This was weird (to me): I’m testing an MSI Z590/i7-11700K build which I have been having issues with (not enough power from onboard HDMI port to successfully drive a 4K display it would seem).
I swapped the board and the processor trying to track down the problem.
Now that I have the 2nd board and processor operational with Windows 10 I was just doing some benchmarking. I had installed 1 16GB stick of memory in DIMMA2 on the motherboard, as per documentation, to keep things simple while trying to identify my problem (the one with the build…).
The AIDA64 memory read/write/copy throughput was half what it was when I tested the first version of the motherboard and processor with 2 sticks of RAM installed (DIMMA2 and DIMMB2).
Putting the other stick of RAM back in couldn’t cause the read/write/copy benchmark scores to double could it?
Oh yes it could.
I presume that since it’s a dual channel memory system (like virtually all consumer level motherboards and processors) that this is expected behavior: the IMC can’t double up on or pipeline memory transfers when there’s only one channel with any memory in it, even if it is dual rank. Since I’ve not tried running a system with a single DIMM installed before (at least if I have I can’t remember it), this rather surprised me.
The memory is G.Skill RipJaws F4-3200C16D-32GVK DDR4-3200 CL16-18-18-38 Dual Rank.
I’m using the trial version of AIDA64 Extreme.
Can anyone confirm this is totally expected behavior?