The shadows in Odyssey are particularly problematic in VR, because they, as well as projection lights, tend to "detach" from the world, and lag behind it when one look around, which looks rather ridiculous, even if that silliness goes quite well with those not rare times when your character and chair cast shadows, but not the vehicle they are inside...
Horizons shadows were also subject to the limitations of their nature, but at least they didn't exhibit these particular problems... Don't know whether there are some optimisations involved, which e.g. selectively limit shadowmap update frequency, and require some sort of locked viewport...
There is inevitably always going to be some dancing shadows when terrain LODs transition, of course, but I think I'd rather have that terrain "pop" between LODs, than the current ground constantly "inflating" in front of you (even with LOD blending turned off), with rocks falling into place onto the morphing surface, as if we were playing BoulderDash...
I happen to like the harsher lighting in space in Odyssey (as well as space being black), and find it much more realistic than what the "foggy" full-screen post-effect colour filters, that were introduced in Horizons at one point, produced, but it doesn't come without problems -- there is frequent overexposuring, not helped by aggressive tonemapping that often seems to do the opposite to what you'd expect from something that simulates one's pupils contracting and dilating -- again something that is exacerbated in VR, because there the left eye and right eye views (which have stereo separation, so each has things that are visible, which are out of frame for the other) are tonemapped individually, so you often get a dark image in one eye, and a bright one in the other. This is, of course, a full-screen effect, so it affects exterior, interior, and GUI overlays (diegetic, but still...), indiscriminately.
E
always, since the very beginning, had this problem where the lighting of a planet does not seem to depend on whether
it is illuminated, but on whether
you are. I had a wierd edge-case the other week, where I was on the day side of a close-orbiting moon; The whole surface I could see was in full daylight during my approach, but as I descended, this must have made something line up, or being brought into evaluation, in such a way that the various limitations of the lighting engine made the whole moon suddenly turn entirely black. -Ascend back up a bit, and it lit back up like a lightbulb.
So my totally unrealistic pie-in-the-sky dreaming is that while the current engine build will be the base for the game for the next couple of years, with only minor tweaks and fixes, I very, very much wish that by the time the next refresh comes around, hardware and techniques will have matured enough that it has become possible to relegate rasterization to a museum, and have the rendering engine doing ground-up pathtracing -- no other game needs this quite as much as Elite Dangerous, which by its nature has to be mostly dynamic lighting and content -- no overnight prebaked lightmaps here...
Proper realtime shadows, realistic reflections/refractions, global illumination, etc, etc, and finally more than a single star contributing light in systems that have multiple.
No more characters that look like they are photoshopped onto the background, but instead exchange light with the environment they inhabit, and don't have their eyelids illuminated by light sources behind their heads.
The current engine would in this scenario remain as a legacy fallback, for players who do not have required new hardware.
"Wait", one might think now... "Didn't you say you're a VR player? Doesn't VR demand higher
everything, and isn't raytracing a performance hog?".
-Well, yes - these things are true...
...on the other hand; Given the nature of raytracing allows you to perform it randomly, on a per-subpixel basis, and that upcoming VR headsets are expected to include eye-tracking as a standard feature, several foveated rendering-, and dynamic quality techniques become possible, which save both rendering work and buffer memory -- greater relative savings acceptable the higher the resolution, and wider the field of view. This should be able to offset a fair bit of the overhead, if only somebody would develop it.
Yes -- it's never going to happen, of course, and many would class it meaningless graphics-tech-pron, but one can dream... :7
