Benchmark Results: Efficiency Analysis
Our efficiency run consists of several applications that we cycled through a batch script in the following order:
- 3ds Max
- DivX
- Xvid
- Lame
- MainConcept
- PDF Creation
- Photoshop
- AVG Anti-Virus
- WinRAR
- WinZip
Some of these applications don’t take advantage of multiple processing cores; others do heavily. The batch file creates time stamps at the beginning of the run and when it finishes. This way we track the total time required for the test run to complete.
In the meantime, we also track power consumption in one-second intervals during the test run. This allows us to shed light on power consumption for each application, and to create power consumption profiles for each system as we did before using PCMark Vantage. Compared to our old run, this workload is based on real world applications, and it actually shows performance benefits on the efficiency diagram.
The result is runtime performance score and power consumption readings that allow us to analyze efficiency. Idle power is taken out of the equation, as the differences are rather small.
The average power required to complete our entire efficiency workload does not differ much between the 133 MHz and 160 MHz base clocks (2.66 to 3.20 GHz). However, power consumption starts to increase considerably around 3.8 GHz. From this point on, we were forced to work without the Turbo Boost mode.
Of course, we also tracked the total power in watt-hours required for the workload. It turns out that the highest overclock with Turbo Boost enabled required the least power to complete the test! The static overclocks don’t look as promising.
Finally, we need to look at the runtime, which equals the performance of each test configuration. Clearly, the statically overclocked systems were fastest in the order of their clock speeds. However, the dynamic overclock on the Core i5 with Turbo Boost and 3.84 GHz peak clock speed was exactly as fast as the static overclock at 3.6 GHz. Recall the power consumption numbers and you’ll immediately see who should win the efficiency battle.
This is the real deal. Systems running high overclocks within the limits of the Turbo Boost feature are most efficient in the case of our workload that consists of multiple applications. Overclocking a Core i5-750 from 133 MHz to 160 MHz BCLK is safe on the majority of all upper-end LGA 1156 motherboards, it will have the processor reach up to 3.84 GHz peak clock speed if one or two cores are stressed or 3.36 GHz for three or four cores, and the overall power consumption stays at very acceptable levels. High static overclocks only make sense if you continuously run demanding applications that take advantage of the generated processing power. Otherwise, the power required to keep all four cores running at high speeds is simply wasted.
The diagram illustrates runtime and power consumption for the workload we described at the top of this page.
