50
Chapter 3: Maintenance and Component Installation
Replacing Individual GPUs
Prior to submitting an RMA for GPUs, Nvidia requires that their Field Diagnostic tool is first run
to isolate hardware failures and obtain a log file for review. Check with Supermicro Technical
Services for more details. In addition, more details can be found in Nvidia reference document
Baseboard Field Diagnostics Software Guide DU-09163-001. Use these tools to determine
the failed GPU and locate using the GPU Mapping Table. Be sure to identify the failed GPU
correctly and if any questions reach out to Supermicro Technical Service team to help.
Note:
DO NOT remove the GPU heatsink from the GPU. Supermicro does not recommend
GPU replacement to be done by customers. However, if the customer is familiar with the
product and needs to perform a replacement themselves, SMC can provide full instructions
as needed.
Figure 3-5. GPU Baseboard Replacement