On 04/30/14 21:56, Jonathan Gray wrote: > On Wed, Apr 30, 2014 at 07:28:16PM +0200, Thomas Bohl wrote: >> Am 30.04.2014 05:23, schrieb Jonathan Gray: >> >On Tue, Apr 29, 2014 at 10:22:29PM +0200, Thomas Bohl wrote: >> >>cpu0 at mainbus0: apid 0 (boot processor) >> >>cpu0: Intel(R) Atom(TM) CPU D2550 @ 1.86GHz, 1867.07 MHz >> >>cpu0: >> >>FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CFLUSH,DS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE,SSE3,DTES64,MWAIT,DS-CPL,TM2,SSSE3,CX16,xTPR,PDCM,MOVBE,NXE,LONG,LAHF,PERF,ITSC >> > >> >We only do speedstep if the processor advertises that speedstep is >> >supported in cpuid (ie there should be a 'EST' flag above). >> > >> >According to >> >http://ark.intel.com/products/65470/Intel-Atom-Processor-D2550-(1M-Cache-1_86-GHz) >> >it doesn't do speedstep as well. >> > >> >i386 fakes a table with high/low values for older processors that >> >still have a fsb, which was mostly used before the code to fetch >> >tables from acpi was added. >> >> Thank you for your explanation. >> i386 it is then. > > It wouldn't hurt to check with md5 -tt and/or a power meter > to see if there is actually a difference between > hw.setperf=0 and hw.setperf=100. >
A power meter would be more "useful" -- at least the first generation of Atom systems, the Northbridge chip drew more power than the CPU (really -- the heatsink and fan was on the Northbridge chip, NOT the CPU!! This may explain the lack of speedstep); if you could wack the CPU down to zero power consumption (you can't), it would hardly have changed the TOTAL system power draw at all. Nick.
