Rendering software comparison

solo

Sounds like a reasonable idea actually.

Make it happen.

Max_B

Sounds like a reasonable idea actually.
Make it happen.

I'm quite new on SU, while more experienced with mechanical nurbs modelers.
So, I'm not the best qualified to submit a SU file, and opening a new thread without it is pointless. Hope someone here, if not the original poster, could propose a good test file. …and open a new thread, which should be kept on subject: submitting rendering test images and reports.

Jon

hmm... alright, but we can't just post any old model. Should be a model that incorporates every possible minute detail to see how different rendering engines handle different attributes.

-A variety of different materials glass, metal (aluminum / chrome), wood, plastic, ect.
-A surface to test bumpmapping
-soft shadows / hard light / spots / omnis / sun
-self illumination / lamp shades / translucency
-sub surface scattering
-reflections / specular stamping
-many different textures to add a material to later
-a good model, high poly with a lot of detail, but not too high

You get the drift.
Maybe a kitchen? Or a simpler scene like a chair and a shelf with a lot of nick knacks?

Gaieus

If such a comparison is made there should be both an interior and an exterior scene IMO.

Edson

max's suggestion is quite reasonable but allow me to say, as a novice in rendering, that given a number of capable rendering applications (we all have seem tremendous results being achieved in a variety of apps: KT, Podium, Maxwell, VRay, Vue, Indigo, IRender, and a long etc) perhaps the most important features in that comparison are:

how easy it is to use each of them? meaning: can you easily understand where are the main switches and how to achieve the basic tasks of rendering without a 6-month learning curve?
in other words, how intuitive its use is?
must you be an expert in rendering to be able to use it? some applications' interfaces are plain cryptic...
what about its integration with the modeller you use, in our case SUp?

other questions could be added to those but they are sufficient for you to get my drift.

dylan

It is always going to be a pretty pointless exercise in some respects, but one that we all would enjoy seeing the rsults of

It is a nice idea though Max submitting a default scene to be rendered by the variouse apps.

I just see too many variables to be able to ever get a definate result.

remus

Although it is a very good idea, and definetly better than one person attempting to use all the render engines, i still tihnk its too subjective.

When you render a scene, you are effectively interpreting it in a way that you think is appropriate. You do this evey time you apply a material, every time you tweak a setting, every time you alter the camera angle, and the list goes on. What looks right to someone, in terms of say a material, could well look wrong to someone else.

Having said that, it could still be an interesting exercise, and might still provide a better basis for comparison than is currently available.

edit: jsut realised i was a bit hasty there, basically i said in lots of words what dylan managed to distill.

Max_B

My suggestion would be:
The author of the SU file should define base materials in it, as well as the camera position, with, say, two scene/views, to allow interior/exterior. As for the lights, hdri, it's more tricky. I don't know yet if lightning definition is precise enough in SU (edit: no light in SU, right?).

Then the author of the rendering should give indications on modification made on materials and lights, along with main rendering options, radiosity, scattering, caustics and such.

The goal of exercise should not to reach the ultimate capabilities of a renderer, rather see what can be done in a reasonable (Edit: 2-8 hrs maxi? ) session by a skilled user.
For sure the learning curve is not accounted for in such comparison. Could be useful if each poster added a (subjective) comment about this point.

I'm certain that there are appropriate files available around. I have a model in progress and could submit it when finished, but I'm definitely not the most qualified.

Stinkie

@max_b said:

The goal of exercise should not to reach the ultimate capabilities of a renderer, rather see what can be done in a reasonable (Edit: 2-8 hrs maxi? ) session by a skilled user.
For sure the learning curve is not accounted for in such comparison. Could be useful if each poster added a (subjective) comment about this point.

A "reasonable session": I cannot for the life of me imagine that 2-8 hrs would be sufficient to render a grainfree image of an interior using an unbiased renderer.

Wouldn't it be more useful to have two separate comparison experiments? One for biased renderers, one for unbiased ones.

remus

Personally i think we should try and have a more scientific test scene, where a variety of materials and lighting situations occur. I say this as i reckon if we just use a normal scene it will be very ahrd to do a clear comparison of lots of aspects of the render engine.

solo

LOL, see even the parameters are not agreeable.

Think of SU as a drawing set with pencils and stencils, think of rendering as an illustration set filled with paints, easel's and paper textures.
Now there are different illustration sets, some have water paints and others have oil paints, heck some have cheap and quick drying Walmart poster paint. How the artist uses the set of tools makes all the difference, and that really is the crux of the issue.
Like art through the ages many well known artists started as an apprentice to a master in order to learn the disciplines before branching out on their own and developing their own style. Rendering has it's learning curve no matter how simple the illustration pack is, in fact a simpler (less features) pack will require one to be even more artistic to achieve a more realistic result due to it's limitations, but that is not considered a software learning curve as that falls into technique and talent. As for a pack that has everything in it at a press of a button, that takes a long time to learn the intricacies and the tweaking in order to obtain realistic results and can easily create stunning images once the learning demands are met.
Without getting into biased and unbiased engines, lets view them as poster paints versus oil paints instead, they both can give a photorealistic result if used by the right artist however one will take longer to dry than the other which for the sake of the image should not be a constricting factor, however as a choice of tools to do a job it certainly can be if speed is needed.
So summing up I would say that the choice of render engine is very important for what your needs are, render engines are just a tool and equate to roughly 50% of the resulting image as the balance is talent and skill.

If you are to have a model for rendering, do not restrict the artist as the result will be restricted.

Edson

extremely well put, pete. wise words that sum up the issue, IMHO.

Stinkie

I agree. In any case, there's tons of images on this site and elsewhere that show the capabilities of the different rendering apps already.

IMO, this kind of comparison experiments are born out of the 'need' to find out what the 'best' app is. Ain't no such thing. I know that much by now.

Max_B

@remus said:

Personally i think we should try and have a more scientific test scene, where a variety of materials and lighting situations occur.

So where are we?
Do we try to select a test scene? Or do we stop on "comparison is nonsense"?
We all know that such test is not ultimate judge. It could be useful anyway. I'm a bit surprised by the resistance it raises.

remus

Personally id quite like to see it done, although i havent got any time at the moment (20 page report due in 2 days )

Maybe if i get a bit more time later on i'll try and come up with a test scene.

Al Hart

Be sure to include solid materials (such as wood below - which automatically molds to the curved surfaces), and fractal plants - which create random, complex, detail without clogging the SketchUp model in the criteria for a good renderer.

(Just kidding of course - but it is an example of how a renderer comparison could be slanted to favor a particular rendering engine)