I’ve noticed really better performance of 3d apps (in my case Maya 4.5 and Houdini 5.5) if I compile from the source the Xfree 4.3 and the Window Manager (for 3d application it’s better you use a low cholesterol window manager like windowmaker of xfce). Before compiling you’ve to script the make.conf file (take a look at the FreeBSD docs if you don’t know how to do it) ,in which
you configure your Cpu type (p3,p4 or k6,k7..) and all the flags you need. This is very important because you can bypass the Slack build script having the cpu=i386 flags.
how does “better” reflect on X? How much faster/what works faster (ie. wireframe/shaded/artisan)?
I 've done a test on 1k rendering of a 15 Mb scene. The machine with compiled Xfree was 15-20% faster than the same one with prebuilted Xfree.
IBM-dual 3.0 Xeon-4Gb Ram-Ti4600.
I’ll try to find out some better benchmark (maybe dx or ..)
Bye
Thats a really nice increase. I know what I’m doing tonight 
.
Ill download the src rpm files of all my xfree86 files so I can compile them easier and faster. Ill report on how much faster it is under RH9.
xfree and window managers has nothing to do with increasing a speed of rendering, AFAIK.
it could accelrate the viewport renders..
not expecting that big of a gain though
since that comes mostly from closed source GLX drivers..
perhaps your still running OSS drivers that ship with X?
if so then install nvidia’s driver and you’ll notice about a 10fold increase in performance.
Nyone else tried this?
That’s not true–I’d suggest you to study all Xfree docs. Anyway we’ve noticed no real gains with same test rendering on machines with single Cpu (don’t know why). I can tell you that the gain on performance on the dual-cpu box was visible even on other operations (like the speed of opening of heavy-weght scene file, the velocity of the hypershade…).
The big problem was the scripting of the make.conf file (BIG thanks to Michael and Ivan -NET-BSD hackers)-.
Bye,
Sorry for my English.