GPU rendering instances are designed for traditional GPU-based graphics and image processing use cases, such as 3D rendering. Tencent Cloud offers fast, stable, and auto scaling computing services managed consistently with CVM.
Applicable Scenarios
Suitable for high-performance graphics processing and 3D rendering. For example:
Non-linear editing
Game streaming
Cloud phone
Cloud virtual desktop
CloudXR
Graphics and image processing
Overview of Rendering Instances
The Cloud GPU Service rendering series offers the following instances:
Instance | GPU Types | Available Image | Available Regions |
GA01 | Windows Server 2019 Datacenter Edition 64-bit | Guangzhou | |
NVIDIA A10 | Windows Server 2019 Datacenter Edition 64-bit Chinese Edition with GRID 16.2 Windows Server 2022 Datacenter Edition 64-bit Chinese Edition with GRID 16.2 | Guangzhou, Shanghai, and Beijing | |
NVIDIA A10 | CentOS 7.2 and above Ubuntu 16.04 - 20.04 Windows Server 2019 Datacenter Edition 64-bit Chinese Edition with GRID 16.2 Windows Server 2022 Datacenter Edition 64-bit Chinese Edition with GRID 16.2 | Guangzhou, Shanghai, Beijing, Nanjing, and Chongqing | |
NVIDIA T4 | TencentOS Server 3.1 (TK4) GRID16.2 Ubuntu Server 20.04 LTS 64-bit with GRID 16.2 Ubuntu Server 22.04 LTS 64-bit with GRID 16.2 Windows Server 2019 Datacenter Edition 64-bit Chinese Edition with GRID 16.2 Windows Server 2022 Datacenter Edition 64-bit Chinese Edition with GRID 16.2 | Guangzhou, Shanghai, Nanjing, Beijing, Chengdu, Chongqing, Hong Kong (China), Singapore, Bangkok, Frankfurt, Seoul, Tokyo, Silicon Valley, and Virginia | |
Intel SG1 | CentOS 7.6 64-bit + SG1-pv1.3 CentOS 7.6 64-bit + SG1-pv1.4 CentOS 7.6 64-bit + SG1-pv1.5 CentOS 7.6 64-bit + SG1-pv1.6 | Guangzhou, Nanjing, and Chongqing |
Rendering Instance Selection Recommendations
Tencent Cloud offers a diverse range of GPU compute instances to meet the needs of various business use cases. See the table below and select the appropriate compute instance based on your specific requirements.
The selection recommendations for Cloud GPU Service rendering instances are shown in the table below, where ✓ indicates support and ★ indicates a recommendation.
Feature/Instance | GA3 | GNV4v | GNV4 | GN7vw | GI1 |
Graphics and image processing | ★ | ★ | ★ | ★ | ★ |
Video encoding and decoding | ★ | ★ | ★ | ★ | ★ |
Deep learning training | - | - | ✓ | - | - |
Deep learning inference | - | - | ✓ | - | - |
Scientific computing | - | - | - | - | - |
Notes:
The above recommended use cases are for reference only. Choose according to your actual needs.
For NVIDIA-series GPU instances used for 3D rendering tasks, including high-performance graphics processing and video encoding/decoding, you need to install the GRID Driver and configure the License Server. For installation instructions, see Installing NVIDIA GRID Driver. GNV4v, GNV4, and GN7vw instances can opt for specific images with the GRID Driver pre-installed, eliminating the need to install the GRID Driver and configure the License Server separately.
The GNV4v, GNV4, and GN7vw instance families provide vGPU instance types that support vDWS/vWS and are compatible with graphics APIs such as DirectX and OpenGL.
Instance Specifications
Rendering GA3
GA3 is suitable for graphics and image processing scenarios, including 3D rendering and video encoding/decoding.
Notes:
This instance is temporarily available on an allowlist basis. Contact Pre-Sales Online Consultation to request access permissions for purchasing the instance.
This instance requires a pre-installed GA01 driver image for proper functionality. On the purchase page, click Image Market, search for Pre-installed GA01 GPU Driver and select it. For more details, see Using Pre-installed GPU Driver Images.
Hardware Specifications
GPU: Tencent Cloud Star Lake GA01.
CPU: 2.55 GHz AMD EPYCTM Milan processor with a turbo boost of 3.5 GHz.
Memory: Equipped with eight-channel DDR4.
Storage: You can select Cloud Block Storage Type. If you need to scale out, you can create an auto scaling cloud disk and mount it.
Network: Default network optimization is enabled, with instance network performance matching its specifications. Public network can be configured as needed.
Specification | GPU | GPU Memory | vCPU | Memory (GiB) | Private Network Bandwidth (Gbps) | Packet Tx/Rx (PPS) | Number of Queues |
GA3.LARGE10 | GA01 * 1/6 | 32GB * 1/6 | 4 | 10 | 1.5 | 150 thousand | 4 |
GA3.2XLARGE14 | GA01 * 1/4 | 32GB * 1/4 | 8 | 14 | 2 | 350 thousand | 8 |
GA3.4XLARGE30 | GA01 * 1/2 | 32GB * 1/2 | 16 | 30 | 4 | 750 thousand | 16 |
GA3.8XLARGE62 | GA01 * 1 | 32GB * 1 | 32 | 62 | 7 | 1.5 million | 32 |
Rendering GNV4v
NVIDIA GNV4v instances are rendering instances configured with a vDWS License server and pre-installed GRID drivers. They are well-suited for graphics and image processing scenarios, including 3D rendering and video encoding/decoding. These instances eliminate the need for manual GPU graphics and image processing environment configuration.
Notes:
This instance is currently available on an allowlist basis. Contact Pre-Sales Online Consultation to request access permissions for purchasing the instance.
Hardware Specifications
GPU: NVIDIA® A10(FP32 31.2 TFLOPS, INT8 250 TOPS).
CPU: 2.55 GHz AMD EPYCTM Milan processor with a turbo boost of 3.5 GHz.
Memory: Equipped with eight-channel DDR4.
Storage: You can select Cloud Block Storage Type. If you need to scale out, you can create an auto scaling cloud disk and mount it.
Network: Default network optimization is enabled, with instance network performance matching its specifications. Public network can be configured as needed.
Specification | GPU | GPU Memory | vCPU | Memory (GiB) | Private Network Bandwidth (Gbps) | Packet Tx/Rx (PPS) | Number of Queues |
GNV4v.LARGE24 | NVIDIA A10 * 1/4 | 24GB * 1/4 | 6 | 24 | 3 | 500 thousand. | 6 |
GNV4v.3XLARGE58 | NVIDIA A10 * 1/2 | 24GB * 1/2 | 14 | 58 | 7 | 1.1 million | 14 |
GNV4v.7XLARGE116 | NVIDIA A10 * 1 | 24GB * 1 | Day 28 | 116 | 13 | 2.3 million | Day 28 |
Rendering GNV4
NVIDIA GNV4 instances are rendering instances configured with a vDWS License server and pre-installed GRID drivers. They include public images with GRID support, such as Windows Server 2022 Datacenter Edition 64-bit Chinese Edition with GRID 16.2. These instances are well-suited for graphics and image processing scenarios, including 3D rendering and video encoding/decoding. By using these instances, you can eliminate the need for manual GPU graphics and image processing environment configuration.
Hardware Specifications
GPU: NVIDIA® A10 (31.2 TFLOPS single-precision floating-point performance, 250 INT8 TOPS, 500 INT4 TOPS).
CPU: Intel® Xeon® Cooper Lake processor with a fundamental frequency of 3.4 GHz and a turbo frequency of 3.8 GHz.
Memory: Equipped with six-channel DDR4 memory.
Storage: You can select Cloud Block Storage Type. If you need to scale out, you can create an auto scaling cloud disk and mount it.
Network: Default network optimization is enabled, with instance network performance matching its specifications. Public network can be configured as needed.
Specification | GPU | GPU Memory | vCPU | Memory (GiB) | Private Network Bandwidth (Gbps) | Packet Tx/Rx (PPS) | Number of Queues |
GNV4.3XLARGE44 | NVIDIA A10 * 1 | 24GB | 12 | 44 | 2 | 500 thousand. | 12 |
GNV4.6XLARGE88 | NVIDIA A10 * 2 | 24GB * 2 | 24 | 88 | 4 | 1 million | 24 |
GNV4.12XLARGE176 | NVIDIA A10 * 4 | 24GB * 4 | 48 | 176 | 7 | 2.1 million | 32 |
GNV4.24XLARGE352 | NVIDIA A10 * 8 | 24GB * 8 | 96 | 352 | 13 | 4.2 million | 32 |
GNV4.48XLARGE704 | NVIDIA A10 * 16 | 24GB * 16 | 192 | 704 | 25 | 8.5 million | 32 |
Rendering GN7vw
NVIDIA GN7vw instances are rendering instances built on the GN7 foundation, configured with a vDWS License server and pre-installed GRID drivers. They are well-suited for graphics and image processing scenarios, including 3D rendering and video encoding/decoding. These instances eliminate the need for manual GPU graphics and image processing environment configuration.
Note:
This instance is currently available on an allowlist basis. Contact Pre-Sales Online Consultation to request access permissions for purchasing the instance.
Hardware Specifications
GPU: NVIDIA® Tesla® T4 (FP32 8.1 TFLOPS, 130 INT8 TOPS, 260 INT4 TOPS).
CPU: Intel® Xeon® Platinum 8255C CPU with a CPU clock speed of 2.5 GHz.
Memory: DDR4 with a memory speed of up to 2666 MT/s.
Storage: You can select Cloud Block Storage Type. If you need to scale out, you can create an auto scaling cloud disk and mount it.
Network: Default network optimization is enabled, with instance network performance matching its specifications. Public network can be configured as needed.
Specification | GPU | GPU Memory | vCPU | Memory (GiB) | Private Network Bandwidth (Gbps) | Packet Tx/Rx (PPS) | Number of Queues |
GN7vw.LARGE16 | NVIDIA T4 * 1/4 | 16GB * 1/4 | 4 | 16 | 2 | 500 thousand. | 8 |
GN7vw.2XLARGE32 | NVIDIA T4 * 1/2 | 16GB * 1/2 | 8 | 32 | 4 | 800 thousand | 8 |
GN7vw.4XLARGE64 | NVIDIA T4 * 1 | 16GB * 1 | 16 | 64 | 7 | 1.5 million | 8 |
Rendering GI1
GPU rendering GI1 instances are equipped with H3C XG310 accelerator cards, each featuring four Intel SG1 chips. These instances are ideal for scenarios such as Android game streaming, Android Cloud App, and video transcoding.
Notes:
This instance is currently available on an allowlist basis. Contact Pre-Sales Online Consultation to request access permissions for purchasing the instance.
Applicable Scenarios
Android cloud phone
Android game streaming
Android cloud App
Video Transcoding
Hardware Specifications
GPU: Intel® SG1, utilizing H3C XG310 accelerator cards, with each card containing four SG1 chips.
CPU: Intel® Xeon® Platinum 8255c CPU with a CPU clock speed of 2.5 GHz.
Storage: You can select Cloud Block Storage Type. If you need to scale out, you can create an auto scaling cloud disk and mount it.
Network: Default network optimization is enabled, with instance network performance matching its specifications. Public network can be configured as needed.
Specification | GPU | GPU Memory | vCPU | Memory (GiB) | Private Network Bandwidth (Gbps) | Packet Tx/Rx (PPS) | Number of Queues |
GI1.10XLARGE160 | H3C XG310 * 1 (Intel SG1 chips x 4) | 32GB * 1 (8GB * 4) | 42 | 160 | 13 | 2.5 million | 32 |
GI1.21XLARGE320 | H3C XG310 * 2 (Intel SG1 chips x 8) | 32GB * 2 (8GB * 8) | 84 | 320 | 25 | 6 million | 32 |