This is set at 600 samples, 50% resolution, Reflective and Refractive caustics turned off, and filter glossy set to 1.
I’m also using the Principled BSDF instead of the glass shader if that makes any difference.
I’m also using the Principled BSDF instead of the glass shader if that makes any difference.
Are you sure it’s the glass? What happens if you just remove it?
try branched path tracing with both sample all direct and indirect on.
try setting both min and max bounces to the same.
use lamp objects and not material lights. emission materials can be 1 for visual.
try disable multiple importance for lamps.
“But also, think outside the box …”
While one way to “put those automobiles ‘behind glass’” is to actually do it, another approach is to create the plausible-enough effect by other, more controllable and less-expensive means. These techniques will necessarily involve compositing.
Although modern renderers can provide an impressive range of “physically correct” algorithms – specifically including “glass in all its glory” – these can be overkill. The human eye looks for just a few specific visual cues, based on human experience, to say (say …) “glass.” Most of the time it is not necessary to “fully placate the Gods of Physical Reality” to sell a shot.
Mix glossy and transparent shaders for thin panels. There is no visible refraction so glass shader is a massive overkill.
If you do the transparent trick, make sure the normals are the correct way. Preview the fresnel output; it should be darker when facing and white at glancing angles. If you get a weird circular white shape in the middle the normals are in the wrong direction. If you want proper fresnel from both sides, you need a node setup that inverts (1/IOR) for the backfacing face.
I agree, using proper glass for thin panes is overkill (with or without thickness, unless thickness is significant enough to cause a visible and desirable refraction). Also note that glass shader on a thin face (no thickness) would suffer the same problems as mentioned above.