Unlike looking at a screen on a regular monitor, having a 3d model represent your entire perception of reality is quite unforgiving. It has to perform well at all times in order to not induce motion sickness or other discomfort.
In traditional video games the 3D graphic artists work meticulously to fine-tune each small part of the model and carefully plan out scenes for maximum performance.
When importing your 3D-model in runtime, we cannot do many of these optimizations that are traditionally “hand crafted” and therefore have had to sacrifice some visual fidelity for compatibility and performance/comfort.
In a fully automated process, there is still a bit of a way to go in order to make decisions on the same level as a 3D artist.
That being said, we are happy to announce that we are looking into upgrades in visual fidelity.
A few months ago we added a LOD system that, in cases where models are too heavy for the computer to handle - will remove geometry that is less relevant for your current point of view.
In other words, if the object is too small and far away for you to see it anyway - the computer does not render it.
In addition to this, we are looking into adding support for bump maps (a trick to get higher perceived geometric detail with less demand on the computer) and generally better material support.
All of these efforts are taken in small steps to make sure we do not compromise on comfort and user experience.
There will, however, be times where we can make bigger leaps forward rather than small steps.
It depends partially on available computer hardware and the innovation on that side, and also graphical engine development. Both of which are in a generation shift at the moment.
In all our new computer recommendations, we have included various graphics cards in Nvidia’s RTX series, typically ranging from RTX 2060 to RTX 2080 TI.
Previously, NVIDIA’s lineup was called GTX rather than RTX.
The difference between those lineups is that RTX cards now come with so called RTX-cores.
This is hardware dedicated specifically to something called raytracing, which deserves its own article but in short is the highest form of realism in graphical rendering.
It is going to be a while, probably a few years into the future, but it is definitely something that you can expect your scenes and models to utilize in the not too distant future.
Below I have rendered an example, with the first being graphically similar (if not worse) than current level in most models viewed in Dimension10.