Rombo.Metal vs MILA

Rombo.metal sports a multiple-scattering BRDF engine together with rough Fresnel and adaptive importance-based sampling. This allow us to out perform mental ray MILA both visually and computationally.

Multiple-scattering visually means much more bright and energy conserving materials.

mila_vs_rombomultiscatter

You can see how rombo.metal doesn’t look darker at high roughness thanks to its multiple-scattering BRDF. Also you can see how the surface doesn’t get unrealistically dark at its silhouette thanks to ‘rough Fresnel’. We call ‘rough’ a Fresnel implementation that is integrated at microsurface level instead of being computed at the macrosurface point being shaded. It simply means that the normal we use to compute Fresnel reflection attenuation comes from the microface and not directly from the surface normal passed in from mental ray (ie. state->normal).

Rombo.metal also uses adaptive importance-based sampling. You can see how the sampling looks better for example around specular highlights when using adaptive sampling.

rombo_bettersampling

This gives us reduced render times compared to MILA. First image is MILA.

mila_oq4_mq10
rombo_2-8-0002

We can further reduce render times by using the progressive mode in mental ray which is generally faster than standard regular mental ray sampling.

rombo_progr_adapt_2-4-005

You can see how with rombo.metal and progressive mental ray mode we can half the render time (being around 6mins) compared to use the default factory MILA and standard mental ray sampling (around 12mins).

Leave a Reply