What Is Power-On Self-Test Sequence?_
POST is a built-in firmware routine that checks critical components such as CPU, memory, storage controllers, and GPU hardware (e.g., VRAM, PCIe link status) for proper function. It typically reports errors via beep codes, LED indicators, or on-screen messages if a component fails. Successful completion allows the system to proceed to boot the OS, while failure halts the process for troubleshooting.
Technical Details
During POST, the server’s BMC monitors and validates GPU power rails, performs PCIe link training, and checks NVLink fabric connectivity without loading GPU drivers. The sequence typically includes initialization and a basic check of each GPU’s HBM memory, verifying ECC capability and bank mapping. For liquid-cooled systems, POST also checks coolant flow sensors and pump status before allowing GPU power-up. Errors are logged in the BMC’s system event log (SEL) for later analysis.
How Leviathan Systems Works with Power-On Self-Test Sequence
In our field work, POST is the first step after rack assembly and cabling: we power each GPU server and watch for the front-panel health LED to turn green, indicating no critical faults. If the LED stays amber or red, we check the BMC console for POST error codes, often tracing issues to loose power cables or unseated GPUs.
Related service
Network Testing & Commissioning →