hardNewAnthropicFireworks

Estimate the latency of one decode step for a 70B model under tensor parallelism across eight H100s

Each card reads an eighth of the weights in 5.3 ms, then the step pays 160 latency-bound all-reduces and hundreds of kernel launches that do not shrink with sharding. The chain to a 9 to 12 ms step, the communication floor, and why TP8 gives 4x rather than 8x at batch one.

Updated Sep 2026 · Grounded in real AI infrastructure interview loops and written to a senior-engineer editorial bar, with every number worked and every diagram hand-built.

20 answers per topic instead of 10, plus saved progress and bookmarks · no cardor unlock all 283 remaining answers · ₹2,000 / $25

The concepts behind this question

Ranked by how closely each one overlaps this question's topic, so the first card is the thing to read if the answer above moved too fast.

Advanced

🕸️ Distributed Training🔒 Premium

Tensor ParallelismTensor parallelism splits individual weight matrices across GPUs so each rank computes a slice of every layer, which is how a model whose single layer does not fit one GPU gets trained at all. It costs four all-reduces per transformer block on the critical path, which is why it stays inside the NVLink domain and rarely exceeds 8 ranks.

Advanced

🧮 Napkin Math & Capacity🔒 Premium

Bandwidth-Bound Decode ThroughputBecause decode reads every weight once per step, its speed is a division: memory bandwidth over bytes per step. That one formula gives single-stream tokens per second for any model on any card, the batch curve that flattens at the ridge point, the effect of quantization, and the point where the KV cache rather than the weights becomes the thing being read. This page derives it, works it for a 70B model on four accelerators, and shows how to read a vendor throughput claim against it.

Foundational

🕸️ Distributed Training

Collective Communication PrimitivesAll-reduce, all-gather, reduce-scatter, all-to-all and broadcast are the five operations every parallelism strategy is built from, and each has a fixed per-rank traffic cost you can compute before a job runs. Knowing those volumes for a named model is how you decide whether a layout is compute-bound or waiting on the network.

Foundational

🔌 Networking & Storage

NCCL and Collective AlgorithmsNCCL is the library every PyTorch collective lands in, and its choice of ring or tree, channel count and protocol decides whether an all-reduce runs at fabric speed or at a third of it. Knowing what NCCL_DEBUG=INFO prints, and which environment variable changes which decision, is the difference between tuning a cluster and guessing at it.

Explore the full curriculum →

UP NEXT ON YOUR JOURNEY

Next in this trackHow big can the decode batch get before the step becomes compute-bound?Next in this trackA startup wants to train a 7B model on 1 trillion tokens in three weeks. What cluster do they rent?Next in this trackEstimate the activation memory for training a 70B model at an 8k sequence length

FEDITOR'S NOTE

The interviewer wants three terms with numbers: the sharded weight read, the per-layer all-reduce latency times the count, and launch overhead. Then the sentence that the communication term is a floor independent of batch.

DISCUSSION · 0

No comments yet — be the first to share your approach.