We usually benchmark mainboards in two test modes: at the default settings and with the CPU and memory overclocked. Since we had got new memory modules, a new CPU and a new platform, we were curious to see what maximum performance we can expect. That’s why we got down to benchmarking in the overclocked mode first. In fact, we always do it that way, but publish the default mode results first. All mainboards are similar at their default settings, delivering comparable performance and but rarely showing any problems. It is only in the overclocked mode that we can usually see any difference and, occasionally, problems whereas different overclocking results lead to different benchmarking results. Of course, we compared the performance of the Gigabyte GA-Z87X-D3H with that of the ASUS Z87-K. The two mainboards overclocked the CPU to the same level, 4.5 GHz, so we didn’t expect any difference in speed. The memory modules were set up using their X.M.P. data, i.e. 2133 MHz, 1.6 volts, and timings of 9-11-11-31-2N. The results are sorted in the descending performance order.
We used Cinebench 11.5. All tests were run five times and the average result of the five runs was taken for the performance charts.
We have been using Fritz Chess Benchmark utility for a long time already and it proved very illustrative. It generated repeated results, the performance in it is scales perfectly depending on the number of involved computational threads.
x264 FHD Benchmark v1.0.1 (64 bit) allows us to compare the system performance during video transcoding against the results in the database. So far this is the only utility in our testing suite that supports AVX2 instructions implemented in Haswell processors. The average results of the five test runs are displayed on the following diagram:
We measured the performance in Adobe Photoshop CS6 using our own benchmark made from Retouch Artists Photoshop Speed Test that has been creatively modified. It includes typical editing of four 24-megapixel images from a digital photo camera.
In the archiving test a 1 GB file is compressed using WinRAR archiving tool with maximum compression level.
The recently released PCMark 8 suite allows us to assess the overall system performance, the efficiency of its storage sub-system, measure the battery life of a mobile device. We use the “Home” test package that includes typical home use applications: web browsing, creating and editing documents, simple games, photo editing, video chatting.
The next diagram shows only CPU tests from the 3DMark Fire Strike. This score is obtained in a special physics test that emulates the behavior of a complex gaming system working with numerous objects:
The benchmark integrated into the Hitman Absolution game turned out very convenient. It can be launched directly from the game, from the game launcher utility and even from the command prompt. We use maximum quality settings from the “Ultra” profile and a pretty high screen resolution.
Batman: Arkham City game also reacts eagerly to any changes in the CPU clock frequency, and it uses DirectX 11. We ran game’s integrated benchmark five times with “Very High” image quality settings and then take the average of the five runs for the diagrams.
Similar mainboards are prone to deliver similar performance under the same conditions. Indeed, even though the Gigabyte GA-Z87X-D3H is always ahead, the gap is no larger than 1% even. The only exception is the WinRAR test where the ASUS is slower by more than 5%. Trying to figure out the reason for that, we launched the CPU-Z utility on the ASUS Z87-K. The outcome was most surprising, so we even recorded a small 14-second video to show you. Take note of the CPU multiplier or frequency. The computer is idle for the first several seconds, so the multiplier is reduced to x8 and the frequency is 800 MHz. Then we launch the archiving tool, the frequency grows up to 4.5 GHz as expected but then drops to 3.9 GHz. It then goes up and down again in a cycle.
This was completely different from how the CPU had behaved on the ASUS Z87-K in the default or overclocked mode before. We had observed the CPU frequency multiplier drop like that but only for a fraction of a second. Such short-term frequency drops had but a small effect on performance, but here the multiplier is regularly down, so the computer works quite a lot of time at a CPU frequency that’s lower than specified. That’s why the ASUS mainboard is slower in this test. We must note that it is not some special heavy-load benchmarking utility but a popular and not very resource-consuming archiver WinRAR.