Dear Janggut,
Allow me to try to make a very important point clear.
It is not true that a “real 3d runtime engine” rendered picture is better than a pseudo 3d displayed picture.
Let me go verbose here to make it even clearer.
Given any system with a maximum computation power and instruction execution in X MIPS, we need a certain time to calculate a screen.
In real 3D we are forced to finish rendering within the time frame of a single frame, which is typically more than 15 frames per second, optimally 24 frames per second and superbly at 100 frames per second.
With this in mind, it should be obvious that we have a limited resolution that can be handled to form a 3d screen using polygons, the three corners of each, which have each three coordinates in the (x,y,z) Cartesian space. This means that the minimal surface is a triangle, where each corner must be defined in space by three integers. To increase the resolution by a factor of two from 800 to 1600 and from 600 to 1200 means that we have four times the first area not just twice. Therefore we call it a geometrical rather than an arithmetic multiple.
In a three dimensional world this climbs to cubic measures rather than just being the square of the first.
That is why Zooming out in real 3D must be a resolution catastrophe. That is why resolution must be sacrificed because we have a fixed time frame in which the data of a screen’s frame must be ready to display.

When pseudo 3D isometric screens are used, this is not a problem at all, because the entire world’s map is previously rendered at its leisure and slow pace for the highest resolution. Once a perfectly rendered world is ready to save it is saved to disk and later on a specific crop of that whole world is addressed for being displayed.
When the player chooses the highest resolution and zooms in, the map is displayed at 1:1 scale. Zooming out from there simply requires one of three famous techniques to recalculate the map, but it takes time and must be performed during a game pause. An alternative would be to save zoom presets (which demands more disk memory) such that the player may choose one of three or four zoom levels; with one map for each level per screen resolution. The same applies to sprites, which are much better looking than 3D real time rendered polygonal objects and characters. A sprite could be the product of infinite curvature and textured surfaces rendered to perfection, but a real 3D character does not have enough time to be produced in such a high quality.
The consequences are obvious.
Pseudo 3D have superior in graphics quality but they consume disk memory like hell, and are very limited to manipulate in real time.
Real 3D is inferior in graphics quality but they hardly consume disk memory and they can be superbly manipulated in real time during game play.

A mixture of both is obviously the best answer because still scenes suck when they are real 3D rendered. OTOH all moving objects do not need to be rendered at high resolution as the eye never really catches any details due to the motion blurring effects. The art of displaying a still hero’s sprite that is real 3D rendered on motion is something I did not see until today on the game market. It is simply based on a first and last sprite encompassing a set of real time 3D rendered mini-frames. This technology should ensure superior graphics quality with the minimal impact on memory demands. World rotation is yet another horrifying issue. There is no way an isometric map could be rotated simply because there is no data for defining any other viewing angles.
That is why the concept of rotating a world demands that that world must be defined as real 3D polygons and then the graphics quality deteriorates when limited time frames dominate the rendering possibilities of each frame in a scene with rotations, panning, zooming, tilting and yawing.
Add to all those problems (as if they were not enough) the problem of casting true shadows and ray tracing, then you have come to a single frame rendering time that varies from seconds to minutes.
In my old golden days, I remember to have rendered a single picture in three days on Amiga 1000. <img src="/ubbthreads/images/graemlins/biggrin.gif" alt="" />
Blizzard in WCIII took the decision to make a real 3D G. E. and all its characters are very low resolution polygon-based objects but they added a fantastic touch that made that bad resolution become tolerable.
they added a factor of life to those objects, and you would see the characters breath air constantly, flags flutter under the wind as does some of the leaves of trees. This effect gave a new dimension to the animated scene that made low resolution “forgivable”.