Norman Vine wrote:

> Frederic Bouvier writes:
> >
> > FG is not using the latest features new cards are offering and the 'old'
> > features are pretty maxed out now. This is not an explanation for
> > worst performance but just for lack of improvement. In the meantime,
> > you probably switched to another setup/resolution that can eat computing
> > power on your back.
>
> Implying that the underlying rendering library is the cause for 'lack of
> improvement' is a little harsh given the *many* changes since the
> last relese that have little to do with rendering :-)

I was not speaking about the underlying library at this point. It could
also be our use of it, or the use of it by the model loaders. Otherwise,
how could you explain the variations in framerate whether the buildings
and the bridges are in view or not.
Just as an example, it is not possible to share a set of vertices and
vertices attributes between 2 graphic primitives. Indexed arrays are
there but are never used, and if they would, the same vertex used by
2 adjacents primitives would have to be pushed twice. I didn't experienced
myself what could be the gain yet, but I don't see a potential loss.
And all this static data could reside in card memory instead of being
sent to the bus every frame.

> Has anyone profiled FGFS recently ?

That is certainly needed. I was very disappointed myself to see that
switching from GF3 to GF5900 didn't improve my framerate.
I got a small improvement when I upgraded my CPU.

There is another point with graphic cards. On windows, their drivers
can switch on several features that are not managed by the application.
FSAA and anisotropic filtering come to mind. Unexperienced users can
have them to their max ( just because these are default values ) with
a significant impact on framerate.

It is just weird that in Innis case, things are getting worse. FX5200
are not top end cards but ATI 9800pro are. Something in the setup of
these machines must be wrong.

-Fred



_______________________________________________
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel

Reply via email to