Hi all. Very late to the party, but I am finally trying Karma in place of Redshift.

Problem 1: I have a classic 'liquid in a glass' model. The liquid slightly intersects the walls of the glass - as is standard. I've got Karma standard materials set up in Solaris. Nested dielectric property for the glass set to 1 and 2 for the red liquid. Everything renders as expected if I move the liquid outside of the glass - but the liquid is opaque if it's in it's normal position inside the glass.
Can anyone help?

OK... having wasted so much time chasing this, I've now realised it renders OK - to MPlay or disk - but doesn't display correctly in the Karma stage viewport. I'm hoping there is a button / switch or something that is causing this, otherwise a renderer that doesn't render what I see in the viewport isn't much use to me. How am I supposed to adjust materials? Render for every minor tweak?

Maybe the viewport renderer has too low a transmission sampling limit?

I ran into a similar issue when rendering water with bubbles in it, and it turned out to be due to having the transmission samples set too low.

Thanks for the thought. I've ramped up the sampling limits to the max as a test, but after some more experimentation, it looks like an XPU limitation to me, but I'm hoping it might be user error and someone will educate me : )
It renders as expected with Karma CPU, but not with XPU - as below.

Have you tried tweaking the ray bias?

Thanks brians. I was returning to report that I'd discovered this was a scaling ('aka: ray bias') issue - only to find your post.

What I would ask you to note is that the original model - with default ray bias - rendered correctly in both Redshift and Karma CPU - but failed (as above) in Karma XPU. So it looks like there could be inconsistency in default ray bias settings here. That causes confusion.

Mike_A
So it looks like there could be inconsistency in default ray bias settings here

There is inconsistency between CPU and XPU but it is explained in the tooltip for the Automatic Ray Bias parameter as well as in the docs. In case anyone else is wondering:

Automatically compute ideal ray bias. Under Karma CPU, automatic bias applies to everything except procedural mesh and continued rays for partially opaque surfaces and nested dielectrics (the “Ray Bias” property is still used for those cases). Under Karma XPU, automatic bias applies to polymesh path bounce and polymesh shadow rays only. For everything else (eg SSS, rounded-edge, nested dielectrics, points, curves etc…) the “Ray Bias” property is still used.