Resolving a GPU, PCIe adapter, or device problem

Learn how to access log files, information to identify types of events, and a list of potential problems and service actions.

1. To identify the correct service procedure to perform by using operating system log information, complete the following steps:
   1. Log in as the root user.
   2. At the command prompt, type dmesg and press Enter.
2. Scan the operating system logs for the first occurrence of keywords, such as fail, failure, or failed. When you find a keyword that accompanies one or more of the resource names in Table 1, a service action is required.

   Did you find an operating system log that requires a service action?

   If	Then
   Yes:	Use Table 1 to determine the service procedure to perform for your type of problem. This ends the procedure.
   No:	Continue with the next step.

   Table 1. Resource names, examples, and service procedures for different types of operating system logs.

   Resource name	Example of a log requiring a service action	Type of problem	Service procedure
   mpt3sas	PCI error detected 2	RAID	Go to Resolving a RAID adapter problem.
   eth1, eth2, eth3	Failed to re-initialize device	Network	Go to Resolving a network adapter problem.
   NVRM	aborting RmInitAdapter failed!	Graphics	Go to Resolving a graphics processing unit problem.
   nvme	Failed status: ffffffff, reset controller	NVMe Flash adapter	Go to Resolving an NVMe Flash adapter problem.
   ata1, ata2	SError: { RecovComm PHYRdyChg 10B8B Dispar }	Marvell storage adapter	Go to Resolving a storage device problem.
   sda, sdb, sdc	FAILED Result	Storage

3. Are all of the adapters in the system missing or failed?

   If	Then
   Yes:	Perform the following actions, one at a time until the problem is resolved: Ensure that the PCIe risers are fully seated in the system. Replace system processor CPU 1. Replace the system backplane. Note: If your system is an 8001-12C or 8005-12N, go to 8001-12C or 8005-12N locations to identify the physical location and the removal and replacement procedure. If your system is an 8001-22C or 8005-22N, go to 8001-22C or 8005-22N locations to identify the physical location and the removal and replacement procedure.
   No:	Go to Collecting diagnostic data. Then, go to Contacting IBM service and support.