How much do you increase memory controller voltage for 4 DIMMs vs 2 DIMMs?

[onething]

I've been running my current FM1 socket system with a 2x4gb kit. Recently added another 2x4gb kit, and while memtest passed, there were too many random errors, applications closing, etc. in Windows. I noticed that the memory controller voltage was never increased even though the setting is on auto.

So for anyone else using extra DIMMs, how much did you raise your memory controller voltage to get it stable?

Edit:
I'm starting to think it's not the memory controller, but the timings. I reset all the voltages and set the command rate to 2T instead of 1T.

Solved:
Turns out one of my kits were defective despite passing memtest86. I ran HCI's memtest and it errored out immediately. Swapped it out and errors all gone.
No need to raise voltages or loosen timings.

 

[tweakboy]

well when you put all 4 in , you probably might have to raise the RAM voltage. Also loosten timings, and see if you get errors then. gl For example.. my RAM sys 2.10v but I had it to 2.20v on my mobo that blew up then I got this pos used p5k and I cant do anythings on BIOS with hit crappy.......

 

[onething]

Loosening the command rate seems to fix the issue. Not experiencing anymore glitches. Well, lesson learned.
It seems to be a combination of both. Need to raise memory controller voltage and loosen timings.

 

[Soulkeeper]

When I went from 2x4GB to 4x4GB I didn't need to up voltages or loosen timings.
Perhaps I was lucky (both 2x4 sets are the same make/model).

which motherboard and which memory sticks (are they all the same) are you using ?

Here are some stable timings for me with 4x4GB at 1866 on my fm1 GA-A75M-UD2H
mem voltage is at 1.52v and 1.2125v (+.075v from auto) for the mem controller.
The hynix chips on my mushkin 2133 set are spec at 1.5v +/- .075v in their datasheet.

The command timing (CR, T, CT, etc.) and the write recovery (tWR) are important things you might try relaxing. Also loosening ras to cas (tRCD) by 1 clk might help

 

[onething]

Thanks Soulkeeper.

Anyway, it turns out one of my kits are defective. It didn't make sense that I upped the voltages and problems still persisted. So I needed a 2nd opinion on memtest. It passed every test memtest86, but HCI's memtest errored out right away. I swapped the sticks and boom, no more errors.
No need to raise any voltages or loosen timings whatsoever.

My confidence in memtest86 has dropped significantly.

If you're curious, I'm using Asus F1A75-M Pro R2.0 board with Mushkin 4x4GB 1600mhz 9-9-9-24 1.5v

 

[Soulkeeper]

yeah, memtest isn't very definitive.
I like using it to atleast verify that my system is stable enough to boot the os
it just can't cover all usage scenarios, but I still appreciate it.
then I do real stability testing ie: gaming, compiling code, daily usage

 

[tweakboy]

OUCH! rma that thing good thing we have memtest