One-Click On-Chip Memory Stress Test/Diagnosis

Function

Ascend DMI supports one-click on-chip memory stress test. That is, you can run commands only once to perform on-chip memory diagnosis, on-chip memory stress test, and on-chip memory high-risk address stress test and output the test results.

**Table 1** Diagnostic items
Item	Time Required	Whether NPU Training or Inference Is Affected	Application Scenario
One-click on-chip memory stress test	< 2 h	Yes	When a training or inference job is executed, an on-chip memory ECC occurs on the NPU, and a new isolated page is added.

The on-chip memory stress test and on-chip memory diagnosis have different application scenarios. For details about on-chip memory diagnosis, see Table 1. For details about on-chip memory stress test, see Table 1. Perform the on-chip memory stress test or on-chip memory diagnosis as required.
If you want to conduct the on-chip memory diagnosis, on-chip memory stress test, and on-chip memory high-risk address stress test at the same time, refer to One-Click On-Chip Memory Stress Test/Diagnosis.

Parameters

Table 2 lists only test-specific parameters. For details about other common parameters, see Common Parameters.

**Table 2** Parameter description
Parameter	Description	Mandatory
[-at, --at, --auto-test]	Performs an automatic stress test. This parameter takes effect only when [-i, --items] contains hbm and -s is specified.	Yes
[-st, --st, --stress-time]	Specifies the time required by the on-chip memory stress test. The command for combined stress tests additionally performs functions such as on-chip memory diagnosis and high-risk address stress test. Therefore, the actual execution time is longer than the specified time. The value ranges from 60 to 604800, in seconds. This parameter must be used together with [-s, --stress] in the scenario where on-chip memory check items are included.	No

Example

ascend-dmi -dg -i hbm -s --auto-test -q

[***@***]# ascend-dmi -dg -i hbm -s --auto-test -q
Stress test is being performed, please wait.
Summary:
    Arch: aarch64
    Mode: ******
    Time: 20250529-19:08:50

Hardware:
    hbm:
        PASS

Fault Check Items

**Table 3** Parameters in the command output
Parameter	Description
PASS	The one-click on-chip memory stress test is successful and no exception occurs.
EMERGENCY_WARN	If the number of real-time isolated pages caused by multi-bit errors is greater than or equal to 64, the current device is at high risk, and you are advised to use spare parts. If the number of isolated pages caused by multi-bit errors is greater than 0 and is less than 64 and any of the following conditions is met, the current device is at high risk, and you are advised to use spare parts: Within the same Bank (where Stack, PC, Sid, and Bank are all the same), the number of isolated rows shall be > 16. Within the same HBMC (where Stack and PC are both the same), the number of Banks with isolated rows shall be ≥ 4, and the cumulative number of isolated rows within the PC shall be > 16. Within the same Sid (where Stack and Sid are both the same), the number of isolated rows located in different HBMCs (i.e., with different PC) shall be ≥ 4. If the number of isolated pages increases for three consecutive times during the stress test, the current device is at high risk. In this case, you are advised to use spare parts.
SKIP	The product or scenario does not support the one-click on-chip memory stress test.
FAIL	The one-click on-chip memory stress test and diagnosis fail. For details, see Figure 1. Contact Huawei technical support or locate the fault by referring to the section "FAQs". Refer to On-Chip Memory Stress Test Fails Due to Insufficient Device Memory.

Figure 1 On-chip memory stress test and diagnosis

Parent topic: Stress Test