Well, none of them
I'd like if Devs could concentrate on some long standing issues instead.
Here you find a short list, i hope it helps.
DISPLACEMENTTexture Displacement in Octane has severe limitations (honestly, more limitations than any other engine i know) since it requires a single "map" to work. No way to use procedurals or mixing maps, or using more complex setups - i.e. mixing two different displacement maps using a procedural or a Dirt node, just to make a quick example. Everything has to be baked to a huge tileable map before displacement can use it, causing big slowdowns and also "killing" some possibilities.
BTW i'm not considering Vertex displacement as a viable solution. In most cases it needs a too high subdivision of the mesh in advance OR too high subdivisions into the node, for sure it's not a solution which can be largely used.
just in order to provide a quick example, plase check attachment. A tree with snow.
We do this easily in Corona, and there is just no way to do same in Octane.
With a simple setup in Octane, we can get snow on a tree using a Falloff map which mixes bark and snow materials.
Then, we'd like to give snow some thickness, and in order to do this we'd need to plug the Falloff-based mask in displacement: not possible.
I tried same thing with Vertex displacement and it's impossible to get a good result again without a ridiculously high polycount, which makes tweaking just too slow.
I could report several examplex of this kind of limitation but hopefully this is enough to understand.
CHAOS NODEAgain, current Displacement implementation is making Chaos node useless - and this is another highly needed feature.
When we map a large terrain with a PBR material (i.e. rock) we need to break tiling and Chaos node does this, but you just cannot plug chaos node into displacement since you have to bake it.
Baking will BTW result *again* in a tileable map, which will show tiling in displacement, killing the purpose of Chaos node itself. Plus, it will not even match with color, specular, etc, since these are not baked and thus not tiled.
The good new is that fixing displacement you will also fix this one automatically
VOLUMETRIC CLIPPINGActually i use Scatter voluemtric for distance haze or such atmospheric FX, but it has a nasty clipping issue which makes it unuseable for subtle effects such a delicate distance fog.
If you ever worked on a large environment scene you have met this for sure.
Try to use any volumetric tool for partially dissolving something at 50km from camera (i.e a mountain). You will see that when you try to fade the volumetric, it will suddently disappear.
Basically there is no way to get a subtle volumetric effect which is under a given level of intensity since it's "clipped out". No matter of ray epsilon or anything you can try, this will happen 100% of the times, both if you use volumetric in Render target or onto a geometry with Specular material. Hopefully this can be fixed.
This brings to another FR:
ATMOSPHERIC EFFECTSMost engines offer solutions for Z-depth fog, subtle distance haze and such.
Hopefully you manage to add something similar to Octane (working with HDRs and not just with physical sky) in order to make environment artist's life easier and avoiding boring post-processing operations.
BTW this could be part of a general Volumetric improvement, up to you how to implement this, but in general for environments a separate tool for this kind of FX is needed.
Thank you for your attention,
Paolo