Copyright by William Lloyd Bircher 2010 - The Laboratory for ...
Copyright by William Lloyd Bircher 2010 - The Laboratory for ...
Copyright by William Lloyd Bircher 2010 - The Laboratory for ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
5.3.4 CPU<br />
To test the extensibility of per<strong>for</strong>mance counter power modeling across processor<br />
architectures, the methodology is applied to an AMD Phenom quad-core processor. Like<br />
the Intel Pentium 4 processor used in sections 3.2 and 3.3, fetched instructions is a<br />
dominant metric <strong>for</strong> power accounting. Differences in architecture and microarchitecture<br />
dictate that additional metrics are needed to attain high accuracy. Two areas are<br />
prominent: floating point instruction power and architectural power management. Unlike<br />
the Intel server processor which exhibits nearly statistically uni<strong>for</strong>m power consumption<br />
across workloads of similar fetch rate, the AMD desktop processor consumes up to 30%<br />
more power <strong>for</strong> workloads with large proportions of floating point instructions.<br />
To account <strong>for</strong> the difference in floating point instruction power, the desktop processor<br />
model employs an additional metric <strong>for</strong> retired floating point instructions. A still larger<br />
power difference is caused <strong>by</strong> the addition of architectural power management on the<br />
desktop processor. <strong>The</strong> older, server processor only has architectural power management<br />
in the <strong>for</strong>m of clock gating when the halt instruction is issued <strong>by</strong> the operating system.<br />
<strong>The</strong> newer, desktop processor adds architectural, DVFS. This leads to drastic reductions<br />
in switching and leakage power. To account <strong>for</strong> these power reductions, the desktop<br />
model includes tracking of processor frequency, voltage and temperature. <strong>The</strong> details of<br />
the model can be found in Section 3.4.5.<br />
93