| |||||||||||||
LCA: Server power managementLCA: Server power managementPosted Jan 28, 2011 16:29 UTC (Fri) by robert_s (subscriber, #42402)Parent article: LCA: Server power management
"Another interesting feature of contemporary CPUs is the "turbo" mode, which can allow a CPU to run in an overclocked mode for a period of time. Using this mode can get work done faster, allowing longer sleeps and better power behavior, but it depends on good power management if it is to work at all."
From all I could find on the web about Turbo mode, the policy all seems to be done "automatically" by the hardware. If true, this seems like the worst design decision ever. At best, surely the kernel would be able to do a much better job with the extra context it has about the workload. At worst, doesn't this cause absolute havoc with the kernel's scheduler?
(Log in to post comments)
LCA: Server power management Posted Feb 1, 2011 21:22 UTC (Tue) by martinfick (subscriber, #4455) [Link]
No, it usually involves OS support.
http://en.wikipedia.org/wiki/Intel_Turbo_Boost
"Turbo Boost activates when the operating system requests the highest performance state of the processor. "
LCA: Server power management Posted Feb 1, 2011 22:41 UTC (Tue) by mjg59 (subscriber, #23239) [Link]
Right, but that just appears as another P state to the OS - it's entirely possible that the BIOS will have programmed that before the OS asserts any control over the functionality. The problem with turbo mode is that without knowing the full state of the processor (the C state of every thread, the temperature of every sensor, the current power draw of every core) and without knowing the algorithms used (which the vendors consider secret), the OS has no way of knowing whether asking for the highest performance state will give it performance identical to the next highest state, or anywhere up to 60% higher.
|
|||||||||||||