Docs
gpu
Model and Performance
Spot GPU Cloud Server

Spot GPU Cloud Servers

Product Introduction

Spot GPU cloud servers offer a super low discount compared to the pay-as-you-go GPU cloud servers, aiming to provide customers with more convenient low-cost GPU computing resources in some scenarios.

The list price of spot GPU cloud servers will change with market prices and stock balances. If backend resources are insufficient, or the customer’s bid is lower than the current list price, the customer will be notified by email, and the resources will be reclaimed 30min later.

Applicable Scenarios

Spot GPU cloud servers are suitable for stateless, checkpoint-restartable application scenarios, such as image rendering, high-performance parallel computing, big data businesses, etc. The more distributed, scalable, and fault-tolerant the application program, the more suitable it is to use spot GPU cloud servers to save resource costs and increase throughput.

Stability-demanding services should not run on spot GPU cloud servers because factors such as insufficient system resources or failed bidding may cause spot cloud servers to be released, interrupt tasks, and possibly affect normal business operations.

Life Cycle

After creation, the spot GPU cloud server can run smoothly for one hour, and during this hour, it is not affected by the lack of system resources or price adjustments, and can be used smoothly.

If backend resources are not sufficient or the market price is higher than the customer’s bid, the system will notify the customer by email that the spot GPU cloud server will be released 30min later. Please backup your data in advance.

When creating a spot GPU cloud server, you can also turn on the snapshot service. If you need to continue using it, you can restore the GPU cloud server through the snapshot.

Billing Instructions

Currently, only cloud servers support spot bidding, that is, CPU, memory, and GPU card can enjoy ultra-low discount, cloud servers associated with EIP, disks are still sold at the original price;

If the resources of the spot GPU cloud server are in arrears, they will immediately enter the recovery list and notify the customer by email to renew the fee immediately. If the fee is not renewed after 5 minutes, the resources will be shut down and recovered;

The pricing method of spot GPU cloud servers cannot be changed, and upgrade modifications are not supported;

The spot GPU cloud server is not affected by customer sales discounts and cannot enjoy the stack discount.

Supported Machine Configuration & Region

RegionMachine & Card TypeConfiguration
North China 2AGPU type cost-effective graphics card 516 core 32G GPU* 1
North China 2AGPU type cost-effective graphics card 516 core 64G GPU* 1
North China 2AGPU type cost-effective graphics card 532 core 64G GPU* 2
North China 2AGPU type cost-effective graphics card 532 core 128G GPU* 2
North China 1B/Shanghai 2AGPU type V100S10 core 32G GPU* 1
North China 1B/Shanghai 2AGPU type V100S10 core 64G GPU* 1
North China 1B/Shanghai 2AGPU type V100S10 core 96G GPU* 1
North China 1B/Shanghai 2AGPU type V100S20 core 64G GPU* 2
North China 1B/Shanghai 2AGPU type V100S20 core 96G GPU* 2
North China 1B/Shanghai 2AGPU type V100S20 core 128G GPU* 2
North China 1B/Shanghai 2AGPU type V100S20 core 160G GPU* 2
North China 1B/Shanghai 2AGPU type V100S20 core 160G GPU* 2
North China 1B/Shanghai 2BGPU type P404 core 8G GPU* 1
North China 1B/Shanghai 2BGPU type P404 core 16G GPU* 1
North China 1B/Shanghai 2BGPU type P408 core 16G GPU* 1
North China 1B/Shanghai 2BGPU type P408 core 32G GPU* 1
North China 1B/Shanghai 2BGPU type P408 core 64G GPU* 1
North China 1B/Shanghai 2BGPU type P408 core 16G GPU* 2
North China 1B/Shanghai 2BGPU type P408 core 32G GPU* 2
North China 1B/Shanghai 2BGPU type P4016 core 32G GPU* 2
North China 1B/Shanghai 2BGPU type P4016 core 64G GPU* 2
North China 1BGPU type T4S4 core 8G GPU* 1
North China 1BGPU type T4S4 core 16G GPU* 1
North China 1BGPU type T4S4 core 32G GPU* 1
North China 1BGPU type T4S8 core 16G GPU* 1
North China 1BGPU type T4S8 core 32G GPU* 1
North China 1BGPU type T4S8 core 64G GPU* 1
North China 1BGPU type T4S8 core 16G GPU* 2
North China 1BGPU type T4S8 core 32G GPU* 2
North China 1BGPU type T4S8 core 64G GPU* 2
North China 1BGPU type T4S8 core 128G GPU* 2