Skip to content
Launch imminent

Official date to be announced shortlyGPU offers available at launch; figures shown are indicative.

Priority access: sign up

What is NVSwitch? All-to-all GPU links explained

NVSwitch connects every GPU in a server over NVLink at full speed. See why 8-GPU HGX systems use it for large-model training.

Glossary3 sections1 min read

What is NVSwitch?

NVSwitch is a switch chip that connects many GPUs over NVLink so that every GPU can talk to every other one at full NVLink speed, instead of only to its direct neighbors.

Why it matters when renting a GPU

In an 8-GPU server, all-to-all traffic (for example gradient exchange) would bottleneck on point-to-point links alone. NVSwitch removes that bottleneck.

When you rent a multi-GPU machine for training, ask whether the GPUs are SXM boards on an NVSwitch baseboard or separate PCIe cards.

Worked example

With NVSwitch, each H100 SXM in an 8-GPU server keeps its full NVLink 4 (900 GB/s) to the others, so the 8 GPUs pool 8 × 80 = 640 GB of HBM for one sharded model.

Frequently asked questions

Is NVSwitch the same as NVLink?
No. NVLink is the link; NVSwitch is the switch that connects many NVLinks so all GPUs reach each other at full speed.

Be ready on launch day

Create your account now: you can top up your balance and launch a GPU as soon as the market opens.