Irradiance vs Brute Force

[glorybound]

When you say 1/4 res., what do you do? Do you turn off your reflections, create your low resolution IRR map and use that for your final.

I use almost same IRR settings with most my work, I rarely have to tweak them. The Image I compared I used 1/4 res on the IM map and I cant tell the difference from the BF one.

[glorybound]

I have turned off trace-reflections, using VMC, and rendering it out for both LC and IRR. I leave trace-refraction on when I have windows. It is a pain to turn things off/on, but it was always much faster.

[eyepiz]

When you say 1/4 res., what do you do? Do you turn off your reflections, create your low resolution IRR map and use that for your final.

Just turn the min max in Irr down. I do -5/-2 on a 4k renders, that's it..I could be wrong though I read this somewhere a long time ago. 0 is full res -1 half and so on.

I don't turn off reflections seems like to much work.

[dlparisi]

Next I took irmap and went from -1 to 0 and render time went up to 8 min 40 sec. This is mostly on prepass as the actual render was very fast. I can say that with this quality its acceptable for the edges and precision, I've also reduced the hsubdivs and interp samples to about 3rd for 0 res its applicable.

Thanks Dimitry for all of your tests and lengthy post. Can you please clarify what you meant about the reducing the subdivs by "3rd for 0..."? I didn't quite follow what you were getting at.

[Donfarese]

@Morbid Angel, I tried your scene, took forever to render on a 3930k, seems to be the displacement slowing it down from the carpet tremendously. I wanted to just test it in Vray RT with BR, LC. rendered of course really fast at 1min30sec. But i have some really weird shading at room edges and ceiling, like it has bad AO turned on or something. never seen this before.

[Morbid Angel]

Thanks Dimitry for all of your tests and lengthy post. Can you please clarify what you meant about the reducing the subdivs by "3rd for 0..."? I didn't quite follow what you were getting at.

What I think is, lets say for a given resolution of 1280x720 your irradiance map can produce good quality result with reasonable time. But if you simply raise the resolution of your render, then you obviously produce a much higer density irradiance map, which is almost like brute force anyway. So I think the power of irradiance map is that for larger resolution renders you can reduce the number of samples for it.

[Morbid Angel]

@Morbid Angel, I tried your scene, took forever to render on a 3930k, seems to be the displacement slowing it down from the carpet tremendously. I wanted to just test it in Vray RT with BR, LC. rendered of course really fast at 1min30sec. But i have some really weird shading at room edges and ceiling, like it has bad AO turned on or something. never seen this before.

hmm this is odd, the ao is not on. As you can see in my tests. It appears you might have detail enhance on with bf in irmap is that the case?

[Donfarese]

hmm this is odd, the ao is not on. As you can see in my tests. It appears you might have detail enhance on with bf in irmap is that the case?

No detail enhancement, that is just with IRmap, and RT doesn't support IRmap. don't know what it is.

[Morbid Angel]

well the point of this exercise is not really bound around RT or use of RT. I was just showing that at 1 to 1 res, irmap can be as slow as bf.

[dlparisi]

What I think is, lets say for a given resolution of 1280x720 your irradiance map can produce good quality result with reasonable time. But if you simply raise the resolution of your render, then you obviously produce a much higer density irradiance map, which is almost like brute force anyway. So I think the power of irradiance map is that for larger resolution renders you can reduce the number of samples for it.

Thanks Dimitry. I understand the relationship of the IMap to resolution and have reduced the min/max samples accordingly as I scale the image up but I don't understand how this applies to lowering the subdivs. Wouldn't you want Hemisph. subdivs of each sample (and probably the Interp. samples) to remain the same to maintain the quality of each sample?

[Morbid Angel]

You can, but at a cost of longer render time. That is why people render irmap at half or quarter of the frame res, then render final frame.

[Donfarese]

well the point of this exercise is not really bound around RT or use of RT. I was just showing that at 1 to 1 res, irmap can be as slow as bf.

Yeah I know, was just giving it a test, thought it would be fun. Good to test things like this to find bugs in RT, which maybe this might be.

[eyepiz]

This is an old article I read a while ago.. it's probably still relevant since Irradiance map has not been updated.

Irradiance min/max settings vs. Image resolution

http://vray.info/topics/t0101.asp

[Morbid Angel]

actually I did a test with progressive and it reached 0.006 noise thresh in about 7 minutes, which is faster then dmc...

[Morbid Angel]

I would wish irmap worked in a different way - ditch the resolution method and just simply have brute force samples, and their biasing. More bf samples - slower render time but better quality and smoother solution.