AI Infra Interviews logo
Hardware, Cabling & Cluster Build-Out / 13
hardNewNVIDIACoreWeaveMeta

After a maintenance window, two-node bandwidth halved. Single-node is unchanged. Diagnose it.

Single-node fine and two-node halved is a specific signature, and it points at one thing before anything else. Why the copy path costs exactly about half, the five causes that produce it, and the one log line that confirms which.

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.

Single-node fine and two-node halved is a specific signature, and it points at one thing before anything else. Why the copy path costs exactly about half, the five causes that produce it, and the one log line that confirms which.

more free answers with an account · no card

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.

Foundational
🖧 Hardware & Cluster Build-Out
Cluster Bring-Up: Firmware, Drivers and the StackBring-up is an ordered dependency chain and skipping a step produces a symptom that points somewhere else. Firmware first, then the operating system and kernel, then the GPU driver, then the fabric manager, then the network stack, then GPUDirect, then CUDA and NCCL, then the container and scheduling layer. On Blackwell HGX systems the fabric manager reaches the NVSwitches through a bridge device and therefore depends on the InfiniBand stack being present, which is a dependency that surprises almost everyone the first time.
Core
🔌 Networking & StorageSign in
RDMA, InfiniBand and RoCEv2Training across nodes moves hundreds of gigabytes per step, and a CPU-driven TCP stack cannot feed a 400 Gb/s link. RDMA lets a NIC write straight into a remote GPU's memory with no kernel and no copies, and it runs over two fabrics: InfiniBand, which is lossless by design, and RoCEv2, which is Ethernet made lossless by configuration. The choice is operational as much as technical, and the numbers that decide it are per-GPU bandwidth, the collective's volume, and who will debug a pause storm at 3 a.m.
Advanced
🔌 Networking & Storage🔒 Premium
GPUDirect RDMA and GPUDirect StorageBy default a byte leaving a GPU for the network or the disk makes a detour through host memory, crossing PCIe twice and costing a CPU copy. GPUDirect RDMA lets the NIC read and write GPU memory directly, and GPUDirect Storage does the same for NVMe. The win is not raw bandwidth (PCIe is the ceiling either way) but the halving of PCIe traffic and the removal of the host as a bottleneck, which is what makes collectives run at NIC rate and checkpoints run at drive rate. When it is silently off, everything still works, at half speed.
Advanced
🕸️ Distributed Training🔒 Premium
Ring vs Tree All-ReduceA ring all-reduce moves the minimum possible bytes per rank but takes 2(N-1) steps, so its latency grows with the number of GPUs; a tree finishes in a logarithmic number of steps but is harder to keep bandwidth-optimal. NCCL keeps both, chooses per message size and rank count, and reading its choice is how you diagnose a collective that is slower than the fabric allows.
UP NEXT ON YOUR JOURNEY
FEDITOR'S NOTE

Scored on reading the signature rather than starting a general investigation, on knowing why the host bounce roughly halves bandwidth, and on the NCCL log line and module checks that identify the cause.

DISCUSSION · 0

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