Biased Or Non-Biased renderers

[maleemk]

Hello to everybody here,
I've been looking around for a while and found that this forum (PP2) is the best forum to share and discuss Architecture related ideas and thoughts effictively,

I would like to begin a discussion about biased and Non-Biased renderers.
I have used Accurender ( a non biased renderer ) extensively for Lighting analysis, Material analysis ( for example whether to use glazed tiles or matte tiles on certain lighting conditions ) and to decide color schemes to say the least.
Now I am using Mexwell ( a non biased renderer ) and V-ray ( a biased renderer )
In my opinion non-biased renderers are very use for Design purpose and biased renderers are good for presentation purposes what’s your opinion and I would like to categorize all the currently available ( and under development ) ones under the Biased and Non-Biased Renderers because I would be help full for us Architects; what to use as a design tool ? and what to use as Presentation tools ?

[Nick Fox]

It would seem to depend on how much time you have to render and the nature of the project your doing. Unbiased rendering engines require longer rendering times with often very realistic results (depends on the skill of the end user) whereas unbiased renderers are generally a lot faster but are also capable of equally realistic results and again that depends a lot on the skill of the user. I don't think it's right to say non biased is good for design purposes and nothing else or that biased renderers are better for presentation. I really don't think architects care if their renderer is biased or unbiased and I think you'll find they use the renderer they can afford to buy that will give the results they like and have seen and whether or not it's V-Ray, mental ray, Maxwelll or whatever isn't an issue. What seems to be important to most users is ease of use in operation.

[A_Minima]

Time, time, time...
The time is the nerve of renderers war...

[kwistenbiebel]

Time, time, time...
The time is the nerve of renderers war...

...And ease of use.
Some developpers just like to make it as complicated as possible.:bang head

I have abandoned the piste of unbiased renders because of the time issue and the according noise. (I would consider Fryrender though because it seems to have a very good quality/speed ratio due to improved algorithms. A sketchup bridge isn't implemented yet unfortunately)

I like the new biased solutions that are trying to gain a piece of the market.
However the best quality engines are most often the ones with too many whistles and bells...killing workflow and the joy to work.

[Melvyn]

A sculpter can have the best carving tool, the final masterpiece solely depends on the artist himself.

kwistenbiebel have some cool renderings by using podium and vray. Definitely worth looking at.

Cheers.

[maleemk]

hi biebel,
Fan of your's buddy, also hi to Minima
biebel you seems to better getting my point,:rock on:
First of all the time factor will be solved sooner or later ( as the war of cores and nm is heating up between Intel and AMD ) for sure and the recent development in the hardware and software front,
I used to be a user of Accurender and it was way ahead of it's time ( consider the P2's and P1's era )
My actual concern is for sure biebel rendres are amazing to look at but do they play any "Decisive" role in his design process? after all their should be a difference between Architect and Graphic Designer or Visualizer
As for Melvyn consider these renders as initial sketches made by a sculptor you simply can't compare them with the final product ( The sculpture ) itself.
As it is true for the Architecture, In my opinion working drawings are more important for "us" than the rendered images in this regard.( or can these renderers also take the role of working drawings ? ).
My point is we should analyze these Rendering Engines as Architects not as
Graphic designers or Visulizers
As for biebel again i love your process of integrating the real projects with renderers that's i am talking about, you simply can't match these renderes to the final product itself.
The whole idea behind this discussion is that:
" In the presence of so many rendering engines and so many to come which one, two or three of them fulfills the need of Architectural Visualization to what extent "

[sigue2000]

Here's a list of Renderers on the site of the Sunflow open source rendering system.
They are not sorted by biased and non-biased, but the list is impressive.

Link (http://sunflow.sourceforge.net/index.php?pg=link)

[sigue2000]

To this day I am no render crack like Kwist, but at the pace that these render engine releases are going, who knows.

[Nick Fox]

To this day I am no render crack like Kwist, but at the pace that these render engine releases are going, who knows.
I think it's a bit like comparing one apple with another. Most of us here are architects or wanna be architects so this argument is a bit of a wank.

[am_i_wry]

You know i find it kind of absurd that we cant render in (near) photoreal quality in more or less real time. I have perhaps unrealistic expectations of technology but it seems that the whole rendering scene is about extracting ever more perfectly realistic results out of ever more obtuse software. I want something simple that you click a button and get good useable images out of quickly. Surely that isnt beyond the wit of man, and with the way software development seems to proceed it should be available by now!

[SWANK-E]

Can someone please explain what's Biased and Non-Biased rendering? :confused:

[Ross Millaney]

Non-biased would be for example Maxwell-you input realistic real world values to achieve your results. For example the shutter speed of the camera etc....

Biased renderers is basically (IMO) a workaround aproach!Its approach is more like programming perhaps?!I'd call artlantis biased....but hey maybe Im wrong!?! :wondering

[Nick Fox]

Can someone please explain what's Biased and Non-Biased rendering? :confused:

Non biased renderers are concerned with the physics of natural light and it's behaviour and are more intensive in CPU use (so render slower) than biased renderers that are generally faster and less CPU intensive so faster to render - though this depends on the settings you're using. For example I can turn a 10 second rendering in mental ral into one that takes a couple of hours just by changing it's settings. I think you'd be hard pushed to tell the difference between the two where their quality settings were the same. Here's a link that explains how Maxwell works (sort of):

http://cgarchitect.vismasters.com/catalog/viewproduct.aspx?product=6

[sigue2000]

I think it's a bit like comparing one apple with another. Most of us here are architects or wanna be architects so this argument is a bit of a wank.
For one Nick, Kwist is an architect I respect. Second, I am talking usability and input efficiency here.

[pitrak]

Non-biased renderers like maxwell simulate the physical behaviour of photons. For output they even simulate cameras and lenses. So lens effects, depth of field, ... are all the effect of modelling a real camera and reproducing the light rays going through rather than the result of tricks.

In a biased renderer, depth of field would be achieved by progressive blurring related to the Z-axis (away from or towards the camera). Lens effects are post-processed to make the render look more like a photo.
The behaviour of light is simulated in such a way that it yields convincing images in as little time as possible.

Theoretically, biased renderers are most interesting, as they allow you to simulate how a textured building will look in real life. In practice, you need to model and texture everything very accurately to really be able to use it as a light and materials study. That takes a lot of time and practice. If you are after images to seduce rather than to study, I think biased is the way to go.

[Nick Fox]

For one Nick, Kwist is an architect I respect. Second, I am talking usability and input efficiency here.
I don't know what you're talking about when you refer to 'input efficiency' so maybe you could exlain that to me? Makes no difference to me if Kwist is an architect or a pelican. Most architects only care about how to get the best rendering, in the quickest time, for the least dollars and I doubt they are going to evaluate a renderer on whether or not it's biased or non biased.

[Nick Fox]

Non-biased renderers like maxwell simulate the physical behaviour of photons. For output they even simulate cameras and lenses. So lens effects, depth of field, ... are all the effect of modelling a real camera and reproducing the light rays going through rather than the result of tricks.

In a biased renderer, depth of field would be achieved by progressive blurring related to the Z-axis (away from or towards the camera). Lens effects are post-processed to make the render look more like a photo.
The behaviour of light is simulated in such a way that it yields convincing images in as little time as possible.

Theoretically, biased renderers are most interesting, as they allow you to simulate how a textured building will look in real life. In practice, you need to model and texture everything very accurately to really be able to use it as a light and materials study. That takes a lot of time and practice. If you are after images to seduce rather than to study, I think biased is the way to go.

I'm inclined to agree with you and in the end it indeed comes down to the skill of the end user. Getting good renderings also requires a lot of research and looking at how textures behave under different lighting conditions in the real world. This is something a lot of people assume is what you get when you stick a couple of lights into your scene when in fact it's a whole lot more.

[pitrak]

I'm inclined to agree with you and in the end it indeed comes down to the skill of the end user. Getting good renderings also requires a lot of research and looking at how textures behave under different lighting conditions in the real world. This is something a lot of people assume is what you get when you stick a couple of lights into your scene when in fact it's a whole lot more.

Sure, it is very complex. I guess that's why most people use biased renderers...

If you look at how to define materials in maxwell, you see the same complexity as in the real world. You have your texture or colour map, the parameters that define the behaviour of light on the surface (very simplified: everything between rough and glossy, and that's a lot) but you can also add several films. Those act likes layers of varnish on a painting - they don't necessarily change the colours, not even the view from a certain angle, but they can make a huge difference when looking from another angle (both for the reflection and for the depth of the colour).

So making one realistic material in maxwell can take you more time than making a bunch of renders in a simpler program. The experts have made a whole lot of materials already, so that makes it all much easier. :cheers:

There's more of course, I hope the experts can add to this. I only know most of the principles - i've been using render engines since some years, but didn't have the time and continuity to get good at one specific renderer...
Feels like I have to start over and over again :rolleyes:

Maybe I should focus on the simpler renderers after all :D

[Bricklyne]

Can someone please explain what's Biased and Non-Biased rendering? :confused:

To put it in layman's terms: unbiased renderers are "brute force" renderers that try to simulate the exact physical behaviour of light and light interaction with geometry to produce "physically correct" renders, which tend to be more photorealistic than those produced with biased renders.

As the "brute force" definition implies, these renderers tend to be very CPU intensive and processesor power- hungry and usually take long to clear up (clean up the "noise" in resolving the render), but the light quality is typically very very difficult to match with biased renders unless you're really experienced and have a keen eye for details and realism.

Unbiased renders usually have very few settings and dials to tweak, since everything works like a real-world camera where, you only focus on things like exposure, f-stop, shutter speed etc.


Biased renders on the other hand focus more on speed rather than on light quality or photorealism. To this end, they are
a) rarely ever physically accurate....
b)considerably optimizable for speed through a whole bunch of settings and dials and ....
c) typically very fast (relative to unbiased renders). Commensurately, the light-dispersion quality of renders produced by biased renders tend to be considerably lower than those of unbiased renders albeit much cleaner ( less noisier). You can, of course tweak the settings to approach realism as much as possible, but of course at the cost of speed.
d) have a whole host of settings and things to tweak ( which is both good and bad seeing as this is what allows to to optimize the renders, but conversely makes biased renders notoriously difficult to learn and master, with all there is to remember.)

In essence, biased renderers, allow the user to "cheat" the physical limitations of light simulation and calculation by allowing one to sacrifice things like specific object shadows, caustics and cautic relfections/refractions, and by placing limits on the depth of light dispersion or how far the light is allowed to bounce around the room or scene before cutting off (unbiased renders theoretically calculate this to infinity, if allowed to go as long; biased renders operate on the premise that the human eye and human perception does not need that much resolution - which is true to a certain extent).

Consequently while biased renderers (Vray, Mental ray, FinalRender etc) tend to look cleaner and render faster, they also tend to look very "CGey", "plastic" and not as photorealistic as unbiased ones; especially when done by novices and less experienced users. While on the other hand, unbiased renders (Maxwell, Fryrender, Indigo, Busyray, etc) produce more realistic renders, with less effort, but conversely tend to be "noisier" or speckled, probably because the render was cut off before it could be completely resolved. Unbiased renderers are starting to become more popular thanks to their lower requirement for technical know-how ( i.e less effort and brainpower required) and also due to the fact that computers are getting faster and cheaper to allow users to overcome the slow-speed issue with networking and multi-core PCs.

I hope this helps.........