QoS Power Management
Submitted by srlinuxx on Thu, 09/27/2007 - 14:32
"The following patches implement a more generalized infrastructure (than latency.c) for connecting drivers and subsystem's that could implement power performance optimizations with the data needed to implement such policies," began Mark Gross, describing his Quality of Service power management patchset. He added, "these patches are following up on the discussions and presentations at the power management summit last summer." Mark continued:
"The idea is that from an abstract point of view how much to throttle hardware can be expressed as a function of QoS types of parameters; Latency, throughput, and idle time outs.
Theoretical QoS is not practical for continued cpu service ?
In order to continue service of a cpu overheated, cpu is interrupted to cool down by idling. Some has speed step to further the cooling down process.
But in the end, saving the cpu is priority number one. Quality of service is futile unless cpu is saved thru proper cooling cycle, learned by experience of any cooling cycle possible(Microsoft shadowed bios adjustment in auto-update service provided).
Many people had experience of overclocking the cpu; and their experience is always larger cooling fans and optimum heatsink(base thickness not more than 3/16 inch thick) which is fast reacting with large thin fins. Then cpuidle software is used for safety net. Idling period is not based on QoS but cpu thermal runaway prevention by delta temperature limits(85/95 degrees C thermal runaway temperature).
Today, power management is lowering core voltage and speedsteps, switching circuits on/off to balance the current density flowing thru the silicom chip.