0
0

Some of our designers put together some performance calculations, I thought you might find interesting. It presents the case of why HW mode is appreciated on the PC platform. CPU numbers as reported by Fmod

Level One
Banks Loaded: Enemy1.fsb
Peak Channels 45+

Force SW —- DSP —- Peak CPU —- Avg CPU
Y Y 16% 7-10%
N Y 11% 5-6%
Y N 7% 3-4%
N N 1% 0.5-0.6%

Level Two
Banks Loaded: ENV1.fsb, ENV2.fsb, ENV3.fsb
Peak Voices 32+
Force SW —- DSP —- Peak CPU
Y Y 9%
N Y 4%
Y N 5%
N N 1.9%

Best regards,
Michael Marks

  • You must to post comments
0
0

Curious as to what DSP effects you were running in software vs. hardware mode?

  • You must to post comments
0
0

Our DSP effects were Low Pass, High Pass, and Pitch.

Not every event in the project used a DSP filter, so some events could be run in hardware mode. That explains the different between 16% and 11%, which is similar to the 7% and 1% which is the penality for switching sounds from HW to SW.

The different between 16% and 7% is the penality for using LP, HP, and the Pitch DSP effects.

Best regards,
Michael Marks

  • You must to post comments
0
0

Use ITLOWPASS, not lowpass for performance. I have also doubled speed of ITLOWPASS in the next release, 4.04.27. (same with parameq, though not highpass).

You are not seeing directsound driver cpu usage, which is usually quite high, so the hardware numbers you are quoting are probably not accurate.

edit: pitch filter! Ok that is a pretty expensive effect and not normally used in games :) It is FFT / iFFT based. We have a lot of room for optimization on this one though.

  • You must to post comments
Showing 3 results
Your Answer

Please first to submit.