Should increase RAM usage and map load time a bit, but reduce sound decoding
CPU usage (Vorbis decoding) substantially. Expecting this to be a net positive
tradeoff, and should fix issues like #2500 in a much safer way, with no quality
lposs whatsoever.
set snd_soundradius 1200
set snd_softclip 1
set snd_maxchannelvolume 0
-set snd_streaming_length 2
+// Only streaming-decode music, not any sound effects. RAM is cheap nowadays.
+// Sadly, our longest sound effect is 36 seconds.
+set snd_streaming_length 40
seta menu_snd_sliderscale 2 "0: decibels; 1: linear percent; 2: 0..10 scale; 3: slider size percent"
seta menu_snd_attenuation_method 1 "Use exponential instead of linear falloff for sound attenuation"
alias snd_attenuation_method_0 "set menu_snd_attenuation_method 0; set snd_soundradius 1200; set snd_attenuation_exponent 1; set snd_attenuation_decibel 0" // Quake default