You are tasked with designing a highly available AI data center platform that can continue to operate smoothly even in the event of hardware failures. The platform must support both training and inference workloads with minimal downtime. Which architecture would best meet these requirements?

• A.Deploy a single, powerful GPU server with redundant power supplies and network interfaces
• B.Implement a distributed architecture with multiple GPU servers and a load balancer to distribute the workload
• C.Set up a warm standby system where another data center mirrors the primary one and is manually activated
• D.Use a cluster of CPU-based servers with RAID storage to ensure data redundancy and protection

Comment: *

Name: *

Email: *

Verification: *

What is a key consideration when virtualizing accelerated infrastructure to support AI workloads on a hypervisor-based environment?

• A.Enable vCPU pinning to specific cores
• B.Disable GPU overcommitment in the hypervisor
• C.Maximize the number of VMs per physical server
• D.Ensure GPU passthrough is configured correctly

Comment: *

Name: *

Email: *

Verification: *

Which of the following statements is true about GPUs and CPUs?

• A.GPUs are optimized for parallel tasks, while CPUs are optimized for serial tasks.
• B.GPUs have very low bandwidth main memory while CPUs have very high bandwidth main memory.
• C.GPUs and CPUs have the same number of cores, but GPUs have higher clock speeds.
• D.GPUs and CPUs have identical architectures and can be used interchangeably.

Comment: *

Name: *

Email: *

Verification: *

Your organization is planning to deploy an AI solution that involves large-scale data processing, training, and real-time inference in a cloud environment. The solution must ensure seamless integration of data pipelines, model training, and deployment. Which combination of NVIDIA software components will best support the entire lifecycle of this AI solution?

• A.NVIDIA RAPIDS + NVIDIA Triton Inference Server + NVIDIA NGC Catalog
• B.NVIDIA TensorRT + NVIDIA DeepStream SDK
• C.NVIDIA Triton Inference Server + NVIDIA NGC Catalog
• D.NVIDIA RAPIDS + NVIDIA TensorRT

Comment: *

Name: *

Email: *

Verification: *

The data center administrator is asked to deploy infrastructure to support training of a large natural language processing model with a billion parameters and they need the fastest method to train the model. What should the administrator recommend to support this?

• A.Single server with 4-8 NVIDIA GPUs.
• B.Two Node NVIDIA DGX Cluster.
• C.NVIDIA DGX SuperPOD.

Comment: *

Name: *

Email: *

Verification: *