Frequently Asked
Questions
A GPU has more arithmetic logic units (ALUs) than a CPU and supports large-scale multi-thread parallel computing. It is most suitable for the following use cases:
- AI computing: Deep learning inference and training.
- Graphics and image processing: Cloud game, cloud phone, cloud desktop, and CloudXR.
- High-performance computing: Fluid dynamics, molecular modeling, meteorological engineering, seismic analysis, genomics, etc.
You need to select an instance model based on your use case:
- AI training: GN10Xp, GN10X, GT4, GN8, and GN6/GN6S
- AI inference: GN7, GN10Xp, GN10X, PNV4, GI3X, GN6, and GN6S
- Graphics and image processing: GN7vw, GNV4, GNV4v, and GI1
- Scientific computing: GN10Xp, GN10X, GT4, and GI3X
- For more information, see Computing Instance and Rendering Instance.
NVIDIA GPU instance models include physical passthrough instances (having entire GPUs) and vGPU instances (having no entire GPUs, such as 1/4 GPU).
The GPU on a physical passthrough instance can use the Tesla or GRID driver (several models don't support the GRID driver) to accelerate computing in different scenarios.
A vGPU can use only the GRID driver on certain versions to accelerate computing.
For detailed directions on how to install a driver on an NVIDIA GPU instance, see Installing NVIDIA Tesla Driver.
GPU instance models PNV4, GT4, GN10X, GN10Xp, GN6, GN6S, GN7, GN8, GNV4v, GNV4, GN7vw, and GI1 support instance configuration adjustment in the same instance family, while GI3X does not support instance configuration adjustment.
You can try the following:
- Change the region
- Change the AZ
- Change the resource configuration
If the problem persists, please contact us.
GPU instances are pay-as-you-go, where the instances are billed by second and settled hourly. You can purchase or release the instances at any time. For more information, see Pricing Overview.
When you purchase GPU instances, you can check the estimated price in the Price Calculator.
A private IP is a connection address that provides services for a client with a source IP from the private network. A public IP is a connection address that enables public network communication for a client with a source IP from the public network. They can be directly mapped to each other through network address translation. GPU instances in the same region can communicate over the private network, while those in different regions can only communicate over the public network.
An EIP is a static IP specifically designed for dynamic cloud computing. It is region-specific. You can quickly remap an EIP to another GPU instance (or a CVM/NAT gateway instance) under your account to block instance failures. For more information, see EIP.
Tencent Cloud provides different types of data storage options for GPU instances, including cloud disk, local disk, COS, and block storage device mapping. Different options differ in performance and price and are suitable for different use cases. For more information, see Storage Overview.
Some of the CPM GPU instance models support local storage, and you can use remote storage as needed.
1. Local storage
Certain CPM GPU instances are equipped with NVMe SSD disks with excellent read/write performance, which is three times higher than that of a general model, so as to guarantee the stability of high-performance computing jobs.
2. Remote storage
CFS: You can mount Turbo CFS through the smart ENI technology, which enables flexible expansion of storage capacity and guarantees strong consistency between three replicas.
COS: COS works with the distributed cluster architecture of GooseFS to improve the data locality and uses the high-speed cache feature to improve the storage performance and increase the bandwidth for writing data to COS.
1. If your GPU instance uses a cloud disk, you can back up your business data by creating a custom system disk image and a data disk snapshot
- For more information on how to create a custom image, see Creating Custom Image.
- For more information on how to create a snapshot, see Creating Snapshot.
2. If your GPU instance uses a local disk, you can back up your business data by creating a custom system disk image. However, you need to customize backup policies for the business data in your data disk. Usually, you can back up the data in your GPU instance over FTP. For more information on FTP deployment, see the following documents:
- For Windows: Setting up an FTP Site (Windows).
- For Linux: Setting up an FTP Site (Linux).
3. In addition, if you require high data security, you can purchase more professional third-party custom backup services.
Tencent Cloud regions are completely isolated. This guarantees the maximum cross-region stability and fault tolerance. We will gradually deploy nodes in more regions for a wider coverage. We recommend you select the region closest to your end users to minimize the access latency and improve the download speed. For more information on supported regions, see Regions and AZs.
We recommend you select the region closest to your users, and choose the same region for your GPU instances so that communication can occur over the private network.
- Close to your user: By selecting a region that is geographically close to your users, you can reduce their access latency and increase their access speed. For example, if most of your users are located near Southeast Asia, Singapore or Thailand will be a good choice.
- Intra-region communication: GPU instances in the same region can communicate over private network free of charge. When they are in different regions, they can only communicate over the public network, which incurs fees. To enable instances to communicate over the private network, you must select the same region for them.
Each AZ runs on its own independent and physically distinct infrastructure and is designed to be highly reliable. AZs do not share common vulnerable equipment such as power supplies and cooling equipment. In addition, they are physically independent of each other, so if a natural disaster such as a fire, tornado, or flood occurs, only the AZ at that location would be affected while other AZs remain operational.
Tencent Cloud provides various network and security services such as security groups, encrypted login, and EIP to ensure the secure, efficient, and free operations of your instances. For more information on GPU security, see Network and Security Overview.
You can control the access to your GPU instances by adding them into a security group. You can also configure communication across security groups and specify which IP subnets can communicate with your instances.
When security risks are detected, you can troubleshoot the issues as instructed in Security Violation Handling Guide and resolve it as instructed in Cloud Workload Protection.
If you select a vGPU or Rendering instance, you can select an image with a preinstalled GRID driver in “Public images” on the purchase page.
Tencent Cloud provides public images, shared images, and custom images. For more information, see Image Types. You can select appropriate images based on their different features.
You can share a custom image with other users, or obtain images shared by others. For more information on the limits and usage of shared images, see Sharing Custom Images.
An image can be shared with up to 50 users. Shared images do not count towards your own image quota.
Frequently Asked
Questions
A GPU has more arithmetic logic units (ALUs) than a CPU and supports large-scale multi-thread parallel computing. It is most suitable for the following use cases:
- AI computing: Deep learning inference and training.
- Graphics and image processing: Cloud game, cloud phone, cloud desktop, and CloudXR.
- High-performance computing: Fluid dynamics, molecular modeling, meteorological engineering, seismic analysis, genomics, etc.
You need to select an instance model based on your use case:
- AI training: GN10Xp, GN10X, GT4, GN8, and GN6/GN6S
- AI inference: GN7, GN10Xp, GN10X, PNV4, GI3X, GN6, and GN6S
- Graphics and image processing: GN7vw, GNV4, GNV4v, and GI1
- Scientific computing: GN10Xp, GN10X, GT4, and GI3X
- For more information, see Computing Instance and Rendering Instance.
NVIDIA GPU instance models include physical passthrough instances (having entire GPUs) and vGPU instances (having no entire GPUs, such as 1/4 GPU).
The GPU on a physical passthrough instance can use the Tesla or GRID driver (several models don't support the GRID driver) to accelerate computing in different scenarios.
A vGPU can use only the GRID driver on certain versions to accelerate computing.
For detailed directions on how to install a driver on an NVIDIA GPU instance, see Installing NVIDIA Tesla Driver.
GPU instance models PNV4, GT4, GN10X, GN10Xp, GN6, GN6S, GN7, GN8, GNV4v, GNV4, GN7vw, and GI1 support instance configuration adjustment in the same instance family, while GI3X does not support instance configuration adjustment.
You can try the following:
- Change the region
- Change the AZ
- Change the resource configuration
If the problem persists, please contact us.
GPU instances are pay-as-you-go, where the instances are billed by second and settled hourly. You can purchase or release the instances at any time. For more information, see Pricing Overview.
When you purchase GPU instances, you can check the estimated price in the Price Calculator.
A private IP is a connection address that provides services for a client with a source IP from the private network. A public IP is a connection address that enables public network communication for a client with a source IP from the public network. They can be directly mapped to each other through network address translation. GPU instances in the same region can communicate over the private network, while those in different regions can only communicate over the public network.
An EIP is a static IP specifically designed for dynamic cloud computing. It is region-specific. You can quickly remap an EIP to another GPU instance (or a CVM/NAT gateway instance) under your account to block instance failures. For more information, see EIP.
Tencent Cloud provides different types of data storage options for GPU instances, including cloud disk, local disk, COS, and block storage device mapping. Different options differ in performance and price and are suitable for different use cases. For more information, see Storage Overview.
Some of the CPM GPU instance models support local storage, and you can use remote storage as needed.
1. Local storage
Certain CPM GPU instances are equipped with NVMe SSD disks with excellent read/write performance, which is three times higher than that of a general model, so as to guarantee the stability of high-performance computing jobs.
2. Remote storage
CFS: You can mount Turbo CFS through the smart ENI technology, which enables flexible expansion of storage capacity and guarantees strong consistency between three replicas.
COS: COS works with the distributed cluster architecture of GooseFS to improve the data locality and uses the high-speed cache feature to improve the storage performance and increase the bandwidth for writing data to COS.
1. If your GPU instance uses a cloud disk, you can back up your business data by creating a custom system disk image and a data disk snapshot
- For more information on how to create a custom image, see Creating Custom Image.
- For more information on how to create a snapshot, see Creating Snapshot.
2. If your GPU instance uses a local disk, you can back up your business data by creating a custom system disk image. However, you need to customize backup policies for the business data in your data disk. Usually, you can back up the data in your GPU instance over FTP. For more information on FTP deployment, see the following documents:
- For Windows: Setting up an FTP Site (Windows).
- For Linux: Setting up an FTP Site (Linux).
3. In addition, if you require high data security, you can purchase more professional third-party custom backup services.
Tencent Cloud regions are completely isolated. This guarantees the maximum cross-region stability and fault tolerance. We will gradually deploy nodes in more regions for a wider coverage. We recommend you select the region closest to your end users to minimize the access latency and improve the download speed. For more information on supported regions, see Regions and AZs.
We recommend you select the region closest to your users, and choose the same region for your GPU instances so that communication can occur over the private network.
- Close to your user: By selecting a region that is geographically close to your users, you can reduce their access latency and increase their access speed. For example, if most of your users are located near Southeast Asia, Singapore or Thailand will be a good choice.
- Intra-region communication: GPU instances in the same region can communicate over private network free of charge. When they are in different regions, they can only communicate over the public network, which incurs fees. To enable instances to communicate over the private network, you must select the same region for them.
Each AZ runs on its own independent and physically distinct infrastructure and is designed to be highly reliable. AZs do not share common vulnerable equipment such as power supplies and cooling equipment. In addition, they are physically independent of each other, so if a natural disaster such as a fire, tornado, or flood occurs, only the AZ at that location would be affected while other AZs remain operational.
Tencent Cloud provides various network and security services such as security groups, encrypted login, and EIP to ensure the secure, efficient, and free operations of your instances. For more information on GPU security, see Network and Security Overview.
You can control the access to your GPU instances by adding them into a security group. You can also configure communication across security groups and specify which IP subnets can communicate with your instances.
When security risks are detected, you can troubleshoot the issues as instructed in Security Violation Handling Guide and resolve it as instructed in Cloud Workload Protection.
If you select a vGPU or Rendering instance, you can select an image with a preinstalled GRID driver in “Public images” on the purchase page.
Tencent Cloud provides public images, shared images, and custom images. For more information, see Image Types. You can select appropriate images based on their different features.
You can share a custom image with other users, or obtain images shared by others. For more information on the limits and usage of shared images, see Sharing Custom Images.
An image can be shared with up to 50 users. Shared images do not count towards your own image quota.