Skip to navigationSkip to main contentSkip to footerScaleway DocsAsk our AI
Ask our AI
Log inSign up

Scaleway GPU Instances internet and Block Storage bandwidth overview

Scaleway GPU Instances are designed to deliver high-performance computing for AI/ML workloads, rendering, scientific simulations, and visualization tasks. This guide provides a detailed overview of their internet and Block Storage bandwidth capabilities to help you choose the right Instance for your GPU-powered workloads.

Why bandwidth matters for GPU Instances

GPU workloads often involve processing large datasets, requiring high-bandwidth networking and storage access.

  • Internet bandwidth: Determines how fast your Instance can send or receive data from external sources, enabling smooth data transfers for model training or streaming.
  • Block bandwidth: Impacts data access speeds for datasets stored in Block Storage, optimizing performance for AI pipelines, rendering workflows, and other data-intensive applications.
Important

Bandwidth specifications listed here are for informational purposes only. To validate the exact bandwidth capacity of your GPU Instance, refer to the block_bandwidth field in the Instances API. Use the endpoint /instance/v1/zones/{zone}/products/servers to retrieve GPU Instance specifications.

Tip

To maximize compatibility with Block Storage Low Latency 15k, select an Instance with at least 3 GiB/s of Block bandwidth.

GPU Range

Instance TypeInternet Bandwidth (up to)Block Bandwidth (up to)
H100-SXM-2-80G20 Gbit/s5 GiB/s
H100-SXM-4-80G20 Gbit/s5 GiB/s
H100-SXM-8-80G20 Gbit/s5 GiB/s
H100-1-80G10 Gbit/s2 GiB/s
H100-2-80G20 Gbit/s4 GiB/s
L40S-1-48G2.5 Gbit/s1 GiB/s
L40S-2-48G5 Gbit/s1.5 GiB/s
L40S-4-48G10 Gbit/s2.5 GiB/s
L40S-8-48G20 Gbit/s5 GiB/s
L4-1-24G2.5 Gbit/s1 GiB/s
L4-2-24G5 Gbit/s1.5 GiB/s
L4-4-24G10 Gbit/s2.5 GiB/s
L4-8-24G20 Gbit/s5 GiB/s
GPU-3070-S2.5 Gbit/s1 GiB/s
RENDER-S2 Gbit/s1 GiB/s
Still need help?

Create a support ticket
No Results