[SOLVED] Memory bank A unusable, only memory bank B is

[igorjrr]

I have an ASRock 970 Extreme4 and AMD PhenomII processor. It's about 10 years old. There are four RAM slots: A1, A2, B1, B2. I have four RAM sticks exactly the same (brand, model, size, specs and even manufacturing date).

If I install any RAM stick on either or both slots A, I get error 54 ("unspecified memory error") in the motherboard and it doesn't boot.
Installing individual or both slots B work, and the computer starts and can be used normally, including gaming with a RX590 video card. Even if there are RAM sticks in bank A, as long as there is at least one stick in bank B, the computer boots. However, memory sticks from bank A are never usable.

When all slots (banks A and B) are used, BIOS show the correct total memory (for all four slots, that is, 4x4 GB = 16 GB). However Windows and USB-booted memtest report only the available memory from banks B (2x4GB = 8GB), that is, BIOS recognizes memory sticks in bank A but don't allow them to be used by Windows or memtest.
The memory banks A are physically fine, inspected, cleaned.
Each stick has been tested separately in bank B, and they all work fine. They are not the source of the problem.

It puzzles me that memories in bank A are identified by BIOS (it shows all specs correctly) but are not usable by any Operating System. If the socket had an issue, then I'd assume it would never identify these sticks in bank A, or something weird would come up.
Strange that when memory is installed only in bank A (leaving bank B empty), then there is that BIOS error 54 and the computer won't boot.

Nothing related to software as USB-booted memtest sees all four RAM sticks, but it shows only the memories in bank B as "available memory".
There is no setting in BIOS reserving the memory (I don't have an internal graphic processor, it's a PCIe card).

My question is: where is likely the problem? Why BIOS see these memories, but mark them as unavailable? Is there any other test, check that can be done?
Maybe BIOS runs some basic testing and because they fail, they are not available for use. I would like to understand what might be causing it (even if it's something electric in the motherboard).
Is it possible the CPU has some issue that is no longer able to use any memory from bank A? Or more likely a physical damage in/around bank A?

I tried with a brand new PSU, same problem.

My point here is to investigate what is the source of the problem, even if it's not fixable.

Thank you for any ideas!

 

[punkncat]

Given the age of this computer it could be that the memory controller is failing?

Was this a recent thing, such as been working fine for 'x' and then "yesterday" starting doing this?
Any changes inbetween?

 

[igorjrr]

Given the age of this computer it could be that the memory controller is failing?

Was this a recent thing, such as been working fine for 'x' and then "yesterday" starting doing this?
Any changes inbetween?

Is the memory controller unique for all banks (2xbank A + 2xbank B) or by bank? If it's failing, why it would affect only the slots A? I would also expect some random issues with bank B, but bank B is working perfectly.

Recent thing. It was running with full 16 GB until two days ago. I had cleaned the computer (dust), and reinstalled the CPU (removed to clean fan). The computer has been running perfectly fine except this issue with memory bank A.

Yes, it's an older CPU, and will be replaced, but as an IT, I'd like to try to understand where the issue is. Unfortunately I don't have another CPU or motherboard to troubleshoot further.

So I'm asking here in case there are more knowledgeable friends that might have an idea why memory bank A (2 slots) recognize the memory sticks, but they are not usable. And when sticks are used only in bank A, the BIOS error 54 comes up.

 

[punkncat]

If you took the CPU out and this issue presented I would check the mounting. Likely something is wrong there, or you damaged some pins, trash/dust, etc.

 

[igorjrr]

If you took the CPU out and this issue presented I would check the mounting. Likely something is wrong there, or you damaged some pins, trash/dust, etc.

You mean pins, trash/dust, mounting of the CPU? This morning I removed it again, checked PINs and used a vacuum to remote any possible dust.

If the CPU had any PIN issue, trash/dust, wouldn't it prevent it from working? Because the computer is fully usable if there is at least one memory stick in bank B.

I'm just trying to understand a little more how memory banks (A and B) are connected with motherboard and CPU, to try to find where the cause might be.
So far I can't think it's an issue with the CPU because that would certainly have much bigger consequences. Everthing is working fine, even for gaming, except the two slots in bank A (BIOS recognize the memory sticks there, but make them unavailable for use).

 

[punkncat]

In short, your CPU has a number of PCI-E lanes that "communicate" back and forth between components and the CPU. This includes your memory channel(s), your GPU, expansion cards, drives, and so forth.
IF during the process that you were taking this whole thing apart (which was working fine before) anything got damaged, static electricity, ghosts, and other such phenomenon and you NOW see that the A channel for your memory isn't working as opposed to it's correct operation before certain assumptions have to be made.

With that said. Given the age of that system and it's current operational status, I would suggest you limp along until new(er) relevant replacements can be procured. I would not spend money chasing or fixing this platform CPU/mobo unless you find stuff free or nearly so.

 

[igorjrr]

In short, your CPU has a number of PCI-E lanes that "communicate" back and forth between components and the CPU. This includes your memory channel(s), your GPU, expansion cards, drives, and so forth.
IF during the process that you were taking this whole thing apart (which was working fine before) anything got damaged, static electricity, ghosts, and other such phenomenon and you NOW see that the A channel for your memory isn't working as opposed to it's correct operation before certain assumptions have to be made.

With that said. Given the age of that system and it's current operational status, I would suggest you limp along until new(er) relevant replacements can be procured. I would not spend money chasing or fixing this platform CPU/mobo unless you find stuff free or nearly so.

Thank you. If you don't mind I just have another question: CPU you mean the processor or the whole processor+socket?
In this scenario I described, is it more likely it's a problem within the motherboard (socket, etc.), or the processor?
I'm asking in case I'm able to find a replacement just for testing, for learning/curiosity. Would you say it is possible that everything would work fine just by replacing the processor? Or more likely it would be motherboard (keeping same processor)?
To me, in my limited knowledge, I can't picture that the processor would be the problem since everything works fine, except for memory bank A. To me, it would make more sense that motherboard got damaged in a way that memory bank A identifies the memory stick(s) installed, but not all channels work for using them, so BIOS make them unavailable. Would you agree or see something different?
I know, it's a guess, I'm just looking for your opinion. I'm just trying to learn more. Thank you!

 

[punkncat]

Total guess.

I have no idea where you are located, but spending ANY money to fix this old a platform is wasted. Even a lowly i3 10100 would beast mode that Phenom.