On Fri, Aug 28, 2015 at 4:09 PM, Eric Griffith <egriffith92(a)gmail.com> wrote:
Literally the very last sentence of the documentation explicitly
states that
thermald will control the P-States of the CPU to decrease temperature
without affecting the T-states (clock cycles) thus not harming performance.
I think that's misleading. P-states themselves adjust voltage and
frequency, so performance can be impacted by them alone. T-states
essentially lock the CPU even further through clock gating. So the
documentation seems to be implying that the daemon was able to reduce
temp just through p-state adjustment and not resort to t-states would
would have impacted performance even more. In other words, it was
able to reduce temp while limiting performance loss but not eliminate
performance loss all together.
josh