Our Haswell system
For this workshop, we used an Intel Haswell system. We initially combined our Intel Core i7 4770K with a number of standard components. We picked the MSI Z97S SLI Plus as our motherboard. In addition, we equipped our system with 8 GB of DDR3-1600 1.5V memory, an old Seagate 7200.10 500 GB hard disk, and a Corsair CX750M 80 Plus Bronze power supply. At the start of the test, we reset the BIOS to factory settings. All tests were performed with Windows 8.1 as the computer's operating system.
First, we attempted to reduce the system's power consumption by replacing various components with more energy efficient alternatives. After this, things got serious: we asked our resident overclocker Joost 'Rsnubje' Verhelst to do something that is the complete opposite of what he normally does: to make various BIOS adjustments that, rather than making the system as fast as possible, make it as energy efficient as possible.
After each adjustment, we ran a pair of benchmarks: Cinebench 11.5 and 3DMark Vantage. Cinebench only stresses the CPU, allowing us to determine whether our adjustments influence CPU performance, and if so, how large this influence is. While 3DMark Vantage might not exactly be the youngest member of the 3DMark family, it does exhibit good scaling with both CPU and GPU performance, which is why we opted to include this benchmark as well.
Along with the two benchmarks, we took three different measurements of the PC's power consumptions: the average power consumption when idle, the maximum power consumption during Cinebench 11.5, and the average power consumption during 3DMark Vantage. Power consumption was measured using a calibrated Emu 1.x4 power meter. Just to be clear, we measured power consumption at the wall, so the numbers in this article reflect the power consumption of the entire system.
Our system consists of a Core i7 4770K, an MSI Z97 motherboard, an 80Plus Bronze power supply, an HDD, and some run-of-the-mill DDR3 memory.
When using the factory settings, our configuration scores 8.61 points in Cinebench 11.5 and 6575 points in 3DMark Vantage. According to our measurements, the system's idle power consumption when using factory settings is a little over 42W. The meter goes up all the way to 131W during Cinebench 11.5, and the system uses a little below 80W on average during 3DMark Vantage.
Before we began swapping components, we opened up the Windows control panel and changed the energy profile from the default setting (balanced) to power saver. This forces Windows to make a large number of changes to the way in which hardware is being driven. Important differences include limiting the CPU's Turbo-mode, reducing the clock frequency of the GPU, and earlier suspension of components such as the hard drive. Changing to Windows' power saver profile turns out to have a pretty negative impact on the computer's performance. For example, our Cinebench score drops by over 7% to 7.98 points, whereas our 3DMark Vantage score is reduced by over 50% (!) to 3202 points. This is offset by the fact that our PC is about 3.5% more energy efficient when idle, maximum power consumption during Cinebench is 1.5% lower, and average power consumption during 3DMark Vantage is some 26% lower.
Given that the reduction in power consumption clearly isn't worth the massive hit in performance, we switched our Windows energy profile back to balanced, and left it that way for the remainder of our tests. We don't mind a small performance hit, but we shouldn't go overboard.
Configuring Windows to use the power saver profile results in a (slight) reduction in power consumption, but also substantially lowers the system's performance.