Updated: Tuning C'n'Q: Maximize Power And Performance, Part 2

Editor's Note: This story first went live in November of 2009. After its publication, we went back and forth with AMD's power experts, discussing some of the conclusions drawn. With feedback from AMD, we've retested to clarify where optimizations are most likely to be made. We've also addressed a problem with Cool'n'Quiet observed in the previous piece, which turned out to be caused by an implementation of the technology, rather than the technology itself. Special thanks to AMD for its input. You can find Part 1 of this exploration right here.

Earlier this week we discussed the inner-workings of AMD's Cool'n'Quiet technology and what you could do as an enthusiast with the right tools to further-optimize an Athlon or Phenom processor.

Now it's time to talk about performance. Like most of the processor-oriented reviews you read, we're going to use several benchmarks to measure the speed of our various samples. But we're going to go even further this time around. In addition to performance numbers and test results, we're also going to include power consumption measurements for each benchmark.

The purpose is actually quite simple. Different applications utilize processors differently. Some will make full use of all available cores, while many others will not. This also means power consumption with each scenario will likely be different. By measuring the power consumption of single- and multi-threaded applications, we can get a reasonable idea of the performance and power consumption in a handful of scenarios, addressing whether or not our optimizations had a positive effect.

There’s also a different type of scenario we wanted to test. Many benchmarks apply a 100% load to each core, so the processor will be in its highest p-state during the test. We wanted to observe performance with idle p-states, or at least intermediary p-states. After all, the processor is idle through much of time we spend on the desktop, whether you're Web browsing, writing e-mails, playing music, or watching standard definition video--all of which hardly load the processor. For our purposes, we'll be measuring system power consumption when playing high definition video with and without hardware acceleration.

We're also going to look at the core/processor utilization levels. Why? To see whether or not the benchmarks really make use of all available processor resources. We can see just how often the processor is idle and to what degree. We can also see whether or not all cores are fully-utilized. To do this, we record processor utilization by using Vista's built-in Performance Monitor.

These measurements were taken with different power profiles: No Management (we turned off Cool'n'Quiet in the BIOS), Balanced (Cool'n'Quiet is enabled in the BIOS and the Windows power policy is set to Balanced), and Optimized (Cool'n'Quiet is enabled in the BIOS, the Windows power policy is set to Balanced, and the processor is running with lower core and northbridge voltages, shorter p-state transition times, and synchronous clock changes or “Ganged” in K10Stat). These settings are selected because we want to focus on performance and consumption with power management features enabled.

Note: We re-tested the Athlon II X2 250 and Phenom II X4 955 on a new platform, swapping in a Biostar TA790GX A3+ (a dual-power plane, AM3-equipped motherboard) and four modules of Team Elite DDR3 memory. The power supply was also swapped to a lower-wattage unit (Enermax's 405W Tomahawk). We also obtained an Athlon II X4 620 to see how it compares to the Phenom II X4 955 BE.

Create a new thread in the UK Article comments forum about this subject
This thread is closed for comments
Comment from the forums
    Your comment
  • silverblue
    It might be worth re-running the DivX and XviD benchmarks with the 710; specify core affinity with cores 0 and 1.

    Anyway, I thought they were looking at this issue?
  • vexun11
    I love my i7 930