A customer needs to run a generative AI workload at multiple, dispersed edge locations. Each location has significant space and power constraints. The workload is inference-only and does not require the absolute highest performance, but rather a balance of good performance and energy efficiency.
Which HPE ProLiant server and NVIDIA GPU combination is specifically positioned for this type of edge AI use case?

• A. HPE ProLiant DL320 Gen11 server with NVIDIA L4 GPUs
• B. HPE Cray system with NVIDIA Grace Hopper Superchips
• C. HPE ProLiant DL384 Gen12 server with NVIDIA H200 NVL GPUs
• D. HPE ProLiant DL380a server with NVIDIA H100 NVL GPUs

An AI application is being developed to analyze financial time-series data to predict stock market trends. The model needs to understand long-term dependencies and the sequence in which data points occur.
Which neural network architecture is specifically designed to handle sequential data like text or time- series?

• A. A fully connected, feedforward neural network
• B. Transformer model
• C. K-Means Clustering Algorithm
• D. Convolutional Neural Network (CNN)

An architect uses the HPE Intelligent Configurator for a customer with 120 concurrent users for a text generation task with RAG. The tool recommends a "Medium - Expanded (4-node)" configuration. The customer then reveals they want to use a smaller, more efficient 7B parameter model instead of the 13B model the tool defaulted to.
How will this change in model size likely affect the sizing tool's recommendation?

• A. It will have no effect, as the number of users is the primary factor.
• B. It will upgrade the recommendation to a "Large - Standard (4-node)" configuration.
• C. It will downgrade the recommendation to a "Medium - Standard (2-node)" configuration.
• D. It will require the architect to switch to the "application capacity" sizing method.

An organization is deploying a multi-tenant AI environment using HPE Private Cloud AI. They need to run several smaller, independent AI inference workloads on a single, powerful NVIDIA H100 GPU to maximize resource utilization. Each workload must be securely isolated with its own dedicated portion of the GPU's compute and memory resources.
What key NVIDIA technology, supported on the Hopper architecture, allows for this secure partitioning of a single physical GPU? (Select all that apply.)

• A. Confidential Computing
• B. Multi-Instance GPU (MIG)
• C. GPUDirect Storage (GDS)
• D. NVLink Bridge
• E. Remote Direct Memory Access (RDMA)

A customer is expanding their HPE Private Cloud AI "Medium" configuration to support a new generative AI inferencing workload. They are concerned about network congestion and latency, as the AI workload is known to generate large, sudden bursts of traffic between the compute nodes and the storage system.
The solution uses NVIDIA Spectrum SN4700M switches for the AI interconnect.
Which feature of these switches is specifically designed to handle bursty traffic and prevent packet loss in a lossless Ethernet fabric?

• A. A fully shared buffer architecture that can dynamically absorb traffic bursts from any port.
• B. Integrated Silicon Root of Trust to validate the switch firmware integrity.
• C. The ability to route traffic based on application-layer metadata.
• D. Support for Fibre Channel over Ethernet (FCoE) encapsulation.