VOGONS


First post, by clueless1

User metadata
Rank l33t
Rank
l33t

I noticed on my slower DOS PC that Duke Nukem 3D took a noticeable hit in framerate when bumping the sound quality to 44khz. Duke allows you to adjust between 44, 22, 16, 11, and 8 khz. I tested 3 ISA sound cards: Soundblaster 16 Vibra (CT2800), Audician 32 Plus, and an Aztech SBPro-compatible. I was curious if any of the sound cards were more efficient.

Test system:
Packard Bell Multimedia C110
Pentium 120 underclocked to P75 (50x1.5)
32MB EDO
onboard PCI Cirrus Logic GD5430 1MB

Software:
Duke Nukem 3D in 320x200 VGA resolution

Showing framerate:
type 'dnrate' in-game for frame counter to show in upper left corner

Method:
Use Duke's setup.exe to toggle sound settings--
1) for each sound card, start with no sound, no music
2) start E1M1, type 'dnrate', wait for rocket to crash and observe framerate when it settles down. If it jumps between two numbers, split the difference.
3) exit, run setup.exe and enable sound card for both digital effects and music (OPL). Set to 44khz, test, then drop to 22khz and 11khz.

Results:
Audician-
no sound: 31.5 fps
44khz: 26.5 fps
22khz: 26.5 fps
11khz: 28.5 fps

Aztech-
no sound: 31.5 fps
44khz: 26.5 fps
22khz: 26.5 fps
11khz: 28.5 fps

SB16-
no sound: 31.5 fps
44khz: 23.5 fps
22khz: 26.5 fps
11khz: 28.5 fps

Takeways:
-The Audician and Aztech are SB Pro-compatible, but the Audician at least declares support of 44khz on the box. But the results don't show any performance hit at 44khz compared to 22khz.
-The SB16 does show a hit at 44khz vs 22khz.

Notes:
-I did try setting both the Aztech and Audician as a SB16 in Duke's setup program, as well as SBPro (new). Both ran normally, and both gave the same results.
-on my faster DOS system (Celeron 333), there is no performance hit for the various sound qualities vs with sound disabled. It appears that once a certain CPU speed is attained, the system is able to absorb any performance impact.
-if I disabled OPL Music, the performance hit was still there, but less. If I used a wavetable MIDI instead of OPL music, the performance hit was the same as when I disabled OPL music, so wavetable music doesn't seem to have any performance impact.
-It appears that as the system gets slower, the performance impact of high sound qualities will increase.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 1 of 1, by realnc

User metadata
Rank Oldbie
Rank
Oldbie

None of these cards do mixing in hardware. And I don't know if Duke3D even supports cards with hardware mixing (the GUS does hardware mixing, but only if the game uses it in that mode.)

So mixing is done in software. Higher sample rate == more CPU time for mixing. But only if the sound data actually is 44.1khz. Duke uses 22.05khz audio data, right? Even when outputting 44.1, I think the samples are 22khz and that's what's mixed.

The SB16 44.1kHz result is puzzling though. At least to me. I'm sure some gurus in here have an explanation. 😀