HwBot.org, one of the world's biggest overclocking and benchmarking sites has banned the use of Windows 8 for posting benchmarking results. With all existing results being disqualified, due to severe validity problems with the Windows 8 real time clock (RTC). The RTC in a PC is designed to keep accurate time and often cases is housed in the Southbridge, other times it is a separate item. This is just one of the reasons whythere is a battery on most motherboards.
This RTC is useful to the majority of benchmarking programs due to its inability to be easily meddled with, due to it being a hardware part rather than software. Because of this bencmarking programs such as 3dMark utilise the RTC to correctly determine when the benchmark begins and ends, the accuracy of the score being heavily dependent on the accuracy of the RTC.
The reason why Windows 8 RTC isn't accurate, and the reason it has been banned as an OS that you can submit scores with, lies with the way that Windows 8 has become a jack of all trades operating system, needing support for a wider array of devices and subsequent different configurations of hardware, some of which do not have the built in RTC hardware. Or as HWBot wrote "Microsoft made changes to how it measures time to be compatible with embedded or low cost PCs that do not have a fixed RTC clock. After all, having a fixed RTC clock adds cost to a platform."
HWBot have already faced similar issues in the past with issues being noticed in a modified version of the Heaven DX11 benchmark that HWBot ran several years ago, although this was resolved using a patch that checked against a second source to verify the benchmark duration before calculating the score. Due to the problem occurring across such a wide array of benchmarking tools however it is not known if a simple solution will be available.
The problem occurrs when down clocking the system under Windows 8, the RTC is affected as well, by down clocking you are effectively slowing the machines RTC down. This then makes the benchmarking programs run for longer, (with a measured 18 seconds difference over the course of 5 minutes), because of this the benchmarking programs and your system can then render more frames in what the programs assume is the same time frame thus improving the score.
HWBot show list several examples some of which are shown below, all of which however show a marked improvement when the system has been down clocked.
Let us make this more practical. On our Haswell test system we downclocked the BCLK frequency by about 6% from 130 MHz to 122MHz. Using a CPU ratio of respectively 32x and 34x, the resulting CPU frequency remains 4160MHz. Then we ran comparison benchmarks. Here are a couple examples:
Result at default
Result after underclock
3DMark Ice Storm
3DMark Cloud Gate
Bus Speed of 130MHz = 7min33.862
Bus Speed of 122MHz = 7min10.799 secs