NCA GPU Architecture & Fundamentals 2

Question 1

What is the primary role of the L1 cache in NVIDIA GPU Streaming Multiprocessors (SMs)?

Accepted Answer

Provide low-latency data access for threads within an SM

Answer

The L1 cache in each SM provides fast, low-latency access to frequently used data for threads executing within that SM.

Question 2

In NVIDIA GPU architecture, what does the term 'warp' refer to?

Accepted Answer

A group of 32 threads that execute the same instruction simultaneously

Answer

A warp is the fundamental scheduling unit in NVIDIA GPUs, consisting of 32 threads that execute the same instruction in lockstep (SIMT).

Question 3

Which memory type on an NVIDIA GPU is shared among all threads within a thread block and has configurable size per SM?

Accepted Answer

Shared memory

Answer

Shared memory is an on-chip scratchpad memory partitioned per SM, accessible by all threads in a block with very low latency.

Question 4

What is 'warp divergence' and what is its performance impact?

Accepted Answer

When threads in a warp take different branches, causing serial execution of each branch

Answer

Warp divergence occurs when threads in a warp follow different code paths (e.g., if/else), forcing the GPU to serialize execution of each path, reducing throughput.

Question 5

What is the function of the GigaThread Engine in NVIDIA GPU architecture?

Accepted Answer

Distribute thread blocks across Streaming Multiprocessors

Answer

The GigaThread Engine is a global scheduler responsible for distributing thread blocks to available SMs across the GPU.

(NCA) NVIDIA-Certified Associate Practice Test

(NCA) NVIDIA-Certified Associate Practice Test

NCA GPU Architecture & Fundamentals 2