PRO B660M-A WIFI - Finicky with RAM or Broken DIMM Slot?

Nagorg

New member
Joined
Apr 15, 2024
Messages
9
I'm trying to get my system up to 128GB of RAM. I've got 4 sticks of Crucial DDR5-4800 32GB (CT32G48C4OU5) and the most I can get the thing to POST with is 96GB.
Every time I populate the DIMMA1 slot, the thing wont POST. Instead, the EZ Debug lights come on. Starts with a red LED and then switches to a yellow LED.

Yes, I have another pair of 32G modules (same make/model) that I've tried in the DIMMA1 slot and the thing just wont POST. I've re-seated the modules many times to no avail. I even updated to the latest BIOS, 7D43vAK, but the problem persists.

Any thoughts about what the problem might be?
 
First, try one module only in slot B1. See if the board boots like that, and see if you get the warning that this is not the correct population order for the slots. Shut down, now try the same module on its own in slot A1 instead. If it doesn't boot (stopping on DRAM LED, the LEDs are CPU - DRAM - VGA - BOOT) and you don't get the warning message, something is wrong with that slot altogether. Or to be more precise, usually it won't be the slot itself, but rather, some pins in the socket (connecting to that slot) which might be bent. So in that case you'd check for bent socket pins, see here and the following posts.

However, if the slot is working individually, then it might have more to do with RAM explained: Why two modules are better than four / single- vs. dual-rank / stability testing.

This is definitely still one of the budget board models too - not the worst one by far, but still, less PCB layers and less optimizations compared to some Z-series chipset boards, and i don't know which CPU you have either, its memory controller might not be the strongest one. But let's see what the test with the single module uncovers.
 
First, try one module only in slot B1. See if the board boots like that, and see if you get the warning that this is not the correct population order for the slots. Shut down, now try the same module on its own in slot A1 instead. If it doesn't boot (stopping on DRAM LED, the LEDs are CPU - DRAM - VGA - BOOT) and you don't get the warning message, something is wrong with that slot altogether. Or to be more precise, usually it won't be the slot itself, but rather, some pins in the socket (connecting to that slot) which might be bent. So in that case you'd check for bent socket pins, see here and the following posts.

However, if the slot is working individually, then it might have more to do with RAM explained: Why two modules are better than four / single- vs. dual-rank / stability testing.

This is definitely still one of the budget board models too - not the worst one by far, but still, less PCB layers and less optimizations compared to some Z-series chipset boards, and i don't know which CPU you have either, its memory controller might not be the strongest one. But let's see what the test with the single module uncovers.
The CPU is an Intel i7-12700K.

Per the mobo manual, it says to "Always insert memory modules in the DIMMA2 slot first". I've got memory populated in DIMMA2, DIMMB2 and DIMMB1 without issue. (Though it did take a few re-seating cycles to get this to work.) It's when I add a module into DIMMA1that the thing wont POST.
Pretty sure I don't have bent CPU pins though I have read several posts about that notion.
 
Yes, i'm fully aware in which order to populate the slots. What we want to do now is test the slots individually. We know that A2 and B2 work fine, and you can also add B1 into the mix. But with A1 there is trouble. So we want to use one module only in B1 first, to see what happens with a known good slot, but in the wrong population order.

What should happen is, the board should boot, but it should complain that the RAM is in the wrong slot, by showing you a message on screen. With that confirmed, you shut it off, and now you put it into A1. If you get the same message as before, all is ok with the slot. However, if you now get the DRAM LED and no picture, instead of booting and the warning message, you should look at the CPU socket next for any bent pins. That's a good idea in general too, just to rule out bent pins early on.
 
Alright, no DIMM slot problems and no bent CPU pins. Just an impatient DIMMwit it seems.

Apparently the EZDebug LED's go through a cycle after installing RAM as the new modules are enumerated, or something to that affect.

It would first illuminate the red CPU LED, then switch to the yellow RAM LED and stay there for a bit. All the while I would have no display output. This is when I ASSumed that there was a problem preventing POST.

I decided to just let it sit for a bit longer and then noticed the white VGA LED illuminate. I happened to glance up at my monitor to see a message being displayed that the amount RAM had changed. Entering the BIOS at that message, I see all my RAM.

So, I'm happy to say that I do have my system running 128GB and all DIMM slots populated.

Thanks for the tips and the patience; I obviously need more of the latter...
 
I'm happy to say that I do have my system running 128GB and all DIMM slots populated.

Glad to hear.
But run some memory stability tests because you don't want to end up with a corrupted OS and apps after a few days or weeks.
Also be aware: Intel does not guarantee stability for i7-12700K and 128GB DDR5-4800.
You're on the edge here!
Intel_DDR5.png
 
@RemusM

The specs for my processor indicate that it should support "up to" 128GB of DDR5-4800 RAM.


I noticed in the BIOS that it defaulted to 4000MHz instead of 4800MHz. So I set the XMP Profile to XMP 1 for 4800MHz. I'm currently typing this post while booted using this profile with 128GB.

Are you suggesting this may be an issue? Should I reset the Memory to 4000MHz (default) in the BIOS?

Also, what type of memory stability tests do you suggest?
 
The specs for my processor indicate that it should support "up to" 128GB of DDR5-4800 RAM.

Yes, see again my screenshot (from Intel).
As general rule: more memory you have, lower speed you can achieve.
So from 4800 for 2x16GB to 3600 for 4x32GB.
Be prepared for 4000 or 4400 in case of stability issues at 4800, that was my warning.

Also, what type of memory stability tests do you suggest?

You can use any (MemTest86, TestMem5, etc), but run it at least for 30-60 min.
 
Apparently the EZDebug LED's go through a cycle after installing RAM as the new modules are enumerated, or something to that affect.

Yes, it does so-called memory training. On Intel it usually doesn't take ages like on AMD AM5 sometimes, but still, it can take a minute or two, especially with such a RAM configuration. But hey, glad you got it solved in the end. As for RAM testing, see point 5) of the RAM thread i linked.
 
This is from the PRO B660M-A WiFi DDR5 User Guide:

Yes, that's correct.
Those are the initial release numbers for the Intel Gen 12 CPUs.
After the release of the Gen 13 & 14 CPUs the max speed is the same (6200) due to a motherboard limitation,
but the numbers for 4 memory modules are much better because Gen13/14 IMCs are 800-1200MT/s faster than Gen12.
Max overclocking frequency:
1DPC 1R Max speed up to 6200+ MHz
1DPC 2R Max speed up to 6000+ MHz
2DPC 1R Max speed up to 5600+ MHz
2DPC 2R Max speed up to 5600+ MHz
 
My post was to point out that the motherboard memory specs for the i7-12700K conflict with the Intel DDR5 Memory Support Summary. Whatever memory speed settings are used, it is best to perform some RAM tests.
 
My post was to point out that the motherboard memory specs for the i7-12700K conflict with the Intel DDR5 Memory Support Summary. Whatever memory speed settings are used, it is best to perform some RAM tests.

The real situation about these numbers is much more complicated.
The Intel official numbers are for DDR5 native voltages (VMEM = VDDQ = VDD2 = 1.1V)
The MSI numbers are from the motherboard point of view (cherry-picked CPU, overclocking and overvolting).
So the real range for the Gen 12 CPUs starts from 3600 and ends around 5600.
But for an overvolted Gen 12 CPU the expected range (keeping stability) is between 4000 and 4800 in case of 128GB memory.
Of course, an extensive memory test is a MUST here.
 
FWIW, I ran MemTest64 while my 128GB of RAM was set for 4800MHz and encountered an error within ~15 minutes.

I then set it down to 4000MHz and MemTest64 ran for an hour error free. So I feel I'm in a good place now. Thanks for all of the help folks! I've definitely learned some things I didn't know before.

In case anyone is wondering what I need 128GB for, I need it for running some VM's for various things AND Windows 11 plus multiple apps, browsers with many tabs etc. simply uses more memory than prior OS's. And those VM's, yes those are mostly Windows too with the same hunger for memory. If it weren't for the VM's, 64GB would be plenty; today anyway.. lol

Thanks again @citay and @RemusM !
 
I ran MemTest64 while my 128GB of RAM was set for 4800MHz and encountered an error within ~15 minutes.
I then set it down to 4000MHz and MemTest64 ran for an hour error free. So I feel I'm in a good place now. Thanks for all of the help folks!

As I said before, that was expected.
4000 for 128GB is not terrible, but did you try 4400 ?
 
did you try 4400 ?
Not yet but I will.

However, based on the memory compatibility reference linked below, it looks like the max "supported" speed for my motherboard and 4-32GB modules is 4000Mhz. Luckily while using the exact modules I have. But it will be fun to test anyway.. :)

 
it looks like the max "supported" speed for my motherboard and 4-32GB modules is 4000Mhz. Luckily while using the exact modules I have.

I know, but they used the default 1.1V
If you're "almost" stable at 4800 with 1.1V, you should be stable at 4400.:biggrin:
You could also try again with 4800 but with a small voltage boost: set VMEM = VDDQ = VDD2 = 1.20V
That should keep the IMC cold but make it much more stable.
Again: 4000 for 128GB is not terrible, but 4800 would be great for your workflow (VMs, Docker, whatever you need).
 
Update:
I set the speed to 4800 and attempted to set memory voltage as you suggested @RemusM , Not sure if I did this correctly or not, screenshot below. Your suggestion of "set VMEM = VDDQ = VDD2 = 1.20V" didn't exactly correlate with options I could find in the BIOS.

1742988131260.png


Running MemTest64 with this setting still produces an error after ~11 - 24 loops. (I ran it twice in a row) So either I didn't set the voltage correctly or this RAM isn't stable with that config.

I then tried the XMP Profile 2 for 4400MHz with the defaults of "Auto" for voltages. This is able to run MemTest54 error free for a 1 hour duration.

If I did something incorrectly related to the DRAM Voltage, please let me know and I'll attempt this again. Otherwise it appears that my system is stable with 128GB of DDR5 at 4400MHz.

Thanks again for all of the assistance!
 
I set the speed to 4800 and attempted to set memory voltage as you suggested @RemusM , Not sure if I did this correctly or not, screenshot below. Your suggestion of "set VMEM = VDDQ = VDD2 = 1.20V" didn't exactly correlate with options I could find in the BIOS.

You did the right thing.
VMEM is DRAM Voltage in your BIOS.
The CPU IMC already reacted to your change (DRAM Voltage and DRAM VDDQ) and set CPU VDDQ and VDD2 to 1.20V.
If that's not stable for 4800 not much to do.
Probably messing around with voltages and timings you might achieve "some" stability at 4800, but that's not guaranteed.

I then tried the XMP Profile 2 for 4400MHz with the defaults of "Auto" for voltages. This is able to run MemTest54 error free for a 1 hour duration.

Well, 4400 is better than 4000 so I would be happy.
good luck
 
Last edited:
Back
Top