Energy conservation solution is a half measure

[ViperCGI]

Ok, this has been something that I've been meaning to send as a ticket for a while, but after I saw Vlado's update about this, I decided to wait. The problem is the solution isn't complete. This is a test scene done with vray 6 and vray 5 for comparison



It has improved from vray 5, but if you compare to renders from redshift and corona, you can see it still has some way to go before we have a full solution. It's specially egregious on the glass.



Whatever fix was applied seems to have only been applied to the reflections. But we need diffuse and refraction fixes as well to make it perfect. A white furnace test for the diffuse roughness shows that the problem persists there:

It would be great if this could be fixed before the full release. IMO, this is very important so that we can have realistic feedback when constructing shaders, and don't have to fight the settings to get correct results.

[vlado]

For the moment we have focused on reflections only; it was complicated as we also had to take into account the GTR gamma together with anisotropy which make the task way more complicated than in Corona or Redshift.

We will get to the diffuse roughness and refractions in an update - luckily they are easier to do but still need to be carefully tested.

Best regards,
Vlado

[ViperCGI]

I'm sure you are aware of it, but the sony paper goes into detail of how they fixed it for their own tools: http://www.aconty.com/pdf/s2017_pbs_...rks_slides.pdf

They even fixed it for SSS which I forgot to mention on my initial post.

[ViperCGI]

For the moment we have focused on reflections only; it was complicated as we also had to take into account the GTR gamma together with anisotropy which make the task way more complicated than in Corona or Redshift.

We will get to the diffuse roughness and refractions in an update - luckily they are easier to do but still need to be carefully tested.

Best regards,
Vlado

You posted this just as I was doing my other post, so apologies for not acknowledging it. But ok, good to know we'll have a solution in the future at least.

[vlado]

Of course we looked at all papers we could find; none of them deals with the GTR BRDF in a practical way so we had to do some serious research

Best regards,
Vlado

[Alex_M]

In the Corona image it looks like you have caustics turned on while in the Vray one they are missing. Could that be the reason why glass looks different in the Corona image? I agree that the glass in the corona and redshift images looks more realistic though. Here's what a similar object looks like in real life. It does seem closer to the corona and redshift images. I vaguely remember Vlado's thoughts/post on this issue too. I am sure they've been working on an improvement and will present it in an update as soon as it's ready.

[ViperCGI]

Fair enough! I have no idea how complex this problem really is, so forgive me if it sounded like I was trying to say it wasn't. Vray is a big investment for me as I'm just a hobbyist, and things like this are important to me. But as I said, as you guys are aware and a solution is coming, all good

[ViperCGI]

In the Corona image it looks like you have caustics turned on while in the Vray one they are missing. Could that be the reason why glass looks different in the Corona image? I agree that the corona and redshift frosted glass looks more realistic though. Here's what a similar object looks in real life. It does seem closer to the corona and redshift images. I vaguely remember Vlado's thoughts/post on this issue too. I am sure they've been working on an improvement.

I rendered the corona image a few months ago, so it's possible that I turned caustics on. Vray with caustics on looked way worse though. I think caustic improvements must happen in the future as well.

[kosso_olli]

Shaders like these are exactly what we would have needed many times in the past. We used tricks to get the right look for frosted acrylic glass, but here it works out of the box.

[matthew999]

this post is so enlightening, maya user here, how educational it is reading the max forum. I have always wondered why my frosted glass goes dark and never the milky look that i expect when comparing it to a referance sample. Cant wait for energy conservation to be fully supported.

[seandunderdale]

I have always wondered why my frosted glass goes dark and never the milky look that i expect when comparing it to a referance sample. Cant wait for energy conservation to be fully supported.

Same...I always felt like I was doing something wrong when I turned off fresnel to get the render to match reference.

[JonasNöll]

Contrary to what was mentioned here in this thread for me the difference seems to only affect refractive materials and a little bit of dielectic materials, while metallic materials seem quite unaffected. Shouldn't they have the biggest change?




1: Metallic Material
2: Refractive Material
3: Dielectic Material

The refractive materials work much better now though

[kosso_olli]

Well, it seems the full evaluation has made it into the beta 2 release. Is this correct?

[JonasNöll]

Well, it seems the full evaluation has made it into the beta 2 release. Is this correct?

Seems, but as said it somewhat disappeared for pure metallic shaders. In the 1st post there is a very big difference in the reflection, and I think it was also there in beta 1 when I played with it, but in beta 2 so far the changes for metallic shaders seem to be minimal.

[seandunderdale]

That second row blurry glass looks super cool though!