OK, got some benchmark results.
Duron 900 @ 1050MHz (140MHz FSB)
Abit KT7A (v1.3)
256MB Crucial PC133 RAM
Geforce DDR @ stock speed
SB Live! with latest SB driver (not Windows Update driver) dated 17th July 2002
Patched UT2003 Demo results:
min / average / max fps
14.246051 / 28.435907 / 58.733757 fps
Score = 28.455732
16.624786 / 40.320992 / 74.216156 fps
Score = 40.340420
10.090744 / 20.063120 / 40.749531 fps
Score = 20.071943
12.460136 / 28.943920 / 47.080616 fps
Score = 28.961233
Hardware 3D + EAX:
10.534564 / 20.067602 / 42.090870 fps
Score = 20.078844
13.097428 / 29.057392 / 49.269928 fps
Score = 29.074318
14.321988 / 29.437492 / 60.295918 fps
Score = 29.453682
18.594841 / 41.385803 / 75.453117 fps
Score = 41.408051
It would be good if someone with an Audigy card did the same benchmarks as this, as the Hardware 3D + EAX option should be much better suited to Audigy cards (ie. EAX HD) according to the demo readme information.
In terms of sound quality Hardware 3D is a little bit better than Software 3D, but even on my Live! the Hardware 3D + EAX option was in a different league regarding the quality of sound placement. It wasn't so much the quality of the actual sound playback, it was the audio positioning that was significantly better (but that is to be expected of course). Unfortunately the same option was also in a different league with performance, ie. the 'little league' along with the normal Hardware 3D option.
A real surprise turned out to be Safe Mode. In terms of sound placement and quality of the sounds as they are played back, it is a little lacking, but overall isn't too shabby. The positioning is pretty much basic stereo and the quality of the sound samples sounds a little rougher, but the gap compared to Software 3D isn't as big as it is moving from Software/Hardware 3D to Hardware 3D+EAX. As far as I can figure this must be similar to the sound playback in UT anyway without the Hardware 3D acceleration option switched on.
I suppose that if your trying to play on a computer with nothing special sound card, speakers or a laptop, it isn't a bad option to use to get that little extra speed as you may not miss the extra that Software 3D gives you.
Edit: oh yeah, the reason I picked those specific benchmark maps is because they were both a good contrast. The DM map had a lot of action going on and made a lot of use of positional audio whereas the BR (bombing run) map timedemo had less on screen action and hence less mixing of sounds from different positions.