My god, it's full of stars!
Well, it looks very pretty. It was already so with just the sprites. Is there a compromise in performance, making them full of real stars on the fly?
In this earlier thread, I have already presented some views that can be expected:
Imagine the Solar System was part of a Globular Cluster
As a "Gedanken experiment" I simply had placed a globular cluster around our solar system. The respective screenshots should give already some feeling about the "sky experience".
My new globular cluster (GC) stars can hardly be compared with my old sprite-based stars from Celestia. Neither as to rendering performance nor concerning their luminous appearance!
Firstly, in celestia.Sci, the GC stars are rendered with the same shader as my normal shader stars, implying that both types merge perfectly as to the spacial distribution of luminosity, apparent star sizes and color. The old sprite stars were not rendered with GLSL shader technology.
Next, for rendering the large amounts of GC stars I am using the most modern (and fastest!) OGL technology , a "VBO = VertexBuffer Object". VBOs allow vertex array
data to be stored in high-performance graphics memory
on the server side and promote efficient data transfer. This is the optimal solution for rendering large numbers of objects in one go, provided inherent updating constraints on these objects "factorize" in some way.
Of course I have done benchmarks and the new stars are rendered much faster.
But the results (also relative ones) do depend on the abilities of the graphics hardware.
The most striking visual improvements concern the luminosity shapes of the individual star disks. While in close-up view the old sprite stars have a "dull" and blurred appearance, the new shader stars are really "shining" stars. Moreover, due to the different technology, the magnitude range is huge from the brightest down to 25m
. Here is a little comparison for the stars of yellowish M 4, both as seen from a closer distance:
The tremendous "shining star" effect is also well visible in this image from the earlier thread I quoted above: