I ran similar tests which you can see on this list in an e-mail entitled
"performance analysis". I actually found triangles rendered faster than
quads in all scenarios. I am using a GeForce II. I think the performance
is due to drivers and card. I have never seen that kind of wierd cutoff at
11000 though. I am pretty sure I have never built a single geometry array
that large.
Dave Yazel
-----Original Message-----
From: Artur Biesiadowski [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, April 03, 2001 6:00 AM
To: [EMAIL PROTECTED]
Subject: [JAVA3D] Quad/Triangle performance strangeness
I'm working on particle system. After experimenting with immediate mode
(which turned out to be about 4x slower) I've decided to use single
geometry with byref coordinates array for all particles of single type.
Now I have to decide if I want to use QuadArray or TriangleArray (with
two triangles creating a particle). QuadArray is more intuitive and
updating it is a bit faster, but I wanted to check the hard numbers. Fps
below are for static case, without updating coordinates of particles -
just plain throughput.
Particles TriangleFPS QuadFPS
4800 58-104 69-84
8000 41-70 48-82
11200 32-33 4
Ok, quads seem to perform better - after all there is less data to send
to graphics card. But what happens at high number of particles ???? With
10800 particles quads still have 30+fps. Just over 11000 performance
drops to 4-3 fps. What is a problem ? Have I hit some cache overflow, gc
treshold array size or anything like this ?
I use Java3d 1.2.1/directx, jdk 1.3.1 beta, win 2000 and geforce 256.
Under opengl I do not see this behaviour - framerate constantly drops
with increasing number of particles and quads are always a bit faster.
Artur
===========================================================================
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".
===========================================================================
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".
