On 2008.02.22 22:56:59 CET, David Fox wrote:
> On 2/22/08, Johannes Gajdosik <johannes.gajdosik@...> wrote:
> > In the meantime I believe that real speedup is only possible by
> > 1) moving the projection calculation from CPU to GPU (vertex
> > 2) use the other cores in multicore machines
> > 3) make massive use of OpenGL vertex buffer objects.
> Would it work well on lesser-capable cards such as Nvidia FX5200 or
> does one need something that's more capable to really take advantage
> of the GPU facilities? FX5200 is plenty for me right now, and I'm no
> bleeding edge gamer :).
Vertex buffers combined with point sprites could also speed up star
drawing on FX5200, while shader programming gives no advantage,
although it is possible. I have 2 FX5200 on my dual seated system at
home, too, and I am quite satisfied for ordinary use, including
> With the latest svn rev (2974) it appears that the frame rate is
> already sped up compared to what it used to be (17-20 fps most of the
> time) on my Athlon 1000 mhz box. But I'm now getting usually between
> 25-30 fps. I set things up so that time would be sped up fast enough
> to allow for a smooth slow scroll of the sky, and it kept up above 25
> fps, except just around sunset when the frame rate dropped a little
> bit, but not by much.
I like many stars:
mag_converter_mag_shift = -3.0
mag_converter_min_fov = 0.001
mag_converter_max_mag = 30
mag_converter_max_scaled_60deg_mag = 9.5
and get down to 5fps on dense star fields.
> > take some time, I am quite sure that when I am finished lots of
> > will have multicore machines with NVidia series 8 cards or better.
> I'd like having the hardware too :)
I don't have it at home right now, but at work. Anyway nearly every new
computer you can buy is already 64bit dual core and has NVidia 8
(provided it has NVidia at all).