Integrate Compute for GPU Nodes into an Existing System Architecture POWEREDGE-XE-OPERATE Practice Questions (Page 1)

Questions 1–5

1. 1application · medium

   A team has installed a Dell PowerEdge XE9680 with NVIDIA H100 GPUs into an existing SLES 15 SP5 cluster. After boot, the GPUs are not visible in `nvidia-smi`. The administrator has confirmed the PCIe devices are present in `lspci`. What is the most likely next step?

   • AReboot the node to re-enumerate the PCIe devices
   • BInstall the NVIDIA driver package that matches the SLES kernel and the H100 architecture
   • CEnable the GPU node's onboard VGA controller in the BIOS
   • DCheck the GPU node's power cables and reseat the GPUs
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   During GPU node integration planning, which consideration is essential to ensure the existing system architecture can support the data flow to and from the GPU nodes?

   • AThe number of GPU nodes that can be physically stacked in a single rack
   • BThe operating system version installed on the GPU nodes
   • CThe network bandwidth and latency between the GPU nodes and the rest of the system
   • DThe storage capacity of the GPU nodes' local drives
   Reveal Solution Discussion Select an answer first
4. 3expert · hard

   An administrator is adding a Dell PowerEdge XE9680 to an existing rack that has a 60 A three-phase PDU and air cooling. The XE9680 requires 2400 W per power supply and uses direct liquid cooling (DLC). The administrator has verified that the PDU has enough capacity. However, the rack's air cooling is insufficient for the node's heat output. The administrator has a limited budget and cannot install a new liquid cooling loop. What is the best course of action?

   • AInstall the node and rely on the existing air cooling, monitoring temperatures closely
   • BInstall additional fans in the rack to improve airflow
   • CReduce the GPU power limit via nvidia-smi to lower heat output
   • DUse the node's air-cooled variant if available, or relocate the node to a rack with DLC capability
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   When configuring software for a GPU node, which step is necessary to ensure applications can leverage the GPU for parallel processing?

   • ASetting the GPU to operate in legacy BIOS mode
   • BEnabling the GPU's integrated graphics output
   • CInstalling the appropriate GPU compute toolkit (e.g., CUDA or ROCm)
   • DConfiguring the GPU as a network bridge
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   An organization is adding GPU nodes to an existing HPC cluster that uses a Lustre parallel file system. The GPU nodes will run I/O-intensive preprocessing jobs that read large datasets from Lustre. The existing compute nodes use 10 GbE to access Lustre. What should the administrator configure on the GPU nodes to avoid a storage I/O bottleneck?

   • AUse the GPU nodes' local NVMe drives as a Lustre OST
   • BMount Lustre on the GPU nodes using the management network
   • CConfigure the GPU nodes to use the same 10 GbE network as the existing compute nodes
   • DUse a dedicated high-bandwidth network (e.g., 100 GbE) for Lustre client traffic on the GPU nodes
   Reveal Solution Discussion Select an answer first