Test render w/ engine 4

[Rich Rosemann]

Engine 4 w/ hdr sky and image as background. 1900 passes.  The tank walls have displacement added to the image.  Those pesky speckles.  Poor choice of displacement?

 

VERT%20SKID-1.jpg

[Roy Hirshkowitz]

The displacement's fine-- although a lot of unnecessary work (bump map is probably better.)  The fireflies are definitely caustic paths-- mostly direct sun coming off of the shiny pipes.  Looks like another 2000 passes to go before they resolve(!).  

I guess I could throw in a switch to exclude reflective caustics.   Kind of defeats the purpose of the new engine though.

[Rich Rosemann]

Sounds like a weekend of rendering.  As a side comment, I'm doing another rendering of the same view using Path Tracer and most of the speck's have resolved after 110 passes.  Does this sound right?

Will change displacement to bump on next test.  Thanks.

 

[Roy Hirshkowitz]

The Path Tracer doesn't include any caustic paths-- so you shouldn't see the same type of fireflies.

[Rich Rosemann]

Test rerun using bumps instead of displacement.  Cleaned up well.  Can I assume displacement works best on flat surfaces.

 

[kingjin]

Rich Rosemann, Hello, E4, rendering it, how long, how many PASS

[Roy Hirshkowitz]

Displacement works a little better on flat surfaces.  

The main problem is that it's an expensive option to calculate and will significantly slow down the rendering.  This will prevent the renderer from being able to resolve things like caustics in a reasonable amount of time.  Only use it in cases where it's worth it.  In other cases use bump mapping.  In your example, you're not close enough to the tanks to notice a difference.

A secondary problem is that displacement mapping is more prone to aliasing artifacts than the other methods. For curved objects it does depend on the quality of the tesselation I get from AutoCAD (cylinders are likely to send in really crappy triangles.)  A linear pattern, like the one you're using, is even more prone to displacement mapping artifacts.

Rich Rosemann said:

Test rerun using bumps instead of displacement.  Cleaned up well.  Can I assume displacement works best on flat surfaces.

 

[Rich Rosemann]

kingjin said:

Rich Rosemann, Hello, E4, rendering it, how long, how many PASS

 

 

Rendered at about 150 passes.  Not sure of the time. Over night.   1978x1280.

The .jpg itself looks better then the image shown here.   Will run it about 900 passes.

 

 

[Rich Rosemann]

Roy Hirshkowitz said:

Displacement works a little better on flat surfaces.  

The main problem is that it's an expensive option to calculate and will significantly slow down the rendering.  This will prevent the renderer from being able to resolve things like caustics in a reasonable amount of time.  Only use it in cases where it's worth it.  In other cases use bump mapping.  In your example, you're not close enough to the tanks to notice a difference.

A secondary problem is that displacement mapping is more prone to aliasing artifacts than the other methods. For curved objects it does depend on the quality of the tesselation I get from AutoCAD (cylinders are likely to send in really crappy triangles.)  A linear pattern, like the one you're using, is even more prone to displacement mapping artifacts.

Lesson for the day.  Thanks Roy.