I know it's jumping the gun, but you could actually buy one of these boards now:
[link lost in import]
Wait a month, and these boards will be under $300 from several manufacturers.
Is it reasonable to assume that the offloading of some OpenGL processing to the
video card will improve the frame rates in GPL? I get pretty good frame rates
with my Celeron 300A at 450MHz, but that damn mirror refresh BS is annoying. I
wonder if the GEForce256 driver will also have reduced performance with the in
car view.
Keith.
Probably not, GPL really isn't designed to take advantage of a lot of the
card's features. You might see a framerate improvement over say a TNT2
Ultra simply because the board, at least on paper, should be a lot faster so
you might be able then to set the refresh rates of the mirrors to 1 and
still get 36fps in***pit.
But thats just speculation.
>Is it reasonable to assume that the offloading of some OpenGL processing to the
>video card will improve the frame rates in GPL?
Don't assume that T&L processing will be offloaded without a patch. I
did a "strings rastOGL.DLL | grep gl" to see which OpenGL functions are
being called. To my inexpert eye, it looks like Papy wrote their own
T&L code rather than using the OpenGL functions that the GeForce
will accelerate.
When supported, hardware T&L will help a lot when you're CPU-bound and
a lot of those CPU cycles are dedicated to T&L. It won't help much if
fill rate or the drivers are the main bottleneck. OpenGL GPL with mirrors
seems mostly driver-bound on my system (TNT1, Celeron 366 at 539 MHz) up
to 1024x768. Beyond that, the TNT1 starts to run out of fill rate. A
GeForce should help a lot when running at higher resolutions, merely
because of the higher fill rate. But 1152/1280/1600 with laggy mirrors
and reduced frame rates at the start of the race still wouldn't be ideal.
I asked Nick Triantos at Nvidia if he was aware of the GPL / TNT mirror
issue. He said that he'd try to talk to Papyrus and figure out the
problem. No guarantee that we'll see a fix anytime soon, but a good
sign.
--
spamgard(tm): To email me, put "geek" in your Subject line.
Currently most developers do. It gives the better ones a good way of
optimizing their rendering pipeline. There has been quite a bit of
discussion on many of the developer lists about whether to use T&L on
hardware.
Hey, thats a good piece of information.


































