AFID Event List - AFID Event List - UG-70122

Each AFID listed in the table below is associated with distinct Category, Type, and Severity fields, which are explained in the following sections.

Category

Below are the definitions for the five categories, each corresponding to specific events.

• Off-Package Link Errors: Covers events such as WAFL and XGMI, ranging in severity from Corrected to Fatal.
• HBM Errors: Encompasses issues like Bad Page Retirement Threshold and On-die ECC, with severities from Corrected to Fatal.
• Device Internal Errors: Includes Hardware Assertion and Watchdog Timeout, assessed from Corrected to Fatal.
• CPER Format Errors: Indicates events related to Malformed or Invalid ACA Data, marked as affecting all severities.
• Unidentified Errors: Represents general unidentified events impacting all severities.

Type

The Type can refer to a specific component, message, event, or error within the specified Category. Examples include On-die ECC and Watchdog Timeout.

Severity

Severity of the failure is linked to specific events. The four severity levels defined in this list are:

• Corrected: These are events that have been corrected by the hardware and are reported for diagnostic purposes.
• Fatal: These are events which trigger a device reset for recovery.
• Uncorrected, Non-fatal: These are events which cannot be corrected but can be contained.
• ALL: This can be of any of the three aforementioned severities.