Power Consumption Results

System Idle/Max Power Consumption

Let's first look at the idle and maximum system power requirements.

The AMD platform clearly requires more power when idle (102 W vs. 87 W equals 17% more) and the power requirement difference is even larger when under load, at 210 W vs. 163.5 W, or a 28% difference. Remember that the results cover all other system components, which also includes the motherboards' voltage regulators.

3DMark06 And PCMark05 Loops

In this first test we created a batch file to run the 3DMark06 with its default settings, and the system, CPU and memory tests of PCMark05. Since both terminate within minutes, we decided to loop the whole enchilada four times.

This chart shows the average power requirement while 3DMark and PCMark were looped. Clearly, the AMD solution required more energy.

We calculated the total power consumption for the duration of the 3DMark06/PCMark05 test: 339 Wh for the Athlon 64 X2 5000+ is 25% more energy than the 269 W of the Intel system. The question now is: which one took longer?

Click to enlarge

The diagram portrays the power consumption of the systems at the plug during the four repetitions of 3DMark06 (first) and PCMark05. You can observe that the spikes are repeated four times, which corresponds to the benchmark runs. You'll also notice that the Intel system (blue) took 2:15 to complete this run, while the AMD system required almost 2:20, while having an average power consumption that was ~30 W higher than that of the Intel system.

Conclusion

Our two power consumption tests, which we created to track the actual system power consumption over time, are rather simple, and don't necessarily reflect application scenarios, or user habits over a period of time longer than 90 minutes. We did not install anti-virus software, messengers, download clients or other software that could increase the system power consumption in idle and when under load. All we did is to track the power consumption for the entire duration of each test run and calculate the average power requirements (in watts) and actual power required (in Wh).

The results speak clearly, showing that performance is still a major issue, even when the primary goal is to save power in average desktop environments. The results leave us with serious doubts about the strategy of some processor makers that release products with the single focus on energy efficiency. These may be great for specific applications, but certainly not for most desktop PCs.

Leading people into buying a low-power, high efficiency processor for desktop use is misleading, because the user could end up in a scenario in which her or his low-power components will have to run at substantial loads all the time. In such a case, all the power savings could be gone, and an average system might turn out to be more efficient.

Here are the facts from our test results:

• Power consumption must not only be looked from a minimum and maximum power consumption standpoint, but must be tracked using applications and benchmarks that truly simulate PC use over time. Performance and energy efficiency are closely related, and you cannot have ideal energy efficiency without a certain level of performance.
• The Athlon 64 X2 system we used consumed more energy than the Core 2 Duo E6400 machine, whether it was idle, running our power consumption benchmarks, or under maximum load. I want to make clear that faster Athlon processors would look better, but they would also require more power. The same applies for faster Core 2 processors.
• The Core 2 Duo E6400 system completed the SYSmark 2007 Preview run 14 minutes earlier than the Athlon 64 X2 5000+. As a consequence, the Intel system went back to an idle state earlier, which of course results in considerably less total power consumed. During the same time, the AMD system had to stay at a higher activity level for a longer period of time, which eventually meant that it required as much as 50% more power than the Intel system!

Note: All results include the system components (see test setup table). Power supply and voltage regulator efficiency play an important role in total power consumption.

If you are looking for a system that is energy efficient, it is not enough to pick a processor and system components that falls into this category. Performance is at least equally important, because energy efficiency dictates whether a processor can go into an efficient mode often, or if it has to remain at high loads to deal with the workloads you throw at it. The best choice for an energy-efficient, though fast processor seems to be in the reasonable mainstream segment.

On the one hand, a Core 2 Quad has a much higher idle power and does not always benefit from its four cores. A 45 watt AMD EE processor, on the other hand, might not be able to save power, if it has to work at high loads oftentimes. Looking at the test results, we clearly recommend sticking to a mainstream processor (65 W thermal design point) of the latest architecture and generation, which, once again, is the Core 2 Duo. Now it's really about time for AMD to get its Phenom processor family ready, because the efficiency advantage simply isn't there.