PowerEdge: CPU 1 Machine Check Error on 13G with Intel Grantly When Write Data CRC Enabled

During Q2FY18 EST testing, EST found "CPU 1 machine check error detected" log as photo [1] occurred in R430 server on every warm boot after changing BIOS setting "Write Data CRC" from default (Disable) to Enable.

1. This issue could be reproduced in the condition "Dell 13G servers (since MLK RTS) and Intel Grantley CRB."
2. The feature of "Write Data CRC" was defined from the DDR4 spec and was introduced since Intel’s Broadwell CPU.
3. The "Write Data CRC" option is in the page of "F2 (System Setup)" -> "System BIOS" -> "System Profile Settings" -> "Custom," the default setting is disabled which dose not causes this issue.
4. No POST stops responding nor functional impact with this symptom, only "CPU 1 machine check error detected" logged.

[1] 

 

The root cause is

1. Function which enable WrCRC is not executed in the WarmBootFast boot path WrCRC has not been enabled completely and cause the issue.
2. Intel plans to introduce the solution in the next version of MRC. Expect to have the next version of MRC released by end of May (Target in Q3FY18 BIOS).

1. Do not replace any hardware
2. The root cause is WrCRC function (when enabled) is not executed in the WarmBootFast boot path WrCRC has not been enabled completely and trigger the log.
3. The fix has been included from BIOS 2.5.4 version.