. Thinking about it, it's probably the best way to go, develop multiple sections of the software at the same time.

It's just that things written on paper don't always reflect the true state of the features. For one thing, all the current tools could be buggy...the pdf won't mention that. I presume all current tools are either interactive by now...or very near it. While the animation part will come later...unless Newtek have been working on that area as well :D

PS...there really needs to be a LW10 gallery up and running, so people can post the work they have been doing in LW10/Core....just an idea :D

Kuzey

Kuzey
08-15-2010, 01:37 PM
Oh...I believe someone asked where would the export buttons go?


I would say put on the vpr top bar....for easy access.

This was copied from AF and edited in 5 minutes, but you get the idea. I'm sure Matt can come up with a better solution.


The idea is to get it away from the render options to where it belongs...in front of you :D

Kuzey

yolao
08-16-2010, 07:09 AM
yolao, just a note,
they used a Very fast computer, but yes, from what i know VPR is fast.
(but needs more work, as some features are missing)

Yes, i check that in the siggrapph videos they use a very powerful computer. But in the second video, where it compares to the other renders, is a fast but normal computer and is still very fast.


Speaking of which :D

Will vpr....be feature complete at the time of LW10 release?

You know, save rendered images/animations...render everything the normal render engine can etc. or is that slated for LW11?

Kuzey

I really REALLY hope that VPR will have all the options to save images/animations like fPrime. With that render speed it would be bad not to have those options!!.. Of course is fantastic for tweaking the lights, materials, etc. But again, IMO that render speed should also be used for save images and animations.

*Pete*
08-16-2010, 08:35 AM
I would say put on the vpr top bar....for easy access.

its there...together with wireframe, shaded, textured and other modes to view the...uummm...viewport :D

Amurrell
08-16-2010, 10:27 AM
yolao, just a note,
they used a Very fast computer, but yes, from what i know VPR is fast.
(but needs more work, as some features are missing)

Works fast on my i7 920 at stock speed, so I'm not sure what your definition of "very fast computer" is. I rarely have to wait very long for anything, depending on the settings and geometry.

wesleycorgi
08-16-2010, 10:43 AM
I have a three year old laptop with an Intel Core 2 Duo and VPR render pretty fast to me (my stuff is usually low to mid poly count with UVs, bumps and other kinds of textures). The highest poly I tried was the creature kit and it ran fine as well. For me, the weird thing about VPR is that you're not sure when it's done. In some other thread someone mentioned it's kind of like the old Polaroid pictures, where it goes from extremely unfocused to all of sudden you have a sharp image.

So if you are orbiting around, it may be out of focus. But when you stop, it quickly renders.

yolao
08-16-2010, 11:26 AM
You guys are aware that there will be no significant difference in render speed between VPR and the normal renderer? Just asking... because you guys mention yourself that you know it's based on the same "render engine". Nothing against having a "Save Image" for renders the VPR (Viewport PREVIEW Renderer) has ALREADY DONE in the viewport, but the final renderer will be, as has been said numerous times before, the SAME engine, so where does the idea come from that the VPR would/could be faster? Me confused...

I don`t know much about lightwave 10 and lightwave in general. So if the lightwave render will be as fast as VPR is fantastic!!.

But i was also hoping to use VPR to make preview render animations, so i can specify the time of each frame of the animation for fast preview....then do the final render animation in the LW render.

JAW
08-16-2010, 12:08 PM
this looks great. cant wait.

Kuzey
08-16-2010, 12:38 PM
its there...together with wireframe, shaded, textured and other modes to view the...uummm...viewport :D

You're not talking about the export image and create preview movie widgets that I whipped up(badly I must say), so you can export images/animations from the vpr viewport?

About the different viewport options, I already knew about those :D

Basically, when you enable vpr in a viewport, you get extra widgets to create movies and export images etc.

Kuzey

Kuzey
08-16-2010, 12:53 PM
You guys are aware that there will be no significant difference in render speed between VPR and the normal renderer? Just asking... because you guys mention yourself that you know it's based on the same "render engine". Nothing against having a "Save Image" for renders the VPR (Viewport PREVIEW Renderer) has ALREADY DONE in the viewport, but the final renderer will be, as has been said numerous times before, the SAME engine, so where does the idea come from that the VPR would/could be faster? Me confused...

I think it was mentioned in the original vpr video...the one with the oil tankers, and talked about in some other threads, that it will be close to the final render, but not the same quality....not %100 exact.

I presume they have optimized a lot of stuff and have taken a lot of shortcuts to get the speed in the vpr renders, which you don't want with the main render engine...because it'll loose quality and what not.


Kuzey

UltraViolet
08-16-2010, 04:08 PM
VPR sounds cool, If I knew that they were working on unbiased solution for renderer (besides current one) I would buy into CORE today (huge problem for me is that Thea renderer looks very, very tempting) :D

Kuzey
08-17-2010, 05:25 AM
Funny, you still confuse speed with interactivity... :)
Indeed you still don't quite get it. Remarkable. That's the tiring part of ... "presuming", I guess. Kudos for the persistance. Well, you'll have to wait until you get your hands on it, I guess. This time could be shortened by joining the beta group. :D

Ah :stumped:

I'm only going on what Newtek has said in the past, and as far as I can tell, it's still true to this very day. That a vpr render won't be the same as the final render....it's just a previewer, an interactive one at that....but just a previewer.

Otherwise, vpr would be feature complete already (since it's the same render engine) and able to render everything that the full render engine can and at the same quality...but it can't, that's why everyone has called it a previewer to this day....it's fast, but not final render quality.

Oliver, looks like you have news that I and the public don't have access to....would you care to share the links where Newtek has said that there won't be any difference between a vpr render and normal render?


Kuzey

Kuzey
08-17-2010, 05:30 AM
ps...how about a little game Oliver???

Can you post a picture of a 15 second vpr render and a 10 hour normal render of the same view...so we can try and to spot the difference?

:)

Kuzey

Lightwolf
08-17-2010, 06:46 AM
ps...how about a little game Oliver???

Can you post a picture of a 15 second vpr render and a 10 hour normal render of the same view...so we can try and to spot the difference?

If you have a 10 hour render then you can surely spot the difference to the same render in 15s on VPR... d'oh.

From what Mark told me it is pretty much the same renderer, except that not all features of LW are supported yet (this is the Siggraph version). Once they are it won't be much faster (obviously) - but it's still progressive.

Cheers,
Mike

Kuzey
08-17-2010, 07:10 AM
If you have a 10 hour render then you can surely spot the difference to the same render in 15s on VPR... d'oh.

From what Mark told me it is pretty much the same renderer, except that not all features of LW are supported yet (this is the Siggraph version). Once they are it won't be much faster (obviously) - but it's still progressive.

Cheers,
Mike

Haha...well, it was a joke. I was hoping someone would play along with a response like: What 10 hour render? :D

That's my understanding from the beginning as well, but that they were using tricks under the hood to get the speed they now have. That and it won't match the final render...maybe the colour would be a bit off etc....but that was months and months ago....but I guess that might depend on how long to leave vpr to run?

If you leave a vpr render on for 10 hours (since it's progressive), will it be exactly the same image as the normal 10 hour render?

It would be great if it is, but then again, I can't see people doing that...they would use the normal render instead :D

Kuzey

Lightwolf
08-17-2010, 07:25 AM
If you leave a vpr render on for 10 hours (since it's progressive), will it be exactly the same image as the normal 10 hour render?
It's progressive but it stops, it doesn't refine endlessly.
And it will be different due to the different AA algorithms that must be used.

Cheers,
Mike

Kuzey
08-17-2010, 07:32 AM
It's progressive but it stops, it doesn't refine endlessly.
And it will be different due to the different AA algorithms that must be used.

Cheers,
Mike

Now we are talking Mike!!

Thank you, for the actual and factual information...a breath of fresh air, as always :thumbsup:

ps....Mike, do you know at what point it will stop refining...20...30..60 minutes ?

Kuzey

Lightwolf
08-17-2010, 07:50 AM
ps....Mike, do you know at what point it will stop refining...20...30..60 minutes ?
I'm guessing here, but it looked like a set AA level.

Cheers,
Mike

moussepipi2000
08-17-2010, 07:59 AM
i dont think it stop in term of time but in term of quality. it will stop for example at quality 10. it can take 2 sec or 10 hour depend of the scene

Kuzey
08-17-2010, 08:13 AM
I'm guessing here, but it looked like a set AA level.

Cheers,
Mike


i dont think it stop in term of time but in term of quality. it will stop for example at quality 10. it can take 2 sec or 10 hour depend of the scene

Ah...that makes it clear...thanks guys :thumbsup:

I'm guessing there would be some kind of visual feedback, at the point it stops refining, letting you know it's complete...as it were. Could be useful, if vpr ever exports images/animations in the near future :hey:

Kuzey

praa
08-17-2010, 01:47 PM
come on stop being soo 2009 and add OpenCL functionality already :)

Lightwolf
08-17-2010, 01:52 PM
come on stop being soo 2009 and add OpenCL functionality already :)
But OpenCL is the wet dream for rendering of 2009... and the hangover for 2011. Might as well skip it for the time being ;)

Cheers,
Mike

Salv8or
08-17-2010, 02:26 PM
I know I will probebly not get a response on this question other than join the HardCore program, but Ill fire it away anyhow.
Does LW 10's "screamernet" work in the same way as it always have, or does it support bucket rendering over network?
And the obvious follow up: How about VPR?
Will it render on multiple cpu over a network on the same frame? To me that would be a wet dream come true.
Or at least support for VPM(VirtuellParallelMachine) on Linux?

Lightwolf
08-17-2010, 02:28 PM
I know I will probebly not get a response on this question other than join the HardCore program, but Ill fire it away anyhow.
Does LW 10's "screamernet" work in the same way as it always have, or does it support bucket rendering over network?
No bucket rendering was seen at Siggraph...

Cheers,
Mike

Kuzey
08-17-2010, 02:38 PM
Hey Mike, while you're here and what not :hey:

Was Core demoed at Siggraph or was is just LW10?

Kuzey

Lightwolf
08-17-2010, 02:46 PM
Hey Mike, while you're here and what not :hey:

Was Core demoed at Siggraph or was is just LW10?

Kuzey
I haven't spend that much time at the trade show to be honest, but I mainly saw the Layout part of LW10 (which, as we all know by now, includes Layout, Modeler and Core :D ).

Cheers,
Mike

Salv8or
08-17-2010, 02:50 PM
No bucket rendering was seen at Siggraph...

Cheers,
Mike

Well, Im keeping my fingers crossed. They have it in Modo so I still hope. =)

Kuzey
08-17-2010, 02:53 PM
I haven't spend that much time at the trade show to be honest, but I mainly saw the Layout part of LW10 (which, as we all know by now, includes Layout, Modeler and Core :D ).

Cheers,
Mike

Haha...nice touch that, make me laugh :D

Kuzey

Hieron
08-17-2010, 04:58 PM
But OpenCL is the wet dream for rendering of 2009... and the hangover for 2011. Might as well skip it for the time being ;)

Cheers,
Mike


Aye and I think B. Peebler agrees with you as well :)
But it does sound sexy!

(I hope they spend the massive amount of time needed for such a thing on the bugs, issues and general features first)

As for the VPR being fast or slow, I first of all hope it works *flawlessly* and includes volumetric and 3rd party effects. HDI! And :) at Oliver :) (be zen, let it slide...)

UltraViolet
08-17-2010, 05:25 PM
So, is it safe to conclude that unbiased rendering is not part of the road map for Core (does anyone even talk about or want this?) ? :confused:

Lightwolf
08-18-2010, 01:42 AM
Aye and I think B. Peebler agrees with you as well :)

It took him a few years to come out with it though :D

Cheers,
Mike

Kuzey
08-18-2010, 06:25 AM
No, it's been called the preview renderer, because it renders the image progressively, starting at a low resolution, not with less features. Sure, for now there's no area lights... so what? It's beta. You seriously think they'll leave them out for the final product? Hint: they won't.


It does not render everything that the full render enging can simply for the reason that it is beta, though I suspect VPR in Layout 10 will have the same or similar restrictions as FPrime. This however is not the plan for core... which VPR was originally designed for.


My goodness...do you really think I consider the current state of vpr as final...come on, I said no such thing. I even asked about how complete vpr will be when LW10 is released. I know it's beta and I know it's not finished. Please wake up and stop putting words into my mouth, please don't presume you can read me like a book...you can't :eek:

Oliver, your whole problem with me lies there.



But well, since you have such a hard time understanding the purpose of and the basic concept behind progressive rendering (as shown in this and some other thread), that seems like a rather tedious task... *sigh*
The thing is, the VPR is based on the same render engine. We agree on that? The purpose of using the renderer in realtime in your viewport is for finetuning the final result. Moving a light in the generally correct position can be done in OpenGL... but for "result-tuning" having "the colour a little bit off" is exactly what you DON'T want. Having Hypervoxels showing up "looking a little bit different" is what you DON'T want... can we agree on that? Having a different resolution while tuning and hence not the same amount of Antialiasing is acceptable - that's why the VPR does not refine forever as FPrime (unless a quality level is set) does. That's why it's called a preview renderer. You set your viewport to render with the viewport preview rendere, finetune your settings with it, maybe save out a frame (would like to see that functionality) to have it signed of by a client (remember my comment about "since it's already done"?) and then you can adjust your final resolution, AA settings, output directory, animation related stuff and let the final renderer do its thing.

So this is the reason why in every comment regarding the (unfinished) VPR and (upcoming) final renderer it was pointed out that they are based on the same engine, hence producing the same results. As for links... you might in turn want to share some of those that apparently hint at the other direction? ;) I have never seen those remarks.


Somehow, just because I choose to use the word fast, it just eats you up and you can't handle it...at all. I would look into that if I were you. But I'm not making things up as you think I am, I believe it was mentioned in the first vpr video...as I said before, or in the thread about it, that the vpr render won't exactly be the same as the final render, and I'm pretty sure Jay and Chuck said the same at various times as well....I think colour was mentioned as well...but not sure. Of course, they didn't go into details as to what might be different.

The fact is, that it being the same render engine does not necessarily mean you'll get same exact result. I'm sure Newtek are using other tricks to "speed" things up in a vpr render that they want to keep secret, I find it hard to believe all this "speed" improvement is just the result of using "different AA algorithms" alone and those tricks might affect colour in some areas.



Let me ask you a question: if the VPR will show images that differ from the final render... what's its purpose again? We have that in OpenGL... a rough representation of the scene. No really, why would I want a renderer, that calls itself "preview" renderer, if it shows something different from what is supposed to preview? It's supposed to "preview" the final result, so that you can tweak it. But for you it's just a slow OpenGL viewport? Please enlighten me... :D


Easy, to give you a rough idea of how the final render would look..without doing 10 hour test renders. Again, I'm going on what was mention by Newtek.

Oliver, since I haven't used fprime before, did that a) use the same render engine in LW and b) was the fprime result the same as the LW render??



Look at it this way: the VPR is not fast per se. It's progressive and gives you something to work with instantly. It's direct. It's based on a fast render engine, and the final renderer will be just as fast, or, since the software will probably do nothing but rendering once you fire up the final renderer, it will probably even be faster. There will be other features and benefits for not rendering progressively... cast a glance at modo if you please. But the final renderer can not show you changes instantly while dragging a minislider... and that's where VPR shines.


Fast....instant...bloody fast, it's all the same thing...in that, you don't wait for a result, it's there in pretty much no time.




On a sidenote: You heard about the core beta group? You are aware that there are tons of people that "have news that you and the public don't have access to"... for a reason? Because we payed to participate and were asked to let NT decide which information to share? Just asking... :hey:

PS: You might want to check out http://tv.newtek.com/library.php ... if you sort by product you will find some more siggraph videos than linked from the front page. There are a few core clips there. :)

What is "tons of people" anyway...more than 500....less than 1,500 :)

I want vpr to be better and so far it's a hundred time better that what we saw in the tanker demo, but..I also remember that most of you guys were happy with it was a fprime clone and with all that pixelation as the scene was rotated...too.

So, I am both optimistic and concerned at the same time....optimistic that Core will get there. Concerned, that HC members can't see the future possibilities of the software they are beta testing. Like the idea of creating preview animations in vpr...not just saving images. That has a lot of potential that it seems many people don't recognize :D

Kuzey

Lightwolf
08-18-2010, 06:36 AM
Oliver, since I haven't used fprime before, did that a) use the same render engine in LW and b) was the fprime result the same as the LW render??

Define "render engine" :D

I.e. is surfacing and shading a part of the render engine or not?

Or is the render engine "just" a raytracer that doesn't actually shade anything?

Cheers,
Mike

Kuzey
08-18-2010, 07:01 AM
Define "render engine" :D

I.e. is surfacing and shading a part of the render engine or not?

Or is the render engine "just" a raytracer that doesn't actually shade anything?

Cheers,
Mike

Oooooh....multiple choice questions, I would have to guess.....a raytracer...or not..hehe.

The thing I remember, there were differences between a LW render and a fprime render??

Mike, I'm up for a lesson if you want...I always love this technical stuff :D

Kuzey

Lightwolf
08-18-2010, 07:12 AM
Oooooh....multiple choice questions, I would have to guess.....a raytracer...or not..hehe.

The thing I remember, there were differences between a LW render and a fprime render??

Mike, I'm up for a lesson if you want...I always love this technical stuff :D

Well, that's the thing... a render engine is the sum of its parts. And, just like Fprime uses mechanisms as exposed by LW, so does VPR.
Unlike FPrime though, VPR doesn't have to reverse engineer as much functionality, as it gets to use the native code.
One example would be classic surfaces.
FPrime can extract all the channel values (and also evaluate all the layers, including the plugins) - and the results are identical to what LW computes.
However, FPrime uses those to handle its own shading, which can look different as it can't use the same code here. VPR can (and probably does).

The same goes for (pre-plugin) lights, FPrime can extract the settings but has to compute them on its own (which in turn includes shadowing and that touches raytracing again).

But, in either case you'll see a third party renderer jump in and out of LW continuously to let LW compute parts of the final render.

VPR, essentially, has the chance to use more of the native code that isn't exposed to third parties. But, that also doesn't make it any faster either.

Cheers,
Mike

P.S. Going back to my AA comment: the thing is, progressive rendering requires different AA algorithms because it is progressive. It's got nothing to do with speed.

Kuzey
08-18-2010, 07:33 AM
Well, that's the thing... a render engine is the sum of its parts. And, just like Fprime uses mechanisms as exposed by LW, so does VPR.
Unlike FPrime though, VPR doesn't have to reverse engineer as much functionality, as it gets to use the native code.
One example would be classic surfaces.
FPrime can extract all the channel values (and also evaluate all the layers, including the plugins) - and the results are identical to what LW computes.
However, FPrime uses those to handle its own shading, which can look different as it can't use the same code here. VPR can (and probably does).

The same goes for (pre-plugin) lights, FPrime can extract the settings but has to compute them on its own (which in turn includes shadowing and that touches raytracing again).

But, in either case you'll see a third party renderer jump in and out of LW continuously to let LW compute parts of the final render.

VPR, essentially, has the chance to use more of the native code that isn't exposed to third parties. But, that also doesn't make it any faster either.

Cheers,
Mike

P.S. Going back to my AA comment: the thing is, progressive rendering requires different AA algorithms because it is progressive. It's got nothing to do with speed.

Nice stuff :thumbsup:

I would have thought using native code would have had some effect on the speed. But wouldn't Newtek optimize parts of the code (not AA algorithms) that vpr uses....to increase speed?

Kuzey

Lightwolf
08-18-2010, 07:34 AM
I would have thought using native code would have had some effect on the speed. But wouldn't Newtek optimize parts of the code (not AA algorithms) that vpr uses....to increase speed?
Why shouldn't they optimize them either way to gain speed, if it's the same code?

Cheers,
Mike

Lightwolf
08-18-2010, 07:35 AM
Basically, the main difference between VPR (in a more mature state) and the "normal" LW renderer will be which pixels it renders first... and how it display them.

Cheers,
Mike

Kuzey
08-18-2010, 07:39 AM
Why shouldn't they optimize them either way to gain speed, if it's the same code?

Cheers,
Mike

Ah...I'm still thinking in terms of the results being different, say...if the speed increases then something in the quality might decrease....a trade off for that speed etc.

Kuzey

Sensei
08-18-2010, 08:09 AM
In this (http://tv.newtek.com/player.php?recordID=70) video the speed and the quality of VPR are amazing.


By looking at this video, I am guessing they used trick - while changing surface attribute, or light position, etc. not geometry related thing, it's reusing previously rendered data buffers, so no ray-tracing or other heavy task need to be calculated. This would explain lack of progression while changing Surface Editor settings, and showing it while moving/scaling character or rotating camera.

Second thing which they could use is using OpenGL buffers like geometry position, depth from camera, to skip ray-tracing from camera.

This will be easy to check- if I am right, it will work only with perspective/orthogonal cameras, and not work with f.e. 3rd party custom camera plug-ins (or work very slow in comparison with perspective one).

If they used OpenGL also to simulate shadows (to skip ray-tracing), that will be easy to check, because transparent geometry (especially multi-layer transparency) won't cast correct colored shadows, but always opaque.

Lightwolf
08-18-2010, 08:20 AM
By looking at this video, I am guessing they used trick - while changing surface attribute, or light position, etc. not geometry related thing, it's reusing previously rendered data buffers, so no ray-tracing or other heavy task need to be calculated. This would explain lack of progression while changing Surface Editor settings, and showing it while moving/scaling character or rotating camera.
Either that or it simply doesn't need to update the geometry and thus has no need to update any acceleration structures.

Cheers,
Mike

Sensei
08-18-2010, 09:00 AM
or it simply doesn't need to update the geometry and thus has no need to update any acceleration structures.

This is done by all real-time previewers. It's must have..

I was thinking about using depth buffer, normal vectors, maybe even buffered light samples, for every single rendered pixel in previous frame. This way poor substitution of full render is just width*height dot-product operations between (buffered) light sample(s) and (buffered) normal vector(s) multiplied by raw color.. Ready to show to user in fraction of second, without completely worrying about complexity of scene.

Lightwolf
08-18-2010, 09:02 AM
This is done by all real-time previewers. It's must have..

I was thinking about using depth buffer, normal vectors, maybe even buffered light samples, for every single rendered pixel in previous frame. This way poor substitution of full render is just width*height dot-product operations between (buffered) light sample(s) and (buffered) normal vector(s) multiplied by raw color.. Ready to show to user in fraction of second, without completely worrying about complexity of scene.
This is called VIPER (to a degree) and available in LW already ;)

Cheers,
Mike

Sensei
08-18-2010, 09:17 AM
This is called VIPER (to a degree) and available in LW already

Yes, I know.. ;)

That's why I was wondering in previous post, whether they simply get data (XYZ and normals) from OpenGL viewports reading its internal framebuffers (after all these data are there in viewport, no matter what)..

Much quicker than casting 1920 x 1080 rays to scene.. And not so much scene complexity dependent..

If custom cameras won't work as fast, or at all, trick will be dis closured.

Video showed just point or distant light. These are easy to cheat with regular OpenGL rendering to not visible framebuffer. In distant, just use orthogonal camera. Whatever is visible from camera (=light) is getting intensity color from light, otherwise shadow color.
And ray-tracing bypassed..

That would explain why area light is not supported.. I recall somebody mentioning it?

Lightwolf
08-18-2010, 09:20 AM
That would explain why area light is not supported.. I recall somebody mentioning it?
Yes, but that's basically because the version we see now was just finished before Siggraph. It's probably closer to an initial proof of concept than anything else.

Also, using any OpenGL functionality for the rendering would not explain the initially rough progressive image when geometry is changed.

Cheers,
Mike

Dexter2999
08-19-2010, 11:22 AM
"fast" is relative.

One tanker ship or character in a scene and sure it is fast.
But as I have seen in other posts from people with very few posts and recent member dates on the forums (I shall call them "newbies" based on this information). These newbies have scenes with "I have a model with 1million polygons and I have it cloned 40 times and my machine takes too long" or "I need 1 million particles for my hypervoxel emitters".

In scenes of this magnitude I'm not sure that the VPR will qualify as "fast".

Elmar Moelzer
08-19-2010, 11:41 AM
In scenes of this magnitude I'm not sure that the VPR will qualify as "fast".
Nothing will though.

Kuzey
08-19-2010, 03:31 PM
Sorry, you still misunderstand the whole concept... *facepalm*
It's not a hundred times better. It just looks like that to you, as you are outside the beta group and mix up facts. It still behaves like that. Believe it or not. No... better: believe it, please. And we are still happy with a FPrime clone. It still is nothing more than that (except the advantages we tried to explain to you in the older thread). You think you see something more, but it's not there. It's not "a hundred time better that what we saw in the tanker demo". Sure, the render engine is worked on and it will become faster, that's the natural way how things tend to develop. But apart from that the concept behind it all has not changed one bit!


I can only go by what Newtek decides to show and I will most likely get it wrong as a result.

Why not mention that in your first post, that nothing has changed in terms of the very large pixels...it could have save a lot of effect and time? In the meantime, here I am, thinking some progress has been made in terms of user interactivity...eg, no or very little pixelation....as you rotate/manipulate your scene.



Lesson learned: the tanker demo looks today more or less the same still, with the (not) "new" VPR. It was demoed on a slower system and the tankers are a far heavier scene than the material that was shown at siggraph. I remember you saying something like "that looks much better" about the siggraph stuff back then, and I didn't comment on it. Cause yes... the material at siggraph was refining much faster, but the tanker scene still pixels up a lot... and that's not a big deal, it's still the genious concept behind the whole technique, which still unfortunately is so utterly alien to you. :)


NT cared about presenting cool stuff (Yay!), so they preferred not to show scenes that tiptoe on the brink of what the presenter's systems were able to handle.

There will always be scenes where the final renderer will take hours to render, just as there will always be scenes that are very heavy for the VPR... and it will show huge pixels initially or when navigating, and it will take ages to refine. Got it? Simple, eh? :)


No...I'm not concerned with the concept behind vpr, I like it and it's great. I'm more concerned about trying to tame/control it, so it's only working when you want it and not during navigating etc. It's a little thing in the whole scheme of things, and I'm not sure if it's possible or not...but that's still my concern.



What's my basic problem? What's eating me? It's not about semantics. :hey:
No,... it's this: we tell you this and that, and NT tells you that and this, and all you can do is guess and presume, and please, it's your good right to go on trying to ask about stuff that you have no idea about (and in some cases have not payed to know about this early - see core beta). But why do you think it is so important to tell what we misunderstand about the VPR - if you didn't get it in the first place? What drives you to present your thoughts (incl improvements) on stuff that you - as we tell you - have not understood at a much simpler level. Why do you think we misunderstand you? We have the insight, you have not. You think all of us in this and the other threads lie to you for fun?

I am just caught in wonder: for me that would be far to tedious (as it is becoming for some of us trying to explain to you where you are off track in your guesswork). You must be very, very attracted to core (maybe YOU should look into THAT... probably with your wallet and a little money transfer to NT) to waste so much time indeed knowing all you can do is guess. And guess ridiculously wrong in many cases. I think all I'm advising you to do is: stop trying to fight a futile fight - it's horrible to watch. Please rest assured, you are not playing core-beta-light out here, even if it sometimes might feel like that for you. :D

Conclusion: you are not "going on what was mention by Newtek" - you're simply flying hopelessly blindly, that's all. :)



Mmmm....I must have missed that memo from Newtek...or was there more than one :D

The fact is, all this comes down to the lack of information on LW10/Core and the decisions around it....for almost two years now and no one in the public is still sure of what is going on.

HC program will be closed after LW10 for new members...why, we still don't know. Subscription model looks like it's dead, yet it will kinda continue, if you buy before LW10 release...well almost, since you don't get the final product(LW11) at the end of next year, but you get to play with it. Buy after LW10, and there's no HC and no form of subscription plan at all...why? Why try to force people to upgrade now....is Newtek in some kind of trouble...it sounds like it...but as you know, I'm not in the privilege group to know any better :hey:


Orange Coke - designed by a mediocre VPR! :D Sounds... not so impressive. Well, there was a lemon flavored Coke, so why not an Orange coloured & flavored Coke....Mmmm...maybe not :D

Kuzey