← NVIDIA Interview Insights

NVIDIA·Software Engineer·Technical Phone Screen·Intermediate

Intermediate
Apr 2026

Summary

Interviewed for a SWE role at Nvidia and got hit with a GPU architecture question pretty early on. Not the most elaborate process to describe but the technical depth required was real.

Questions Asked (1)

Q1

How does a GPU work? Walk through the architecture and what makes it different from a CPU.

System DesignTechnical Trade-offs
Author's notes

This is Nvidia so you'd think I'd have prepared harder for this.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by contrasting the CPU's latency-optimized design with the GPU's throughput-optimized architecture, then walk through the GPU pipeline from thread scheduling to memory hierarchy. Use concrete examples like matrix multiplication to illustrate how massive parallelism and SIMT execution enable high throughput.

Pro tip: Emphasize that GPUs are not just 'more cores' but a fundamentally different design philosophy: CPUs minimize latency for serial tasks, GPUs maximize throughput for parallel tasks. Mention that NVIDIA's success comes from co-designing hardware and software (CUDA) to make this paradigm accessible.

1. Define the Core Design Philosophies

Explain that CPUs are latency-optimized with large caches, out-of-order execution, and few powerful cores, while GPUs are throughput-optimized with many simple cores and massive parallelism.

2. Describe GPU Architecture Components

Cover Streaming Multiprocessors (SMs), each containing many CUDA cores, warp schedulers, register files, and shared memory/L1 cache. Mention the global memory (DRAM) and how data flows through the hierarchy.

3. Explain Execution Model (SIMT)

Detail how threads are grouped into warps (32 threads) that execute in lockstep, and how the warp scheduler hides latency by switching between warps. Contrast with CPU's out-of-order execution.

4. Highlight Memory Hierarchy and Bandwidth

Discuss the importance of coalesced memory access, the role of shared memory as a programmer-managed cache, and how high bandwidth (e.g., HBM) feeds the many cores.

5. Summarize Trade-offs and Use Cases

Conclude that GPUs excel at data-parallel tasks (graphics, deep learning) but are poor for serial or branch-heavy code. Mention that modern systems often use both CPU and GPU together (heterogeneous computing).

Key Points to Mention

  • SIMT (Single Instruction, Multiple Threads) execution model and warp scheduling
  • Streaming Multiprocessor (SM) architecture with many CUDA cores and shared memory
  • Memory hierarchy: registers, shared memory/L1, L2, and global memory (DRAM)
  • Latency hiding via massive thread-level parallelism (TLP) rather than instruction-level parallelism (ILP)
  • Throughput vs. latency optimization: CPU for serial tasks, GPU for parallel tasks
  • Coalesced memory access and the importance of memory bandwidth

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.