ram that causes problems in booting the system

[Rosin0416]

Hello everyone,
I have a supermicro X10SL7-F motherboard which has 4 ram slots.
I was running 2x8GB ram (CT102472BD160B) on DIMMA1 and DIMMA2. The ram had been tested and working for 6-7 years.
In order to complete the ram on my system, I purchased two additional ram CT102472BD160B to have the maximum 32GB of ram.

When I inserted the new ram into the remaining slots (DIMMB1 and DIMMB2) to test the whole thing with memtest86, the system seemed to boot "correctly". Then at the next reboot I had 4 short beeps at the time of the system initialization. Series of beeps that are not the usual ones of the card and the system crashed without going further.

I went on a big investigation and I'll pass on some of the details.

I made a first series of tests :
_ Each ram bar was tested alone on DIMMA1 (4 pass with memtest86 version 9.4 free) => no error detected
When I put two ram bars (whatever the bar) on DIMMA1 and DIMMA2, I have no problem (no beeps) and the system boots normally.

I did a second series of tests : each ram bar was tested alone on each of the 4 slots (DIMMA1, DIMMA2, DIMMB1 and DIMMB2)
Test protocol:
1] launch a linux mint liveUSB with short navigation (open software, browse folders, no more than 1 minute)
2] shutdown
3] launch memtest86 (1 pass / 13 tests)
4] Ctrl+Alt+Del to reboot the system
5] on reboot, launch LiveUSB linux mint with short navigation (open softwares, navigate in folders, not more than one minute)

Several remarks on the results:
_ no matter what bar and location (16 tests), no error was detected by memtest86.
_ at no time did I get the 4 error beeps at startup. I only got the normal beeps when everything is fine.
_ the two memory sticks I just bought work on all 4 slots without any problem.
_ I have had problems with both of my existing memory sticks.
_ the problems occurred only on channel B => DIMMB1 and DIMMB2
_ for the pairs of strip/slot where there were problems => I never had latency or other problems at the first boot. It's during the second boot that it's not so good and where it happens (very, very) big latencies.


Third series of tests:
I have tried several pairs of barrette/slot with two, three or four barrettes
I have problems of big latency if I put the two strips that worked alone on DIMMB1 & B2 together.
I have the impression that as soon as I put the B channel in, there are problems.


I have noticed that if I make a change (remove a ram, put one back,...), the system starts up for the first time without any problem.
On the other hand, as soon as the second boot, either the system beeps (4 short beeps) and doesn't go any further, or the system does a looped reboot, or I get to the OS (windows or linux) but it's very very very slow (even freezes).



I would also like to point out that :
I disassembled the motherboard to see if there were bad soldering, if there were pins that could not make contact, inflated capacitor, .... => nothing visible
I dismantled the processor and cleaned the contacts => no effect

Do you have any idea of the problem?

 

[Rosin0416]

Hello,
I'm coming back to you with a lot of new and more accurate data that I've compiled.

_ the BIOS (aptio) is up to date in its latest version.
_ the meaning of the 4 short beeps are not listed in the manual, and I don't have the details.
For my tests, I have a liveUSB of memtest86 version 9.4 free, a liveUSB of linux mint, a hard disk where Windows 10 and Linux Mint are installed in dual boot
I don't have any other system to test the cpu or the chips.
Following the tests I have done below, I noticed that I can boot the system once without any problem whatever the number of ram (>2) installed or their location on the board. If I shut down the system and turn it back on, the problem occurs in one form or another (see the forms mentioned below). I just have to remove a strip, exchange a strip with another one, or put one back in order to start the system again without any problem (only once).

TEST 1 : attempts of different strip/slot configurations
Conclusion:
_ whatever the combination with only 2 DIMMA1 and DIMMA2, no problem detected
_ the problem occurs as soon as there is a third ram bar.
_ the problem occurs in 3 different ways:

either 4 short beeps at the time of system initialization (before the moment when I can invoke the BIOS with the "Suppr" key)
no abnormal beeps at startup, the OS manages to launch itself, but when it arrives on the system, a very big latency occurs which does not allow to do anything on the system anymore
the system restarts in a loop

TEST 2 : each ram bar has been tested individually on DIMMA1 with the Memtest86 software version 9.4 free (4 passes)
Conclusion : no error detected by the software

TEST 3 : each ram strip was tested alone on each of the 4 ram slots
Test protocol :
1] launch of a linux mint liveUSB with short navigation (open software, browse folders, not more than one minute)
2] shutdown
3] launch memtest86 (1 pass / 13 tests)
4] Ctrl+Alt+Del to reboot the system
5] on reboot, launch LiveUSB linux mint with short navigation (open software, navigate in folders, not more than one minute)
Test conclusion:
_ no error on any test was detected by memtest86
_ at no time did I get the 4 error beeps at startup. I only had the normal beeps when everything was fine.
_ for each first boot, I never had any latency problem or other. It's during the second startup that things get bad and there are (very, very) big latencies.
The problems only happened on channel B => DIMMB1 and DIMMB2

TEST 4: I didn't see it at first, but it turns out that there are logs in the bios
There were the following messages:
Smbios 0x01 DIMMB1 Single bit ECC memory error (x20 times)
Smbios 0x02 DIMMB1 Multi bit ECC memory error (x3 times)
Smbios 0x01 DIMMB2 Single bit ECC memory error (x1 times)

After several tests, it turns out that :
_these error codes are generated when I manage to launch a system and it goes into high latency mode. When I have the 4 error beeps at initialization, the system must detect something since it beeps, but does not record in the logs at that time.
After doing a lot of strip/slot combinations, it seems that the error messages in the logs only appear on DIMMB1 and DIMMB2. However, for almost all the combinations tested, I only have the 4 error beeps, which are not referenced in the logs. As said above, the errors seem to be referenced later in the system boot => test results to be taken with hindsight.

TEST 5 : as the errors are recorded in the logs when I manage to launch a system, and as for any first launch the system seems stable, I wanted to know if the system remained stable on a longer duration and if errors could appear.
For that, I reset the bios logs, I managed to boot on windows with a working OS, and with my 4 ram bars installed, I let it run all night => the next day, the OS was still reactive, no latency problem.
So I thought I would push the system a bit, I downloaded the memtest software from this editor : https://hcidesign.com/memtest/
It's a software that stresses the available ram on windows in operation. The software ran all afternoon without detecting an error.
After returning to the BIOS, no 0x01 (single bit error) or 0x02 (multi bit error) errors were detected


In the middle of all this, the motherboard has been disassembled:

I checked the soldering of the RAM slots as well as the contact tabs (with and without ram) => no problem
I checked if there was a swollen capacitor or other => no problem
I dismantled the processor and cleaned the contacts.
I launched the IPDT.exe software, the diagnostic software of intel => no problem
=> the problem persists

I'm starting to run out of ideas. So I appeal to you.
Maybe some parameters to configure in the bios ,.....

I used a translator to transcribe my explanations, sorry if the translation is not accurate. I can try to clarify some points if needed.
Thank you

 

[pschatz100]

From a few internet searches, it appears that 4 beeps may indicate an unrecognized CPU. As you already indicated that your bios is up to date, and that CPU is supported, it is possible that you are experiencing a timing conflict when more than two memory modules are being used. This could be a hardware problem or else a timing problem with the memory modules themselves.

Crucial does not list the Supermicro X10SL7-F as a supported motherboard. That, in itself, does not mean the memory wouldn't work but it would indicate that Crucial did not certify it.

I would try the following:

1) Make certain you are putting the memory modules in the right slots. For instance, I have a Supermicro X9SCM motherboard, and my manual says that for two memory modules, you should use slots DIMM2A and DIMM2B. Check your manual to see if there is a specific order in which the memory slots should be populated. If there is, follow the instructions to see if it makes any difference.

2) In your bios setting, try slowing down the memory speed. Instead of running the memory at 1600Mhz, try running at 1333Mhz. If that does the trick, then you know the problem was related to memory timing. If you are running the system as a file server, or for hosting an application like Plex, you probably won't notice any real difference in performance.

Of course, it might also be that the Crucial memory is not fully compatible with your motherboard. Even though the specs may look OK, sometimes motherboards are "fussy" about the memory, especially if the memory is running just at the edge of compatibility. The tolerances become tighter when running 4 modules than they are for two modules.

 

[Rosin0416]

Hello and thank you for your comments.

Crucial does not list the Supermicro X10SL7-F as a supported motherboard. That, in itself, does not mean the memory wouldn't work but it would indicate that Crucial did not certify it.

Indeed, the crucial ram has not been tested by supermicro for this card. However, there are several members of this forum who have this motherboard with 4x8GB ct102472bd160b without reporting any problems. I am thinking of the members (see signatures): DiViDeR, Jailer

Moreover when you read these posts this ram seems to be recommended for this motherboard :
https://www.truenas.com/community/t...or-supermicro-x10-lga1150-motherboards.23291/
https://www.truenas.com/community/r...ns-for-supermicro-x10-lga1150-motherboards.6/

1) Make certain you are putting the memory modules in the right slots. For instance, I have a Supermicro X9SCM motherboard, and my manual says that for two memory modules, you should use slots DIMM2A and DIMM2B. Check your manual to see if there is a specific order in which the memory slots should be populated. If there is, follow the instructions to see if it makes any difference.

I had read that on some posts. But on the manual and unless I am mistaken, there is no particular indication on the order of placement of the strips.

2) In your bios setting, try slowing down the memory speed. Instead of running the memory at 1600Mhz, try running at 1333Mhz. If that does the trick, then you know the problem was related to memory timing. If you are running the system as a file server, or for hosting an application like Plex, you probably won't notice any real difference in performance.

I also tried this solution (which I forgot to list in my post above). The problem also appears at 1333MHz.

Bien sûr, il se peut aussi que la mémoire Crucial ne soit pas entièrement compatible avec votre carte mère. Même si les spécifications semblent correctes, il arrive que les cartes mères soient "tatillonnes" en ce qui concerne la mémoire, en particulier si la mémoire se situe juste à la limite de la compatibilité. Les tolérances deviennent plus étroites lorsque vous utilisez 4 modules que pour deux modules.

Are there any parameters I can play with to detect this?

Regards