The world fastest unbiased renderer!

[Dimitri]

Hello everyone... there was a lot of time to participate in the forum's ongoing discussions... But visiting the forum and learning about the news never stopped...

Here we have a very interesting, I think, discussion... Pixel Oz's so detailed explanation of the issue is really very well put. Given that the gpu based rendering gives so fast results in unbiased rendering if it would be used with biased ways the rendering times would touch almost the lightspeed... So why not having a fully biased way of gpu rendering somewhere? And why not having a whole alternative section of Lux doing such a thing? It would be a superb novelty after all: the fully biased gpu based rendering choice...

Parthenon is some years old and yet... it has such rendering speed... it is really fascinating what would be if we had a renderer producing a quite high quality biased rendering based on gpu power...

I admittedly do know nothing about programming... I am a simple user, so I can not be sure how serious sounds such a thing, regarding the practical side as to its programming dimension, but the simple logic gets such thoughts to mind... and I just am sharing them...

[SATtva]

So why not having a fully biased way of gpu rendering somewhere? And why not having a whole alternative section of Lux doing such a thing?

The main problem is the lack of manpower. As Lux is primarily a physically based unbiased rendering engine, most development efforts will always be concentrated in that area. If we've had a lot of developers with plenty of spare time, than why not, Lux could become a full-featured rendering solution. Unfortunately (or maybe not ), this is not the case.

[Down Rodeo]

I feel like a broken record, but the Blender developers could probably "bolt on" LuxRays to Blender Internal if they really wanted to. Moving to a solution where everything is done on the GPU runs into... perhaps issues is the wrong word. There are technical challenges, like many cards not having enough memory to store large texture maps on them. Of course there are ways around it - and there are some GPUs with lots of additional memory now. But it's another thing to think about.

Not to derail the thread, but can you guys see LuxRays being used elsewhere? I know it's an early stage of development but if Dade manages what he's set out to do it'd be pretty cool to see it in Blender Internal for instance. I assume that the biggest hangup in BI's raytracer is tracing primitives, as it is in Lux. This is also assuming that it is relatively easy to add LuxRays into existing architectures.

That's enough rambling for now, anyway!

tl;dr: we live in interesting times!

[Dade]

Not to derail the thread, but can you guys see LuxRays being used elsewhere?

LuxRays has been developed as a stand alone component exactly with this intent.

I know it's an early stage of development but if Dade manages what he's set out to do it'd be pretty cool to see it in Blender Internal for instance. I assume that the biggest hangup in BI's raytracer is tracing primitives, as it is in Lux. This is also assuming that it is relatively easy to add LuxRays into existing architectures.

It is a lot easier than totally rewriting a rendering engine as required for a GPU-only approach. However it still require a good amount changes mainly because you trace many rays at the same time instead of just one.

I don't expect to see Blender Foundation consider the GPGPU support problem before 2.6. They are already in the middle of a big transition with the 2.49 => 2.5 step. However they will have to add GPGPU support after that.

[Dimitri]

I do understand the manpower issue... it is a serious issue indeed. I just wanted to share the idea. Having it in mind never harms... A fully biased renderer of a very high quality taking advantage of the gpu power... It could be a very original solution in the present condition things are yet (there is no such a thing around, as far as I know)... It will be nothing to be astonished if those days you hear somewhere appear such a news...

There is Fprime (it is a Lightwave plugin renderer), which is a near real time biased renderer... it is not gpu based, as it seems, but it is quite fast and of a very high quality. It is handling GI solutions very fast and easily. There are some videos showing what it does below... they are worth seeing:

http://www.worley.com/E/Products/fprime/fprime.html

Imagine Lux being a whole studio with both unbiased and fully biased rendering ways using gpu power and... having a studio of itself (staging and material applying)... It would be everyone's choice, no doubt... I would even give my hundred euros without the slightest hesitation for buying such an app.

I am not proposing making Lux commercial but... who would not give such an amount of money wholeheartedly for having such a tool in his/her disposition?... And earning some money from spending time on it it can not be considered a 'sin'... given that the amount would be affordable from nearly everyone.

Just ideas... sharing them is free, after all...

[SATtva]

Imagine Lux being a whole studio with both unbiased and fully biased rendering ways using gpu power and... having a studio of itself (staging and material applying)...

Given the current aim for more tight integration between LuxRender and Blender, Blender itself will become such a studio. If (when) Blender Foundation will consider implementation of LuxRays (or any other means to bring GPGPU support to Blender), you'll in effect have a studio for biased and unbiased rendering.

I am not proposing making Lux commercial but... who would not give such an amount of money wholeheartedly for having such a tool in his/her disposition?... And earning some money from spending time on it it can not be considered a 'sin'... given that the amount would be affordable from nearly everyone.

Making donations in a transparent way is tricky from the legal point of view, there is a non-trivial amount of bureaucracy involved to setup and conduct everything correctly.

[namekuseijin]

is that you, AD?

nice discussion, but it should be better at Blenderartists forum, as lack of GPU-processing in BI is Blender's a issue... not a... huh, off-topic Lux issue... :p

[PixelOz]

Well yes it is kinda off topic but not totally cause we are discussing graphics here, mainly centered around LuxRender of course cause it is the focus of attention here and you are right in that respect but it is very difficult to go into the subject of graphics as when talking about 3D and rendering as we are doing here and not talk about other things related to it.

Consider when you are having a normal conversation with friends when the discussion suddenly veers into anything else like politics, religion or even UFOs. Now that would definitely be very off-topic for sure but remember that we are discussing Blender too which is one of several programs that could be used with LuxRender and there is a lot of talk about the new integration of LuxRender with the new Blender 2.5 which could be very significant cause both open source programs might strengthen their mutual relationship in the near future.

A while ago for example Indigo started to be talked about a lot in Blender forums and vice versa cause the number of people using it with Blender was growing but it went commercial and that changed completely even though there are some Blender users that can afford it and are still using it of course but with LuxRender that can be very different cause it is also open source and I think that Lux is going to take a great part of the position that Indigo formerly took for many Blender users so I expect a lot more discussions about Blender in LuxRender forums and vice versa cause both program are becoming very powerful.

Somebody here mentioned something very important which is the fact that the Blender people are in the middle of a mayor transition toward version 2.5x and then toward 2.6x after that and I would love to see Blender's internal renderer support GPU acceleration but at the moment that could be very difficult for them and one has to understand that cause they are improving their program enormously and I think that the new version will probably gain a lot more users and as Blender users grow I think that Lux users will grow too due to the symbiotic relationship of both programs.

There are many things that could happen. Blender may eventually gain GPU acceleration for it's biased renderer and Lux can be the other renderer that people could use when they want much more photorealism or because some users prefer the particular way in which Lux renders their images. So I don't mind having both options or more options as it is the case with other renderers that one could use like Yafaray for example but Lux it's looking very attractive to me due to the GPU acceleration that they plan to incorporate in version 0.8 hopefully soon. I don't think that many new Blender 2.5x users will ignore that which in turn will create many new Lux users.

Another possibility that I found interesting that was mentioned here is that Lux could gain more developers in the future that could help develop a biased side of the renderer so it became a more complete rendering solution in one but at least it is developing very well as an unbiased renderer which is its main focus and that is OK anyway.

[Dimitri]

Here is a link to some videos showing a non-unbiased (as it seems) real time renderer made for Rhino (the nurbs modelling app)... The speed of its feedback is impressive., as its results' quality too...

http://www.youtube.com/watch?v=ugkiV2xh7Vw

Just mentioned it because of its relation to the present thread's discussion...