[Voidmonster]

I'm working with SJ's original test scene and doing a sequence of renders using different materials. In the process, I came across something that nicely illustrates some stuff that's been said here many times.

Both of these images were rendered in V1. Both have the exact same camera settings. (ISO 1500, shutter 15, fstop 200 [hey, it's what SJ had in there], burn .8, gamma 2.4, auto-exposure).

One image has a uniform material setting of RGB 230,230,230 and roughness 90. The other is RGB 200,200,200 roughness 90.

Can you guess which is which?

...

The answer is that the second image has the higher color value. It's darker and it took about 30% longer to reach the same SL as the image with the lower value texture.

So, the moral to this little story is: overbright textures REALLY, REALLY do slow down renders, make them noisier and generally provide poor results.

Ideally, I'd really like a checkmark on materials for 'clamp to legal values', but it ain't that hard to remember to use legal values.

[mverta]

Turn off autoexposure and re-run.

_Mike

[DELETED]

DELETED

[Voidmonster]

Material 200,200,200



Material 230,230,230

No auto exposure. The difference is even more pronounced.

[Voidmonster]

It is legal. However, the brighter the values the more oomph required.

Really.

[mverta]

That's what I expected on brightness, what were the time/noise differences?

_Mike

[Voidmonster]

The 230,230,230 one was rendered a couple of days ago and I don't remember the full specs for certain, but what I recall is that it took 20 hours to get to SL 19 before the noise smoothed out while the other lost noise by SL 18.61 in about 11 hours.

I wasn't keeping close track of SL/Noise/time stats since that wasn't really what I was testing... This was just an interesting indirect result of my primary testing.

[mverta]

Just to confirm: these compares use identical emitter settings, and exposure settings?

_Mike

[Voidmonster]

The only difference between the two renders are the color value in the materials setting.

[mverta]

Perfect.

_Mike

[mverta]

I'd give this test one more run with Lambertian and one with roughness at around 20, if you're up for it. Keep all other values the same.

_Mike

[Voidmonster]

That'll actually be a part of the larger series of tests I'm doing.

My hypothesis is that the difference that's getting measured between beta and V1 has to do with the way that beta deals with diffuse surfaces.

So, what I'm doing is rendering a series of images with different roughness values to compare the way that light gets distributed using the same method SJ did to compare Beta to V1.

The overbright thing was just a whim to see what'd happen.

[DrMerman]

Damn fine idea mate. Looking forward to it

[samsam]

Great test Voidmaster.

I can't believe that no one from NL has actually done any of this prior to the release of V1. Surely if they had the programmers would have put a levels warning inside MXED about capping the RGB value parameters.

Where I get really lost is what to do about texturemaps in the zero degrees reflectance slot. Do I need to desaturate/darken my texturemaps to below 200 for each RGB channel?

I wonder what other discrepancies the poor suffering userbase is going to discover in time?

SJ, Voidmaster if they invite either of you to become A-teamers please resist - otherwise you'll turn to the darkside and spend your days telling us about the next new fabulous release and how great V1 really looks!

[mverta]

V1 looks pretty freakin' good, actually:



From y_okaue's gallery. We're just all trying to make it better.

_Mike