Eevee:
Cycles:
Eevee is definitely not my friend. I have the feeling that, for any moderately complex scene, Eevee will not suit my rendering needs (unless I’m aiming for some particular rendering style)? And in that case, it feels like a pity to spend tons of time tweaking shadows, reflections, AO, etc etc, when in the end I’m going to do a long, offline Cycles render anyway. Maybe when modeling a single (complex or not) object instead of modeling a full scene, Eevee is an easier first choice?
Nonetheless, after some tweaking, I was pleasantly surprised about the results, especially in those areas of the image where “not much interaction is going on” (e.g., the pin on the right, rather than the bowling ball and all surrounding pins throwing shadows and needing reflections, incidentally right on a highlighted part of the bowling alley). Not sure what’s throwing that dark spot on the top left (some bias parameter or such that isn’t properly tweaked), and on highlight-spots (e.g., on the bowling ball) we’re missing shadows (or, more accurately, not shadows but we don’t see objects occluding the light source) in Eevee.
My main source of frustration with Eevee: increasing values isn’t necessarily “heavier computation but more realistic results”, it’s oftentimes just a tradeoff, and depending on the scene it seems to mean “improving one part could worsen other parts” … I could be focusing on a part of the image, set some bias such that it all looks perfect, render it all out, and then realize that the changed bias caused artifacts in another part of the image (and if you changed a lot of parameters before noticing the artifacts, it becomes harder to track down the cause). Especially with lots going on, you can’t keep checking every single detail for each parameter change … Which is why I would think Eevee is more recommended when modeling a single object or a basic scene, rather than scenes that are moderately complex or worse?
Common issue, blame the lecturer for not making the point of getting them right!