Image Quality Increase Test (with images)

Started by moodflow, January 22, 2007, 10:29:03 AM

Previous topic - Next topic

moodflow

I know there was a discussion about this recently, but I wanted to do some more tests, and even go "all out" on one of them.

I rendered one at quality of 0.6, another at 0.8, and then another at 6.0.  The 6.0 does look better and seems to have a bit more depth, but the render times were incredible (see file names for stats).  The best way to compare them is to have the imags in a folder and switch between them in an image viewer.  You can definitely tell the difference.  My opinion is that the higher numbers ARE better, but the benefits drop off quickly as render times exponentiate. 

More tests to follow.
http://www.moodflow.com
mood-inspiring images and music

moodflow

I forgot to add the tgd file in case anyone else wants to run some tests.

http://www.moodflow.com
mood-inspiring images and music

edlo

Great work on setting up all these tests together Pal thanks a bunch.

moodflow

No problem, this will help all of us.

Also, this is the same .tgd Superza posted in the water render times post, but without the water.
http://www.moodflow.com
mood-inspiring images and music

Dark Fire

Perhaps you could combine all of the groups of images with different settings you have uploaded into a single thread and somebody could make it into a sticky.

§ardine

Quote from: Dark Fire on January 22, 2007, 12:30:27 PM
Perhaps you could combine all of the groups of images with different settings you have uploaded into a single thread and somebody could make it into a sticky.

Yes I agree these are very helpful - Thanks again

rcallicotte

I'm not sure how accurate it is without there actually being Global Illumination (like sunshine) in the environment.  What are you testing?
So this is Disney World.  Can we live here?

Superza

Hey mood great Work! btw 6.0 quality it's really insane lol!!
I don't know, btw to that point i have opinion that the more "depth" you want achieve the more quality you need.
Anyway, if you look to the Rerò work i did some weeks ago
( http://www.renderosity.com/mod/gallery/media/folder_136/file_1355209.jpg ),  1.0 is really enough to photographic quality if you found the right settings.
Imho also 0.9 imho is a great settings and it  reduces 50% the render time.. i mean... currently i'm running a render @ 1200*750 0.8 quality, 4 GI, 4 aa with a Pentium D 945,  by 130 hours and it didnt finish GI anymore if i use on this 6.0 quality i think one year of render is not enough,
Man i need a liquid cooled cray!!!
:D

oggyb

Focussing on one point of the image and flicking between them, there is very VERY little difference between all of them, and I'd happily settle for the 0.6 render because. . . just because.  For all I know you could be kidding us about the quality settings!

Lighting is the only difference I see.  Maybe 0.9 with a small raise in the intialiasing would be all that's needed to bring it up to 6.0 standard.

M.

§ardine

imho there is enough significant difference to raise the quality, although I agree with Superza 6.0 does sound pretty insane  :)
The main difference I see is right at the point of focus in the picture - sun and the horizon

I wonder moodflow, if you had the time maybe include one more render at 1.0?  :D
I wouldn't mind seeing how much it changes between that and the 6.0 render

anyway that my 2 cents

Oshyan

These are indeed useful tests, but I wonder about the sudden jump to *6.0*! Really a pretty "insane" value. ;D Under normal use I seldom go above 1, and never above 2. In the future, with improvements to render accuracy and certain current render issues, going above 2 will almost certainly be unnecessary, although I know currently some have seen improvement by doing so. But I would suggest that in these cases you may want to just accept that TG2 isn't ready to produce your scene properly yet and that future bug fixes and renderer improvements will give you better results without increases in render detail or time. So you may just want to set such scenes aside for now.

I do hope you continue these tests, but I recommend at the least that you cover the full range of the slider before going outside it. Also keep in mind that negative values often work and produce interesting results too.  ;)

- Oshyan

moodflow

Quote from: Oshyan on January 28, 2007, 11:38:13 PM
These are indeed useful tests, but I wonder about the sudden jump to *6.0*! Really a pretty "insane" value. ;D Under normal use I seldom go above 1, and never above 2. ...

- Oshyan

Well, I wanted to see just what would happen, similar to trying negative values.  Might as well go "all out"!   ;D
http://www.moodflow.com
mood-inspiring images and music

Rob Allen

I would consider a render time of 6-8 hours to be ideal for a final render, unless it is a very simple scene.  That said, I use TG2 for professional work and have a dedicated computer for it, so I can afford extra long render times.

Not sure if this was helpful, but I thought it might help to put things in perspective.