Section E · Reference

Domain Context

Vocabulary so you never sound lost. GPU internals, interconnects, precisions, the major accelerator lines, checkpoint formats, KV-cache plumbing, and cloud pricing tiers.

GPU anatomy — the parts you should be able to name

PartWhat it is
SM (Streaming Multiprocessor)The execution unit. H100 has 132 SMs. Kernels are launched as grids of thread blocks; each block runs on one SM.
CUDA coreAn FP32 ALU. Marketing number — "how many CUDA cores" is misleading; SM count and clock matter more.
Tensor coreSpecialized matmul unit. Operates on small tiles (e.g. 16×16) in mixed precision. The reason GPUs are dominant for transformers.
WarpGroup of 32 threads executing in lockstep. SIMT execution model.
Register filePer-SM fast storage. Allocation per thread limits warps in flight.
Shared memory / L1Per-SM scratchpad, programmer-managed. ~228 KB on H100.
L2 cachePer-GPU, shared across SMs. 50-60 MB on H100.
HBM (High-Bandwidth Memory)The big main memory. H100 SXM has 80 GB at ~3 TB/s; H200 has 141 GB at 4.8 TB/s.
NVLinkNVIDIA's GPU-GPU interconnect. H100 SXM has 18 NVLink-4 connections totaling 900 GB/s aggregate.
NVSwitchSwitch fabric inside multi-GPU nodes for all-to-all NVLink connectivity.
Transformer EngineNVIDIA's mixed-precision (FP8/FP16/BF16) feature on Hopper+, integrated with cuDNN and frameworks.

Numeric precisions you'll discuss

FormatBitsTypical use
FP6464Scientific HPC; rarely used in deep learning
FP3232Single precision; default in earlier ML
TF3232 (19 effective)NVIDIA Ampere+ "tensor-friendly FP32" — drop-in faster training
FP1616Mixed-precision training (historic), inference baseline
BF1616Wider exponent than FP16; preferred for training stability
FP8 (E4M3, E5M2)8Hopper+/Blackwell; serious throughput win, modest quality cost
INT88Quantized inference (SmoothQuant, GPTQ, AWQ)
INT44Aggressive inference quantization; meaningful quality cost
NVFP4 / FP44Blackwell-era; bleeding edge
The interview answer

"Default to BF16 for training stability, FP8 for inference where the hardware supports it (Hopper+). Below FP8, evaluate the quality cost case-by-case; expect 1-4% degradation depending on the quantization technique and the model."

Interconnect & fabric

  • PCIe Gen4 x16 — ~32 GB/s. Common for A100 PCIe and lower-tier deployments.
  • PCIe Gen5 x16 — ~64 GB/s. H100 PCIe form factor and modern hosts.
  • NVLink-4 — used in H100 SXM, 900 GB/s aggregate per GPU.
  • NVLink-5 — Blackwell era, 1.8 TB/s aggregate per GPU.
  • NVSwitch — full-bandwidth switching between GPUs in a chassis.
  • InfiniBand HDR — 200 Gbps. Older fabric standard.
  • InfiniBand NDR — 400 Gbps per port. Current default for big training.
  • RoCE (RDMA over Converged Ethernet) — RDMA semantics on Ethernet; 200/400/800 Gbps available.
  • GPUDirect RDMA — NIC writes directly to GPU memory, bypassing the host. Required for IB at full speed.
  • NIXL / NCCL — software layers on top.

Form factors

  • SXM — module form factor, higher power envelope (700W+ for H100 SXM5), full NVLink. Found in HGX/DGX systems.
  • PCIe — card form factor, lower power (350W typical for H100 PCIe), NVLink bridges between pairs only. More flexible for standard server chassis.
  • SXM beats PCIe for big-model training and tensor-parallel serving because of full NVLink. For inference of smaller models or workloads with little inter-GPU traffic, PCIe is fine.

The NVIDIA accelerator line — what's on the menu

GenerationArchitectureFlagshipHBMHighlights
AmpereGA100A100 (40/80 GB)HBM2eTF32, MIG, NVLink-3. Still huge in production.
HopperGH100H100 (80 GB)HBM3FP8, Transformer Engine, NVLink-4, DPX instructions.
Hopper-refreshGH200H200 (141 GB)HBM3eBigger HBM & bandwidth — wins on long-context inference.
BlackwellGB100/GB200B100, B200, GB200 NVL72HBM3eFP4, 2nd-gen Transformer Engine, NVLink-5. Multi-chip module.
Ada LovelaceAD102L40, L40SGDDR6Mid-tier inference; no HBM, but cheaper.

Non-NVIDIA accelerators worth recognizing

  • Google TPU v5p / v5e / v6 (Trillium) — pod-based systolic-array architecture. JAX/PyTorch XLA. Strong for big training.
  • AWS Trainium / Trainium2 — AWS's custom training chip. Neuron SDK. Price-competitive for AWS-native workloads.
  • AWS Inferentia2 — inference-only sibling.
  • Intel Gaudi 2 / Gaudi 3 — Habana labs. HPU. Competitive on certain LLM training workloads, with integrated networking.
  • AMD MI300X / MI325X — large HBM (192-256 GB), ROCm software stack. Maturing; competitive for inference of big-context workloads.
  • Apple Silicon / Cerebras / Groq / SambaNova — niche or specialized.

You won't be deployed on all of them, but say their names confidently when asked.

KV-cache, prefix caching, page tables

The plumbing language specific to LLM serving:

  • KV-cache — per-layer K and V projections retained on GPU memory for an active sequence. The reason decode is fast.
  • Paged attention — vLLM's idea of treating KV-cache as virtual memory: fixed-size blocks in HBM, with a block table (page table) per sequence mapping logical positions to physical blocks.
  • Block size — typically 16 tokens; tunable. Smaller = less fragmentation, more table overhead.
  • Prefix caching — when two requests share a prefix (e.g. the same system prompt), point both sequences' early block-table entries at the same physical blocks. Copy-on-write when they diverge.
  • Chunked prefill — split a long prompt into chunks and process them across multiple iterations alongside ongoing decodes, so big prompts don't starve other requests.
  • Swap to CPU / disk — vLLM can spill blocks to host memory when GPU memory is tight; trades latency for capacity.
  • KV-cache quantization — store cache in FP8 / INT8 to double effective capacity at small quality cost.

Checkpoint formats

  • safetensors — Hugging Face's safer alternative to .pt. No pickle deserialization; bounded memory mapping. The de facto serving format.
  • pickle / .pt / .pth — original PyTorch format; opens arbitrary code execution surface. Avoid for untrusted inputs.
  • GGUF — llama.cpp's quantized format. Common for CPU/edge inference, less so for cluster serving.
  • TensorRT plan files — compiled engines, GPU-architecture-specific. Not portable.
  • Sharded checkpoints — PyTorch DCP (Distributed Checkpoint), DeepSpeed shards, FSDP shards. One file per rank; metadata describes how to reassemble.
  • Megatron checkpoint format — used by Megatron-LM with tensor + pipeline parallelism.

Cloud accelerator pricing tiers

You should be able to discuss these without specific numbers — the numbers change.

  • On-demand — pay-per-second, highest unit cost, no commitment.
  • Spot / preemptible — significant discount (50-80%), interruptible.
  • Savings plans (AWS) / committed-use discounts (GCP) — commit to N hours per month for 1-3 years for a discount. Applies across instance types in the same family.
  • Reserved instances — capacity reservation tied to a specific region/AZ/family. Stronger capacity guarantee than savings plans.
  • Capacity blocks (AWS) — pre-purchase GPU-hours in a future contiguous window. Designed for training runs.
  • Dedicated hosts — entire physical server; useful for licensing or strict isolation.
  • On-prem amortization — bought hardware spread across its useful life (typically 3-4 years for GPUs). Add power, cooling, datacenter, ops loaded cost. Compare to cloud blended cost.
Blended cost model

Real cost = (reserved capacity fixed cost / hours used) + (on-demand burst hours × $/hour) + (spot hours × spot rate) + observability overhead + control-plane overhead. The "$/GPU-hour" question is meaningless without that decomposition.

Quick glossary — the rest

  • CUDA — NVIDIA's parallel compute platform and language.
  • cuDNN — accelerated primitives for deep learning.
  • NCCL — NVIDIA Collective Communications Library; all-reduce, broadcast, all-gather across GPUs.
  • UCX / NVSHMEM — lower-level communication libraries.
  • cuBLAS / cuBLASLt — matrix-multiply libraries.
  • NVRTC — runtime CUDA compilation.
  • DCGM — Data Center GPU Manager; telemetry, diagnostics, health.
  • MIG — Multi-Instance GPU partitioning.
  • MPS — Multi-Process Service for GPU sharing.
  • UVM — Unified Virtual Memory between host and device.
  • PTX / SASS — NVIDIA assembly languages (PTX virtual, SASS native).
  • RDMA — Remote Direct Memory Access; bypass kernel for network transfers.
  • RoCE — RDMA over Converged Ethernet.
  • QSFP — connector form factor for high-speed networking.
  • TFLOPS — Tera-FLOPS; vendor headline number, usually peak FP at a precision.
  • Token — unit of LLM input/output; a piece of a word in the model's tokenizer.
  • TTFT / TPOT / ITL — time-to-first-token, time-per-output-token, inter-token-latency.
  • SGLang / TGI / LMDeploy — alternative LLM serving stacks worth knowing exist.