[Cache & Memory Benchmark - different results]

I have noticed something else, when running Aida64 on WinPE (better results), the northbridge clock is correctly detected as 47x. However, on the main Windows 10 installation (slower results), the northbridge clock is detected as 36x. Could this influence results?

 

 

[Cache & Memory Benchmark - different results]

I am encountering this very strange bug. If I use Aida64 Cache&Memory Benchmark in Windows 10 I get a specific Ram benchmark result, let's say 48100 MB/s read / 53600 MB/s write /  47700 MB/s copy and 52 ns latency. However, if I am booting from an USB live Windows 10 CD (WinPE), I am getting drastically better results, such as 55000 MB/s read / 61300 MB/s write / 55000 MB/s copy / 45 ns latency.

I am not changing any bios / RAM XMP / Ram memory timings settings between reboots, only booting from either main Win 10 installation or WinPE 10 live usb.  I am using the same Aida64 version in both cases.

I am scratching my head, what can explain the huge difference between the results? In both cases, cpu usage is idle and no other background programs are running when doing the tests.