On 08/05/2012 10:04 PM, Ralf Mardorf wrote:
I really wonder why you are all raving about this.
Well, it is hard to generalize but on most modern CPUs + main-boards CPU
frequency scaling is _not_ correlated to audio-dropouts [anymore] and a
The time it takes to change CPU frequencies is very small compared (us)
to [even] low-latency audio (ms). Concerning Hardware, these days,
bus-power-management is most often the cause of xruns on idle systems
followed by SMI.
Disable PCI and/or PCIe power-management in the BIOS and also disable
EIST and C1E halt-states; and, the 'ondemand' governor will works just fine.
If you have an unlimited supply of power and noise of cooling
the system is of no concern: sure, use the 'performance' CPU-freq
governor -- reducing the number of possibilities in complex systems
usually increases reliability... which is indeed a good thing for audio.
NB. frequency scaling _can_ be an issue when using jack2 (or tschak) on
a multi-core machine: The total system-load (over all CPUs) may still be
too low for the CPU governor to react, while DSP load is already at the
limit. -- http://rg42.org/oss/jackfreqd/
Linux-audio-user mailing list