Page 1 of 1

Some thoughts on current and future features

Posted: Fri Jul 08, 2005 2:29 pm
by jamestmather
hi,

It's early days yet for me and maxwell (having defected from vray (which I still love - but something about those stunning maxwell renders draws me to the new renderer like a junkie to drugs). In short; you had me at hello.

Firstly, congratulations on developing a new way of looking at rendering in a field dominated by ray tracing and global illumination - its tough to go out on your own with a new idea but the results look worth it. Well done.

A few thoughts:

1. In my max beta version - the material editor ignores tiling at the bitmap level (no matter how much you bump up the tiling- it stays at the amount set by the UVW modifier - this caused some early distress but we're past that now)

2. The renderer seems to "hit" triangles (ie: in a large quad visible in frame it seems to occasionally render a line where the sub triangle is creating the geometry (as the max scanliner would with a non planar quad (you get a crease)). The polygons are, I assure you, flat (mental ray has no issues).

3. The batch render MXC file seems really messy and needlessly complicated (can the renderer not set these ques up as radio bottons (for those of us not inclined to programming) - also no matter how many times I typed -d in my mxc file I don't seem to get a display of the currently rendering frame (across a five machine renderfarm). Further the renderfarm seems to take some ques from the MXC file and some from the render output in max. Image files are set by the drop down in the maxwell render section in max not by the -output line in the mxc file. It reminds me of the horror of early softimage mental ray. I implore you not to go down the same track. Backburner has a wonderfully elegant and simple set up. A sample mxc file would be unvaluable.

- also maxwell litters the drive with lots and lots of files while rendering - any chance that it could clean them up when it's finished?

4. No where in the docs did it state that for motion blur to work in maxwell that Properties "object" blur must be implemented for it to work (again - some early distress over this)

5. Renders not involving skylight seem to be noisy (displaying similar to brute force monte carlo artifacting). Any tips or ideas on how to solve that?

6. Some sample scene files of various successful maxwell renders would be great so we could learn from them. At the moment I'm a little in the dark as to the best methods.


Thats it for now. Again well done and I look forward to october to seeing maxwell v1.

Many thanks

James Mather

Posted: Fri Jul 08, 2005 3:49 pm
by Maxer
Wow :shock: James those are great points, you get a 100 for the best first post I've ever seen! I agree with all of your points very well stated. :)

Posted: Fri Jul 08, 2005 5:20 pm
by psanitra
It will be nice to have small utility for setting all those parameters. radio buttons will be great with option to choose source file etc...

Posted: Fri Jul 08, 2005 6:33 pm
by jamestmather
mxc files. They're very messy and seem to me a complex solution to something simple that I would hope is given a GUI by the first version of maxwell. While I'm on the subject I wonder is it possible for maxwell to determine how many processors are connected even though a network and simply use them in processing the standard renders? (vray does something similar I believe). - or am I asking too much? :wink:

Thanks

James

Posted: Fri Jul 08, 2005 6:40 pm
by Maxer
I guess it could be similar to the way Backburner finds all of the available render nodes. This would probably work if Maxwell could be run as a service in Windows, but the current Maxwell manager needs some serious overhauling in both it's GUI and code before it can handle this kind of job.

Posted: Sat Jul 09, 2005 12:01 pm
by jamestmather
Yes, vray has a distributed rendering function which connects through the standard renderer using IP addresses (each machine has to run a program called vray spawner which then assigns each machine a bucket of the render - which turns even the most grindingly slow renders into workable jobs - it really is a joy to behold. For something like Maxwell which due to the new technology is always going to be a slower renderer than a ray tracer, implementing this sort of function would make it viable as a production renderer with tight deadlines. In short - distributed rendering (which it already has) but from within one button click in max. :lol: :lol: :lol: :lol: :shock:

Posted: Sat Jul 09, 2005 2:06 pm
by Maximus3D
Many of the points you have are good and i agree on those :)
However.. bucketrendering over a network would probably not be possible due to the way Maxwell works. If you split up a screen you won't get a even calculation all over if you have some systems faster and some slower on the network. That would end up in a very strange looking rendering.. hehe some parts are properly calculated while others are not.

/ Max

Posted: Sat Jul 09, 2005 9:06 pm
by jamestmather
Yes but maxwell already has a network option to use several processors to work on a single frame. I'm only suggesting that this work from within max instead of maxwells ms-dos batchfile system which is complex and time consuming. In maxwells case it''s not a bucket system but each processor seems to assist in handling the iterations of the renderer.

Posted: Sat Jul 09, 2005 9:47 pm
by morbid angel
not to mention that maxwell 3d motion blur only supports 2 subsamples making it 2.5 d blur..

Posted: Sat Jul 09, 2005 10:17 pm
by andronikos916
Very nice comments here...I have only to agree.

Congrats to James for his post and I welcome him aboard.

Maxwell need improvement is some fields - but now other render engines need that too cause of Maxwell ! :lol:


cy,
Andronikos