Supermicro SMI300XE - MI300X Expert (SMI300XE) Certification Exam

TheAGFHC tool for Ubuntumust be installed in the/opt/AMD/AGFHCdirectory.

Option A (Incorrect):The/tmp/directory istemporary storageand is not used forapplication installations.

Option B (Incorrect):/windows/user/desktopis a Windows path and does not apply to Linux-based Ubuntu.

Option C (Incorrect):/usr/home/customer/AGFHCis not the correct standard directory for system-wide software.

Option D (Correct):/opt/AMD/AGFHCis the correct directory for installing AMD-related tools in Ubuntu.

Thecorrect torque setting for GPU screwsin anMI300X systemis6-8 in-lbf. This ensures asecure fit without overtightening, which could damage the GPU.

Option C (Correct):6-8 in-lbf isthe recommended torquefor securing GPU screws.

Option A (Incorrect):4.2 in-lbf istoo low, leading to loose screws.

Option B (Incorrect):9.5 in-lbf istoo high, risking damage to the GPU mounting.

Option D (Incorrect):5.3 Nm isnot used for GPU mounting screws.

ForSupermicro MI300X GPU systems,removing the power supply unit (PSU) improperlycan lead to serious issues.

Why the red lock exists?

Thered locking mechanismensures that thePSU is properly securedin the system.

Before disengaging the PSU, thered lock must first be unlockedto avoid damage.

If you forcefully shift the release lever without unlocking the red lock:

B: The power supply unit will fail→ If the PSU is removed improperly, it can causeinternal damageto power connectors, resulting in PSUfailure or improper functioning.

C: You could permanently damage the lock on the power supply unit→ Thelocking mechanismis designed for smooth disengagement. Forcefully shifting the lever canbreak or warpthe lock, making itimpossible to secure the PSU back in place.

Incorrect Answers Explained:

A (All answers are correct)→ This is incorrect becausenot all answers are valid.

D (The PSU can no longer be secured into the tray)→ This happensonly if the lock is damaged, which is covered underoption C.

Supermicro’s official documentation clearly statesthatunlocking the red lock firstis a critical step before removing or replacing the PSU.

TheGPU in the MI300X systemis secured withT15 Torx screws, requiring aT15 Torx Bitfor removal.

Option A (Incorrect):T20 Torx Bit istoo largefor GPU screws.

Option B (Incorrect):Security Torx Bit isnot requiredfor standard GPU removal.

Option C (Correct):T15 Torx Bit is the correct toolfor removing GPU screws.

Option D (Incorrect):Phillips Head Screwdriversdo not fit the GPU screws.

TheIPMI Dashboard health event loghelpsidentify component failuresusing ared severity status indicator. If a fan or power supply fails, theevent log will mark it as a red alert, allowing users to take corrective action.

Option A (Correct):Red severity status in theevent logsignals a failed component.

Option B (Incorrect):Fansdo not change colorswhen failing.

Option C (Incorrect):There is nopurple status indicatorin IPMI for failures.

Option D (Incorrect):You cannot configure a new fan inIPMI under Configuration; you must replace a failed fan.

IfROCm commands fail to provide output, the primary issue is likelymissing or undetected GPUs. Thebest way to validate this is by checking the IPMI GUI and reviewing the GPU component sectionto ensure that all GPUs are present.

Option A (Correct):Missing GPUsare the most common causeof ROCm failures. The IPMI GUI provides visibility into which GPUs are detected.

Option B (Incorrect):CPU misalignment would causegeneral system failure, not ROCm-specific issues.

Option C (Incorrect):The GPU baseboard status can be checked in IPMI but isnot the primary causeof ROCm failures.

Option D (Incorrect):ROCm can function with multiple missing GPUs, butthe key issue is confirming GPU presence first.

Toverify the AOC NIC installedon the system, the correctSUM (Supermicro Update Manager) commandis:

css

CopyEdit

sum.exe -i 10.1.1.100 -u ADMIN -p ADMIN -c getaocinfo

Option B (Correct):This command retrievesAOC NIC information.

Option A (Incorrect):"getaocnicinfo" isnot a valid command.

Option C (Incorrect):This syntax isincorrect and will not execute properly.

Option D (Incorrect):"getnicinfo" retrievesgeneral NIC details, not AOC NIC-specific information.

When handling theGPU tray, it should be placed ona flat, stable surfacetoprevent damage and ensure safety.

Option B (Correct):Aflat surfaceprovides stability and preventsaccidental damage or tipping.

Option A (Incorrect):Placing the trayon top of the chassiscan cause it tofall and sustain damage.

Option C (Incorrect):Thefloor is unsafedue tofoot traffic, riskingaccidental damage.

Option D (Incorrect):Acardboard box may not support the tray’s weight, making itunstable.