Memory calibration and stability

[Hari_Seldon]

This is both a question and something I noticed regarding memory speeds.

I recently upgraded the BIOS for my Z790-F WiFi from 2703 to 2801.  After duplicating my settings, I ran into instability (i.e., general protection faults) running my memory at 4800 (non-XMP).  I performed a lot of tests, and finally noticed my mistake after slowing the speed of the memory to 4200.

While 4200 was a stable speed, I was disappointed that I could not run at 4800 any longer, which is the JEDEC rated speed for my memory kit (KF556C40BWK4-128, QVL approved).  It turns out that my computer had gotten stuck during calibration, so I had to reboot it.  It took awhile to reboot, so I assumed it calibrated correctly.  It did not.  There must be some hidden values regarding the memory because I fixed the stability by setting the memory down to 4200, recalibrated, set to 4800 and recalibrated.  Since the system took a long time yet did not get stuck, it has resulted in a very stable system.

I hope this helps others who made the same mistake.

Plus, it makes me curious about what is actually recorded during a calibration.  The settings dumped in a text file do not show a difference between the unstable 4800 and stable 4800, yet something changed.  Any ideas?  This is with me *only* adjusting the speed of the RAM.  I am not using XMP and am sticking to what the manufacturer claims to be JEDEC settings.

[Vynra]

um, sir, if you have to lower speeds BELOW JEDEC then are you sure you don't have a bad ram stick?

Your ram should be around 5600 with xmp. 

also you dont even say what cpu you have. which can make a difference as if its a 13th or 14th gen i9 it should be fine. if its lower than a i9 your memory controller may be an issue.

However what is super telling is the fact you have to go below 4800. to me this screams that a stick is bad.

[Hari_Seldon]

Good point!  I did forget the CPU:  i7-14700K.

While the RAM should be around 5600, it happens to be four sticks which notoriously has trouble running at higher speeds than JEDEC stably.  It is a four-stick kit, so I did not mix-and-match.

The odd part is that the system is only unstable at the JEDEC speed of 4800 if the system hangs during calibration after changing the speed.  All the tests I run fail easily and reboots and powering down the system do not solve it.  If it does not hang, then I can run a plethora of tests and reboot and power down the system many times without any issues.  Hence, I think there is something being recorded incorrectly when it hangs.  Changing values other than memory settings in the BIOS does not result in instability (e.g., SR-IOV).

I do use memtest but only to rule out poor memory or memory settings.  Running concurrent software builds several times is the best indicator, for me.

Also note an older post of mine where the BIOS chooses a rather low value for MC Voltage when setting XMP I at 5600:  https://rog-forum.asus.com/t5/intel-700-600-series/how-to-tweak-memory-controller-voltage/td-p/10506...  That is why I went with stable JEDEC values.