[GPU programming] Nvidia Software and Hardware Quiz

1.

질문 1

By comparison, integrated GPUs use less heat than dedicated GPUs and share system memory directly with the CPU.

True

Integrated GPUs are less powerful and use a lower wattage. Another feature of integrated GPUs is that they share system memory with the host machine’s CPU.

2.

질문 2

 

Nvidia GPUs

CUDA is developed for Nvidia hardware.

 

3.

질문 3

There is no way in the Windows OS to determine which type of GPUs are installed on the current machine.

False

There are numerous applications for Windows OS.

4.

질문 4

The general pace of new Nvidia GPU architectures is which of the following?

Every 2 years

New GPU architectures are released every two years.

5.

질문 5

Which measure is consistently improving each generation of Nvidia GPU?

GigaFLOPs per watt

It is Nvidia’s constant goal to lower wattage while maintaining or increasing the computational power, measured in Giga FlOating Point Operations (GigaFLOPs).

6.

질문 6

How can the communications between layers be characterized?

Applications and Libraries can communicate with Runtime and Driver APIs

This statement is correct, because different layers communicate bidirectionally.

 

7.

질문 7

Which of these is not a compiled formats of CUDA code?

.cu

Feedback: .cu is a file extension, which is the same as .cpp or .c except that it indicates that it holds CUDA code. Therefore, this is a correct answer. 

8.

질문 8

Using the nvcc command with the -h flag, does which of the following?

It shows help

 

9.

질문 9

The CUDA Framework’s Runtime API is written to be used directly in many programming languages.

False

This is the correct answer because CUDA Runtime API code needs to be written in C/C++.

 

10.

질문 10

Using the CUDA Driver API will require the developer to manage which of the following?

Contexts, devices, and programming modules

Feedback: Writing Driver API code requires developers to manage the three important components in order to execute code on Nvidia GPUs.