Thanks for the corrections and much better explaination.
Philip Taylor writes:
> answers inline below
> <
> all modern cards, read in the last 3 years, except 3dfx Voodoo1 and Voodoo2
> can do 3D hardware acceleration in a window.
>
> Voodoo1 and Voodoo2 are what are known as "secondary" cards that use a
> pass-thru cable. they do not contain any 2D hardware and thus no VGA/SVGA
> core and cannot render the desktop.
>
> Voodoo1 shipped late '96. there were contemporaneous cards from other
> manufacturers ( like PermII, RIVA128,RagePro ) but at the time 3Dfx was top
> of the heap; fillrate and features.
> >
>
> At home I have a Permedia 8Meg card and at work I have an STB Nvidia
> 16 meg card. Both machines are PII 400's.
>
> <
> PermI or PermII?
>
> if PermI,thats severely dated.
==================
It's the Permedia II...
>
> if its PermII, realize that PermII is a 2+ year old card that has ~75% of
> the fillrate of the Voodoo1 ( which had 50-60mp back in the day) but lacks
> alpha blending modes and so cannot do colored lightmaps ( no
> srcalpha/invsrcalpha blend modes if I recollect correctly ).
>
> for comparison with 2nd half of '99 hardware - nVidia geForce has 500mp
> fillrate.
> >
>
> The 8 meg card at home is much faster than the 16 meg one because it
> can do hardware acceleration in a window. The 16 meg card at work is
> just terribly slow in comparison.
>
> <
> what chipset is in the STB? TNT? its not RIVA128, right? not a RIVAZX?
>
> if thats TNT the result does not make sense, cause a TNT is way better than
> a PermII.
>
> what display mode ( width x height x depth )? what rev of the nVidia driver?
>
>
> are you sure you are getting a hardware rasterizer? and not running on the
> software rasterizer?
>
> do you have the DirectX SDK? if so - what does dxdiag tell you? the caps
> viewer? how do D3D samples run?
>
=====================
It is a TNT card. I was very surprised to see that things were slower
on this card. I was running at 1280x1024 16bit color depth.
The speed varied quite a bit depending on the size of the window.
When the window was full size, rotation was very slow.
I have since switched my work machine to Linux ,those examples were
from WindowsNT, so I really don't know what driver rev I was using.
I have yet to try any J3D apps on Linux. I've only run the static
Mesa/OpenGl book demos.
It seems like there was a good chance that there was something wrong
in my NT machine configuration that prevented me from getting hardware
rendering.
> >
>
> I have not tried the DirectX version with full screen mode turned
> on. I suppose the performance should improve on the gaming cards when
> run in full screen mode.
>
> <
>
> depends.
>
> fullscreen is page-flipped which is usually gated by refresh freq. 60 fps is
> usually sufficient for game framerate though. for stereo 120hz refresh is
> nice.
>
> windowed is blt and blts are typically not gated by refresh freq. so it is
> possible to run faster in windowed mode at the same resolution...
>
> here is a tip to see if you are fillrate-bound or poly-bound:
>
> 1st drop rendering resolution ( fewer bits to fill ). if there is no
> significant speedup, you are not fillrate-bound.
>
> next drop model resolution at the same rendering ( fewer polys to draw ). if
> you dont see a significant speedup, you are not poly-bound.
>
> fillrate is what rasterizer hardware is meant to accelerate.
>
> poly processing typically happens on the CPU, but as consumer hardware that
> accelerates transforms ( like nVidia geForce ) becomes prevalent - that will
> change the cost equation between graphics hardware and CPU.
>
> tools like VTune and IPeak GPT ( both from Intel ) are recommended for
> serious perf optimization work.
> >
>
> <hope this helps>
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
