During our recent tests of Socket FM2 mainboards we found out that we couldn’t measure their power consumption correctly at default settings. The LinX utility we use puts a very high load on the CPU – much higher than real-life applications do. Meanwhile, AMD’s Socket FM2 processors can be called over-overclocked as the manufacturer specifies unrealistic default settings for them. Such CPUs can’t work at their default settings when the load is high, so they drop their clock rate, which leads to reduced power consumption and performance. Intel CPUs have no such problems but the question is whether we need to use extremely high loads to measure power consumption. The LinX utility is okay as a means to check out an overclocked CPU for stability or to test a CPU cooler, but real-life applications, like those we use to benchmark performance, seem to be a better choice for power consumption measurements. We can have several test modes: idle, single-threaded and multi-threaded load, and gaming. Although we do not overclock our graphics card, it is a rather advanced model whose power draw contributes a lot to the overall power consumption. It turns out, however, that it’s not easy to obtain repeatable results. Everyday applications have a fluctuating CPU load. Moreover, power consumption increases along with temperature. It’s even worse with games because the power draw depends not only on the particular game but also on graphics quality settings and even on what goes on in the game at the given moment.
Our solution is simple. We do not carry out special tests. We measure the power draw of our configuration right when it’s going through our performance benchmarks. It doesn’t matter that the numbers fluctuate and grow up along with temperature. We just note the top values, disregarding peak, uncharacteristic surges. After all, we don’t sit down to our computers for just a few seconds. Many people even spend hours playing games, so our results will be as close as possible to the real-life level of power consumption.
As a result, our new updated methodology looks as follows. We performed our power consumption measurements using the same Extech Power Analyzer 380803. This device is connected before the PSU and measures the power draw of the entire system (without the monitor), including the power loss that occurs in the PSU itself. In the idle mode we start the system up and wait until it stops accessing the hard disk. The power consumption in case of single-threaded CPU load is measured while running the Pi-digits performance test, and in case of multi-threaded load – while running the Fritz Chess Benchmark test. We also used Hitman Absolution game to create complex load. The results on the diagrams are sorted out in ascending order.
As we have expected, the mainboards from ASRock and ASUS are close in their power consumption since both are equipped with a PLX PEX 8747 chip. They differ much from the ordinary MSI mainboard without that bridge chip. The MSI is 30 watts more economical in every test mode, which is about one third of the total consumption at zero and low loads.
The picture doesn’t change at overclocking. The ASRock is slightly more economical than the ASUS, but both need much more power than ordinary mainboards such as the MSI model.