Fryrender demo is up!
yes please, show us the results!
I did some test also, with MR and Indigo 0.8, and Mr was a lot faster (Cornell box with a glass sphere). I am testing Fry, for an interior scene, and it doesn't seem to be sooooo fast. But I know from Maxwell that the material tweaking has a very high importance in speed improvement.
another point: the rendering speed is important, but the setup-time for the scene is more important, and often underrated..
I did some test also, with MR and Indigo 0.8, and Mr was a lot faster (Cornell box with a glass sphere). I am testing Fry, for an interior scene, and it doesn't seem to be sooooo fast. But I know from Maxwell that the material tweaking has a very high importance in speed improvement.
another point: the rendering speed is important, but the setup-time for the scene is more important, and often underrated..
Hi CoolColJCoolColJ wrote:Just an SSS test, I still don't understrand Fry's SSS, so not really that similar but speeds are closer with lights. I like the lack of MLT artifact spots in Fry though.
Fry does seem faster outdoors though...
If you look closer you can see I Aero-Noised the cube, to clean it up earlier
could you please render the SSS scene with the cube again but this time use a normal shine plastic material for the cube ? I am asking you for this because i have some suspicion toward Fry and Maxwell regarding material creation .... ( of cause i am not sure it is just a feeling
)please render the scene with the same camera angle because what i want to see have to do with "correct Fresnel" reflection and with this camera angle you should see it quit good .
One thing more .... please don't use coating for reflection ... i want to see a standard Fry plastic material ( with two layers .... the one that comes with Fry )
if i am right , the Indigo material will look different ( and i think more correct ) then the Fry plastic material
I hope it is no problem for you to make this render test
thanks
Greetings Patrick
hey guys, I'll see what I can with some comparisons
for sure, materials effect render speed a lot, reflective heavily bump mapped objects render very slowly and noisey....
the one thing that bothers me in all these supposedly accurate renderers, is the difference in tonal balance and lighting with the same scene....
Maxwell seems to be using D75 colour balance all the time as well
for sure, materials effect render speed a lot, reflective heavily bump mapped objects render very slowly and noisey....
the one thing that bothers me in all these supposedly accurate renderers, is the difference in tonal balance and lighting with the same scene....
Maxwell seems to be using D75 colour balance all the time as well
BTW re digital cameras being blurry, well 90% of digital cameras these days can't create a colour from each pixel of the sensor. They normally use 3 pixels of green, blue and red to do so.
So you will never get a sharp image with interpolation of 3 pixels
but there are a few cameras that use a sensor that will allow proper colour for each pixel and the difference in sharpness between these and the regular ones is pretty dramatic
So you will never get a sharp image with interpolation of 3 pixels
but there are a few cameras that use a sensor that will allow proper colour for each pixel and the difference in sharpness between these and the regular ones is pretty dramatic
Hi CoolColJ
regarding the comparison render ..... i think that a basic shine plastic material on a cube should not take that long ( no bump map) .
The thing is that "Fresnel reflection " is more visible on a cube like object then on a sphere ( having the camera placed in a way that you can see different angles of the cube ) ..... the way Fry and Maxwell make their layered materials make me think that maybe they are not doing it accurate
..... but only a test/comparison render will show the truth ......
Greetings Patrick
regarding the comparison render ..... i think that a basic shine plastic material on a cube should not take that long ( no bump map) .
The thing is that "Fresnel reflection " is more visible on a cube like object then on a sphere ( having the camera placed in a way that you can see different angles of the cube ) ..... the way Fry and Maxwell make their layered materials make me think that maybe they are not doing it accurate
..... but only a test/comparison render will show the truth ......
Greetings Patrick
Well your right Patrick, Something not right about Fry's And Maxwell's implementation
Man Indigo rocks, not only does the material look more accurate, but it rendered it cleaner than the other two
All rendered for 25mins. Used Bidrectional MLT with Indigo and it reached 290 samples per pixel. I tried MLT, and it looks about the same, with the cube shading slightly smoother and reaching 550+ samples, but with noisier shadows.
Just your basic Phong material in Indigo with an IOR of 1.5 and 1000 exponent
Had supersampling on in Fry, and if anything it makes edges more jittery, see the light for example. Had a hard time with Fry's strange blending of diffuse and reflective material to create basic Phong! I strated with the basic plastic preset. If I made the light highlight match Indigo, then the shading and reflection in the body are too sharp, and that's before you have to futz around with the right diffuse and reflective layer blends....
Maxwell, well I just used the plastic preset and again had to futz around a bit, but not as much as Fry. Well it has that look which I hate about Maxwell, this D75 colour balance and that washed out lighting with a dark cast that you can't remove even if you want to!
There is less noise in the diffuse materials though.
One strange thing that has always bugged me, Maxwell always saves bitmaps that are upside down....
funny thing, I originally had the scene 3 times larger, and Indigo rendered it fine, but the other 2 were creating black screens. Worked better after I scaled the scene down to 33%
Man Indigo rocks, not only does the material look more accurate, but it rendered it cleaner than the other two
All rendered for 25mins. Used Bidrectional MLT with Indigo and it reached 290 samples per pixel. I tried MLT, and it looks about the same, with the cube shading slightly smoother and reaching 550+ samples, but with noisier shadows.
Just your basic Phong material in Indigo with an IOR of 1.5 and 1000 exponent
Had supersampling on in Fry, and if anything it makes edges more jittery, see the light for example. Had a hard time with Fry's strange blending of diffuse and reflective material to create basic Phong! I strated with the basic plastic preset. If I made the light highlight match Indigo, then the shading and reflection in the body are too sharp, and that's before you have to futz around with the right diffuse and reflective layer blends....
Maxwell, well I just used the plastic preset and again had to futz around a bit, but not as much as Fry. Well it has that look which I hate about Maxwell, this D75 colour balance and that washed out lighting with a dark cast that you can't remove even if you want to!
There is less noise in the diffuse materials though.
One strange thing that has always bugged me, Maxwell always saves bitmaps that are upside down....
funny thing, I originally had the scene 3 times larger, and Indigo rendered it fine, but the other 2 were creating black screens. Worked better after I scaled the scene down to 33%
- Attachments
-
- CCJ_Phong_compare.JPG (104.48 KiB) Viewed 4010 times
Hi CoolColJ
thank you very much for this comparison render
and finally my suspicions seem to be right and Fry and Maxwell don't do Fresnel reflection accurate ....... i am not saying that Fry or Maxwell don't have Fresnel reflection ..... what i am saying is that their Material Editors don't allow to use them the right way ( Fry and Maxwell have the same Layered Material system and even the basic material is the same )
here i attached CoolColJ image with some indications to make it more clear what i mean ..... if you compare Maxwell cube and Fry with Indigo , you can see the difference very good
( it is curios that Indigo, Kerkythea and Radium ( free ware) are more accurate in this material test then the commercial render engines )
Greetings Patrick
thank you very much for this comparison render
and finally my suspicions seem to be right and Fry and Maxwell don't do Fresnel reflection accurate ....... i am not saying that Fry or Maxwell don't have Fresnel reflection ..... what i am saying is that their Material Editors don't allow to use them the right way ( Fry and Maxwell have the same Layered Material system and even the basic material is the same )
here i attached CoolColJ image with some indications to make it more clear what i mean ..... if you compare Maxwell cube and Fry with Indigo , you can see the difference very good
( it is curios that Indigo, Kerkythea and Radium ( free ware) are more accurate in this material test then the commercial render engines ) Greetings Patrick
- Attachments
-
- ccj_phong_compare_652_b.jpg (220.41 KiB) Viewed 3965 times
What would be the difference between that "layered" plastic and Indigo's phong, wich also is composed with a diffuse substract and an highlight layer ?
Mmh Patricks, if you look carefully in every image there is a different value of color between front and side face, the rest could be a tonemapping issue.
I'm just making myself "the lawyer of the devil" here (OT: does that expression exist in english ?).
Mmh Patricks, if you look carefully in every image there is a different value of color between front and side face, the rest could be a tonemapping issue.
I'm just making myself "the lawyer of the devil" here (OT: does that expression exist in english ?).
obsolete asset
Hi CTZn
Of cause it is good that you make the "the lawyer of the devil" part here
I don't know if you have read my "Kerkythea Material Editor Guide" where i explain why the "traditional" layers system don't work correct with Fresnel Reflection ( without mentioning Maxwell of cause but those that know Maxwell's or Fry Material system should understand it ....)
The reason is because in Maxwell or Fry you have to add reflection to the diffuse component but as you can not exceed 100% layer strength for all layers ( to avoid energy conservation error ) you normally setup a layer with 80% diffuse and 20% reflection .
Here is where the problem start ..... a 20% reflection layer will only reflect max 20% at 90 degree viewing angle
In other words .... the reflection Layer you have setup with an IOR (nd) = 1.5 will only contribute 20% with the diffuse component and this is very little to notice any reflection at all ...... so normally you have to use higher IOR ( nd) values to get any reflection showing up ( like IOR = 3.00 or higher)
The problem with Higher IOR values is that Fresnel effect get less evident ( reflection get more and more uniform the more you increase IOR value )
This is what we can see in the comparison render
i hope i could explain myself good enough
Greetings Patrick ( i predicted the result without even seeing the render comparison .... the problem is on how they implemented the Layered weight system )
Of cause it is good that you make the "the lawyer of the devil" part here
I don't know if you have read my "Kerkythea Material Editor Guide" where i explain why the "traditional" layers system don't work correct with Fresnel Reflection ( without mentioning Maxwell of cause but those that know Maxwell's or Fry Material system should understand it ....)
the problem here is that the shading difference are to small ..... it looks more like a uniform reflection strength and not like a correct Fresnel reflection you can see on the Indigo cube .Mmh Patricks, if you look carefully in every image there is a different value of color between front and side face, the rest could be a tonemapping issue.
The reason is because in Maxwell or Fry you have to add reflection to the diffuse component but as you can not exceed 100% layer strength for all layers ( to avoid energy conservation error ) you normally setup a layer with 80% diffuse and 20% reflection .
Here is where the problem start ..... a 20% reflection layer will only reflect max 20% at 90 degree viewing angle
In other words .... the reflection Layer you have setup with an IOR (nd) = 1.5 will only contribute 20% with the diffuse component and this is very little to notice any reflection at all ...... so normally you have to use higher IOR ( nd) values to get any reflection showing up ( like IOR = 3.00 or higher)
The problem with Higher IOR values is that Fresnel effect get less evident ( reflection get more and more uniform the more you increase IOR value )
This is what we can see in the comparison render
i hope i could explain myself good enough
Greetings Patrick ( i predicted the result without even seeing the render comparison .... the problem is on how they implemented the Layered weight system )
you normally setup a layer with 80% diffuse and 20% reflection
sounds arbitrary, specular should be prevalent if present AFAIK !You explained quite well, what (unbiased) softwares do it that wrong way again plz ?
On the other hand, as reinhard tonemapping is non linear you can tweak the materials to compress colors that way as we see it now. Still acting my role
obsolete asset
Hi CTZn
you are absolutely right that this is arbitrary and should have made people think already ( and also the unusual high IOR ( nd ) values used by Maxwell and Fry ...... normally around 3.00 is the lowers they use )
the 80% diffuse and 20% reflection is a example from a plastic material that comes with Maxwell and Fry Material pack
i will say it in a different way "Indigo , Kerkythea and Radium are doing it correct ....... the other commercial unbiased render engines that where mentioned in this threat , i am not convinced that they do it correct because of the way they handle layered weight materialsYou explained quite well, what (unbiased) softwares do it that wrong way again plz ?
"i don't believe that ton mapping have something to do with this because if the reflections are calculated "wrong" in the first step , there is nothing a ton map operation can do
Greetings Patrick
Who is online
Users browsing this forum: No registered users and 85 guests