Docs

GPU Compute Service

Overview

What is a GPU instance?

GPU instance overview

A GPU instance is a standard ECS instance equipped with an additional GPU accelerator card. While delivering exceptional computing power, it retains the key advantages of standard ECS instances, such as on-demand provisioning and flexible usage. It is suitable for scenarios including AI training, AI inference, scientific computing, and video rendering.

GPU instances offer the same usage and management methods as standard ECS instances. For example, you can integrate them with services like Virtual Private Cloud (VPC), security groups, key pairs, and load balancing to build versatile business architectures. This enables the flexible and convenient purchase and management of GPU instances.

In addition to standard GPU instances, the platform also provides HPC clusters. Built on the basis of existing GPU instance type and enhanced with an RDMA network, these clusters significantly boost network performance and improve the acceleration ratio of large-scale clusters. They are ideal for business scenarios such as high-performance computing, artificial intelligence, and machine learning.

Access methods

You can access, create, and manage GPU instances via the web console or API operations based on HTTP/HTTPS requests:

Management console

You can log in to the ECS console to manage your resources on the web page.

You can use HTTP or HTTPS protocols for request communication to make online API calls or perform custom development. For more information, refer to API Reference.

Last updated: 2025.11.21 16:58:03