NCA-AIIO Practice Questions

A.  

A GPU acts as a PCIe controller to maximize bandwidth.

B.  

A GPU is architected to support massively parallel execution of simple instructions.

C.  

A GPU is a single large and complex core to support massive compute operations.

A.  

To gather and analyze cluster data on a regular schedule.

B.  

To monitor and troubleshoot cluster performance.

C.  

To assign workloads to available compute resources.

D.  

To install, update, and configure cluster software.

A.  

Lower upfront costs and capital expenditure.

B.  

Scalability and flexibility.

C.  

Ensure data security and sovereignty.

D.  

Easy remote management.

A.  

Recommender systems

B.  

Natural language processing

C.  

Data analytics

A.  

It is used to remotely manage and troubleshoot network devices independently of the production network.

B.  

It is used to directly manage the AI model’s learning rate during training sessions.

C.  

It is used to increase the computational power of AI models by adapting additional processing resources.

D.  

It is used to manage the data throughput of AI applications by prioritizing network traffic.

A.  

7

B.  

8

C.  

4

A.  

The maximum amount of time a GPU will be used for a workload.

B.  

The GPU memory in use compared to available GPU memory.

C.  

The percentage of time the GPU is actively processing data.

D.  

The number of GPU cores available to the workload.

A.  

NVLink

B.  

GPUDirect P2P

C.  

InfiniBand

D.  

GPUDirect Storage

A.  

Training requires real-time processing, while inference requires large amounts of data.

B.  

Training requires large amounts of data, while inference requires real-time processing.

C.  

Training and inference both require large amounts of data.

D.  

Training and inference both require real-time processing.

A.  

3

B.  

2

C.  

4

D.  

5