EPYCD8 "reset" multiple time after frozen.

[RE0]

I have a ASRock Rack EPYCD8 motherboard with EPYC 7401P processer and a single micron RECC memory module.

Sometimes the system will be frozen (No matter which software is running on it). Few seconds later the monitor goes black and the debug led on the board shows a binking "00" before it successfully "reset" something and started to boot.

Some cards I plugged into the pcie slot:
XFX RX590
PCIE to USB3.0 converter (NEC chips)

The only memory module on the board:
MTA36ASF2G72PZ-2G1A2IK DDR4 RECC
Memtest shows no error.

It will be appreciated if anybody tells me what is happening on this system.

Sorry for my weak English.

Best regards.

 

[Patriot]

Look up the health logs in ipmi.

 

[RE0]

Look up the health logs in ipmi.

There are dozens of "PCI SERR" lying in the ipmi event log.But the time that these events was generated doesn't match the time that system was frozen.

 

[RE0]

Look up the health logs in ipmi.

Here is the full event logs.
Firefox can handle this file correctly.
Thank you for your help!

 

[RolloZ170]

384  | 05/25/2023, 07:12:48 | DRAM ECC ErrorE1 | Memory                             | Correctable ECC - Asserted

the limit of the corr. ECC count can be reached, im that time server should freeze or reboot.
replace microc RDIMM, memtest86 EFI can poll ECC errors. without ECC polling you see no errors in memtest.

 

[ocfguy]

Just to clarify, MemTest86 as in the proprietary version. The open-source Memtest86+ doesn't have EDAC support yet.

MemTest86 - Download now!

Download your free copy of MemTest86 here for USB.

www.memtest86.com

 

[RE0]

384  | 05/25/2023, 07:12:48 | DRAM ECC ErrorE1 | Memory                             | Correctable ECC - Asserted

the limit of the corr. ECC count can be reached, im that time server should freeze or reboot.
replace microc RDIMM, memtest86 EFI can poll ECC errors. without ECC polling you see no errors in memtest.

Thanks for your reply.
That day I took the mem out and put it into another slot without cleaning the golden finger.It boots up and failed immediately.
Then I noticed that the system is complaining about memory error which wasn't happened before.
I took that out, cleaned the golden fingers then put it back in. Erased the SEL & run memtest under efi.No error reported on both memtest and SEL.
This morning(May_28) the system freezed again.I can't see any valuable items in the log.
Now I am really confused.But I will replace that rdimm anyway.

 

[RolloZ170]

had similar things with asrock rack and intel lga3647 cpu at only one RDIMM. with two the reboot (CATERR) never happened again.
and AFAIK EPYC 7001 looses many performance running with only one mem.stick.

 

[RE0]

had similar things with asrock rack and intel lga3647 cpu at only one RDIMM. with two the reboot (CATERR) never happened again.
and AFAIK EPYC 7001 looses many performance running with only one mem.stick.

Yep.Only one of eight channels are enabled.
I'll go with quad Samsung 8G RDIMM .