Question ASRock B450M Steel Legend PMU Training Error/DIMMs B1 & B2 not working?

[TM1172]

So this may wind up being a sad story of self-imposed stupidity but I'm hoping there's a fix.

Last night, I tried upgrading from a 2600X to the current 3600X. I didn't update the bios, b/c on ASRock's website, it said that for the 3600X, bios 2.1 should be compatible and I was already running 2.7 (not sure how but hey).

Initially, the computer just boot looped for minutes. No splash, no bios screen, just power on - 30 seconds - power off - repeat. In a moment of stupid, I thought maybe the cpu was having trouble with RAM, so I pulled the second pair out (with power off), now operating only with 16GB in slots A2 and B2 per the manual.

I replaced the 2600X into the socket, the computer booted just fine, so I downloaded bios 2.9 from ASRock and flashed the bios, figuring it was probably a bios compatibility thing. I also popped the CMOS battery for a minute.

After flashing the bios, I put the 3600X back in the slot, held my breath, and powered on. It booted! But, it gave me a message,

"PMU Training Error at Socket 0 Channel 0 DIMM 0 & DIMM 1"

Going into the bios, only the memory sticks in slots A1 and A2 are now being recognized. The memory in slots B1 and B2 are not recognized in the bios - reads empty. So with 16 gb installed in slots A2 and B2, only 8 gb was recognized by bios. With memory installed in all four slots, only the 16gb in A1 and A2 show up.

I've taken the following steps to troubleshoot:

-Reseat all memory
-Try memory sticks in different slots (to rule out bad RAM)
-Reseat CPU, loosen cooler
-Clean memory contacts and blow out DIMMs
-Boost DRAM voltage to 1.35v
-Boost SoC voltage to 1.04v
-Reset CMOS by removing battery

In CPU-Z, all 32gb are recognized and each channel shows timing info. Also, each stick is definitely getting power, because the sticks are RGB and they're lighting up just fine.

The potentially tragic part: When I pulled 2 of the 4 sticks of RAM, I think I partially unseated the stick that was in slot B2, so it was half in-half out when the machine was boot looping. Being an idiot, I just pushed it back in until it clicked. I don't think it hurt the RAM itself, since that stick later worked in one of the A slots, but I'm worried that it might have torched the mobo memory controller.

I tried putting the 2600X back in to see if all four memory channels would be recognized (thereby indicating bad CPU memory controller) but it just boot looped due to the updated bios.

So, the question - is this a problem caused by A. Bad MoBo memory controller, B. Bad CPU memory controller, C. bios update, or D. Something else?

Would pulling the motherboard out and re-seating everything from scratch be advisable? Am I totally hosed/need a new component? I don't want to be stuck with 16gb in single channel config.

Specs:
Ryzen 5 3600X
ASRock B450M Steel Legend
HyperX Fury DDR4, 4x8 2666
ASUS ROG STRIX GTX 1080 Ti
Crucial P1 1TB NVMe (Boot)
EVGA Supernova 850w Gold

 

[Lutfij]

I would advise on powering down the system, disconnecting from the wall and tear down the entire system, then remove the CMOS battery from the board and replace it after 30 mins. Next with one stick of ram(in A2) reflash the BIOS but prior to doing so, check and see that you're on the latest chipset drivers(as stated alongside BIOS version 2.5) and then reflash the BIOS with version 2.9.

See if that helps. I think the issue is with the BIOS being a little...off or that there's a problem with the way you did things. Usually we work our way up to hardware upgrades, when flashing, I make sure I'm on the basic BIOS version that accepts all hardware and once I see that all the components are detected, I go ahead with BIOS updates(if pending).

 

[TM1172]

I would advise on powering down the system, disconnecting from the wall and tear down the entire system, then remove the CMOS battery from the board and replace it after 30 mins. Next with one stick of ram(in A2) reflash the BIOS but prior to doing so, check and see that you're on the latest chipset drivers(as stated alongside BIOS version 2.5) and then reflash the BIOS with version 2.9.

See if that helps. I think the issue is with the BIOS being a little...off or that there's a problem with the way you did things. Usually we work our way up to hardware upgrades, when flashing, I make sure I'm on the basic BIOS version that accepts all hardware and once I see that all the components are detected, I go ahead with BIOS updates(if pending).

I’ll try that this evening after work. Thanks for the advice, I’ll update as I go.

 

[TM1172]

So, I did all the things mentioned above this evening. I took the computer off power, completely disassembled it, popped the CMOS battery out and left it for some hours. I took the CPU out, cleaned it thoroughly, and made sure the sockets had nothing in them. I unseated all the RAM and cleaned the contacts. Finally, I reassembled the pieces, and re-flashed the BIOS with 2.9 with only one stick of RAM in slot A2. Before doing all this, I updated my AMD chipset drivers and the ASRock drivers, too.

I ran through each DIMM and found that one of them is probably bad - it booted, but then BSOD'd saying "win32kfull.sys failed" or something similar. Upon the next boot, it BSOD'd immediately, saying PFN_LIST_CORRUPT. Finally, it wouldn't boot past the splash screen, saying "a required device is missing. Please use boot media." Soooo that module is probably bad. However, the other three modules all allowed the computer to boot right up - I'm typing on it right now. Each module functioned in the A2 slot.

Next, the big test - would a known good module function in the B2 slot with one in the A2 slot? And while the computer booted just fine, I once again received a message saying "Memory PMU Training Error at Socket 0 Channel 0 DIMM 1." Windows only recognizes 8gb RAM despite CPU-Z recognizing 16gb in two slots. CPU-Z also says the memory is working in Dual channel.

Interestingly, although probably unrelated, in CPU-Z, only one of the DIMMs comes up with Module Manuf/DRAM Manuf/Part Number/Serial Number/Week&year data. The other one, those windows are blank - and this is a matched pair, purchased as a set.

So, best guess, yes, I will need to RMA that bad stick of RAM, but there's something else at work here, and I'm really puzzled as to what it is.

 

[alexbirdie]

You did all testing which I had done in such situation. Sorry to say, but looks like for me like a bug in the motherboard.

 

[TM1172]

Update: last night, on a whim, I popped the 2600X back into the motherboard using only two sticks of known good RAM and powered up. The machine came right to life, and didn’t display any memory errors on boot. The BIOS recognized the RAM in slot B2, and so did Windows.
So, I believe what I have is a bad DIMM that just decided to die and the memory controller on the CPU is shot, too, since the motherboard seems to be just fine with the memory using the old chip. What are the odds?