GPU-accelerated compute-optimized instance families - Elastic GPU Service

This topic describes the features of GPU-accelerated compute-optimized instance families of Elastic Compute Service (ECS) and lists the instance specifications of each instance family.

Recommended instance families
Other available instance families (If these instance families are sold out, you can use the recommended ones.)
- gn5, GPU-accelerated compute-optimized instance family
- gn5i, GPU-accelerated compute-optimized instance family

gn7s, GPU-accelerated compute-optimized instance family

Features:

This instance family uses new Intel Ice lake processors and NVIDIA A30 GPUs that are based on NVIDIA Ampere architecture. You can choose an appropriate mix of GPUs and CPU resources to meet various requirements for AI business.
This instance family uses third-generation SHENLONG architecture and doubles the average bandwidths of virtual private clouds (VPCs), networks, and disks compared with instance families of the previous generation.
Compute:
- Uses NVIDIA A30 GPUs that have the following features:
  - Innovative NVIDIA Ampere architecture
  - Support for the multi-instance GPU (MIG) feature and acceleration features (based on second-generation Tensor cores) to provide diversified business support
- Uses 2.9 GHz Intel^® Xeon^® Scalable (Ice Lake) processors that deliver an all-core turbo frequency of 3.5 GHz.
- Improves memory sizes significantly from instance families of the previous generation.
Storage: Supports only ESSDs.
Note For more information about the performance of cloud disks, see EBS performance.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios: concurrent AI inference tasks that require high-performance CPUs, memory, and GPUs, such as image recognition, speech recognition, and behavior identification.

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	IPv6 addresses per ENI	NIC queues	ENIs
ecs.gn7s-c8g1.2xlarge	8	60	NVIDIA A30 × 1	24GB × 1	16	6,000,000	1	12	8
ecs.gn7s-c16g1.4xlarge	16	120	NVIDIA A30 × 1	24GB × 1	16	6,000,000	1	12	8
ecs.gn7s-c32g1.8xlarge	32	250	NVIDIA A30 × 1	24GB × 1	16	6,000,000	1	12	8
ecs.gn7s-c32g1.16xlarge	64	500	NVIDIA A30 × 2	24GB × 2	32	12,000,000	1	16	15
ecs.gn7s-c32g1.32xlarge	128	1000	NVIDIA A30 × 4	24GB × 4	64	24,000,000	1	32	15
ecs.gn7s-c48g1.12xlarge	48	380	NVIDIA A30 × 1	24GB × 1	16	6,000,000	1	12	8
ecs.gn7s-c56g1.14xlarge	56	440	NVIDIA A30 × 1	24GB × 1	16	6,000,000	1	12	8

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance family.

gn7e, GPU-accelerated compute-optimized instance family

Features:

This instance family uses Intel Ice lake processors and NVIDIA A100 SXM4 80GB GPUs that are interconnected by using NVSwitches. You can choose an appropriate mix of GPUs and CPU resources to meet various AI business requirements.
This instance family uses the third-generation SHENLONG architecture and doubles the average bandwidths of VPCs, networks, and disks compared with instance families of the previous generation.
Compute:
- Uses NVIDIA A100 GPUs that have the following features:
  - Innovative NVIDIA Ampere architecture
  - Connections established between NVIDIA A100 GPUs by using NVSwitches
  - Support for mixed-precision computing with 80 GB of HBM2 memory per GPU
- Uses 2.9 GHz Intel^® Xeon^® Platinum 8369B (Ice Lake) processors that deliver an all-core turbo frequency of 3.5 GHz.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports only ESSDs.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Small- and medium-scale AI training
- High-performance computing (HPC) business accelerated by using Compute Unified Device Architecture (CUDA)
- AI inference tasks that require high GPU processing capabilities or large amounts of GPU memory
- Deep learning applications such as training applications of AI algorithms used in image classification, autonomous vehicles, and speech recognition
- Scientific computing applications that require robust GPU computing capabilities such as computational fluid dynamics, computational finance, molecular dynamics, and environmental analytics.
Important When you use AI training services that feature a high communication load, such as transformer models, you must enable NVlink for GPU-to-GPU communication. Otherwise, data may be damaged due to unpredictable failures caused by large-scale data transmission over PCIe (Peripheral Component Interconnect Express) links. If you are not sure which communication link topology for training you are using,submit a ticket for Alibaba Cloud technical supports.

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.gn7e-c16g1.4xlarge	16	125	NVIDIA A100 × 1	80GB × 1	8	3,000,000	8	8	10
ecs.gn7e-c16g1.16xlarge	64	500	NVIDIA A100 × 4	80GB × 4	32	12,000,000	32	8	10
ecs.gn7e-c16g1.32xlarge	128	1000	NVIDIA A100 × 8	80GB × 8	64	24,000,000	32	16	15

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance family.

When you create or restart a gn7e instance in the ECS console, the Multi-Instance GPU (MIG) feature of the instance is automatically disabled. For more information about MIG, see NVIDIA Multi-Instance GPU User Guide.

The following table describes whether the MIG feature is supported by the instance types in the gn7e instance family.


Instance type	MIG	Description
ecs.gn7e-c16g1.4xlarge	Supported	The MIG feature is supported by single-GPU instances.
ecs.gn7e-c16g1.16xlarge	Not supported	The MIG feature is not supported by multi-GPU instances for security reasons.
ecs.gn7e-c16g1.32xlarge	Not supported

gn7i, GPU-accelerated compute-optimized instance family

Features:

This instance family uses the third-generation SHENLONG architecture to provide predictable and consistent ultra-high performance. This instance family utilizes fast path acceleration on chips to improve storage performance, network performance, and computing stability by an order of magnitude.
Compute:
- Uses NVIDIA A10 GPUs that have the following features:
  - Innovative NVIDIA Ampere architecture
  - Support for acceleration features such as RTX and TensorRT
- Uses 2.9 GHz Intel^® Xeon^® Scalable (Ice Lake) processors that deliver an all-core turbo frequency of 3.5 GHz.
- Provides memory of up to 752 GiB, which is much larger than the memory sizes of the gn6i instance family.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports only ESSDs.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Concurrent AI inference tasks that require high-performance CPUs, memory, and GPUs, such as image recognition, speech recognition, and behavior identification
- Compute-intensive graphics processing tasks that require high-performance 3D graphics virtualization capabilities, such as remote graphic design and cloud gaming.

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.gn7i-c8g1.2xlarge	8	30	NVIDIA A10 × 1	24GB × 1	16	1,600,000	8	4	5
ecs.gn7i-c16g1.4xlarge	16	60	NVIDIA A10 × 1	24GB × 1	16	3,000,000	8	8	5
ecs.gn7i-c32g1.8xlarge	32	188	NVIDIA A10 × 1	24GB × 1	16	6,000,000	12	8	5
ecs.gn7i-c32g1.16xlarge	64	376	NVIDIA A10 × 2	24GB × 2	32	12,000,000	16	15	5
ecs.gn7i-c32g1.32xlarge	128	752	NVIDIA A10 × 4	24GB × 4	64	24,000,000	32	15	10
ecs.gn7i-c48g1.12xlarge	48	310	NVIDIA A10 × 1	24GB × 1	16	9,000,000	16	8	8
ecs.gn7i-c56g1.14xlarge	56	346	NVIDIA A10 × 1	24GB × 1	16	12,000,000	16	12	8

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance family.

gn7, GPU-accelerated compute-optimized instance family

Features:

Compute:
- Uses NVIDIA A100 GPUs. NVSwitches are used to establish connections between NVIDIA A100 GPUs. The GPUs have the following features:
  - Innovative NVIDIA Ampere architecture
  - 40 GB of HBM2 memory per GPU
- Uses 2.5 GHz Intel^® Xeon^® Platinum 8269CY (Cascade Lake) processors.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports only ESSDs.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Deep learning applications such as training applications of AI algorithms used in image classification, autonomous vehicles, and speech recognition
- Scientific computing applications that require robust GPU computing capabilities such as computational fluid dynamics, computational finance, molecular dynamics, and environmental analytics

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs
ecs.gn7-c12g1.3xlarge	12	96	NVIDIA A100 × 1	40GB × 1	4	2,500,000	4	8
ecs.gn7-c13g1.13xlarge	52	378	NVIDIA A100 × 4	40GB × 4	16	9,000,000	16	8
ecs.gn7-c13g1.26xlarge	104	756	NVIDIA A100 × 8	40GB × 8	30	18,000,000	16	15

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance family.

When you create or restart a gn7 instance in the ECS console, the MIG feature of the instance is automatically disabled. For more information about MIG, see NVIDIA Multi-Instance GPU User Guide.

The following table describes whether the MIG feature is supported by the instance types in the gn7 instance family.


Instance type	MIG	Description
ecs.gn7-c12g1.3xlarge	Supported	The MIG feature is supported by single-GPU instances.
ecs.gn7-c13g1.13xlarge	Not supported	The MIG feature is not supported by multi-GPU instances for security reasons.
ecs.gn7-c13g1.26xlarge	Not supported

gn6i, GPU-accelerated compute-optimized instance family

Features:

Compute:
- Uses NVIDIA T4 GPUs that have the following features:
  - Innovative NVIDIA Turing architecture
  - 16 GB memory (320 GB/s bandwidth) per GPU
  - 2,560 CUDA cores per GPU
  - Up to 320 Turing Tensor cores per GPU
  - Mixed-precision Tensor cores that support 65 FP16 TFLOPS, 130 INT8 TOPS, and 260 INT4 TOPS
- Offers a CPU-to-memory ratio of 1:4.
- Uses 2.5 GHz Intel Xeon Platinum 8163 (Skylake) processors.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports standard SSDs, ultra disks, and ESSDs that deliver millions of IOPS.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- AI (deep learning and machine learning) inference for computer vision, speech recognition, speech synthesis, natural language processing (NLP), machine translation, and recommendation systems
- Real-time rendering for cloud games
- Real-time rendering for AR and VR applications
- Graphics workstations or overloaded graphics computing
- GPU-accelerated databases
- High-performance computing

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	Baseline storage IOPS	NIC queues	ENIs	Private IP addresses per ENI
ecs.gn6i-c4g1.xlarge	4	15	NVIDIA T4 × 1	16GB × 1	4	500,000	N/A	2	2	10
ecs.gn6i-c8g1.2xlarge	8	31	NVIDIA T4 × 1	16GB × 1	5	800,000	N/A	2	2	10
ecs.gn6i-c16g1.4xlarge	16	62	NVIDIA T4 × 1	16GB × 1	6	1,000,000	N/A	4	3	10
ecs.gn6i-c24g1.6xlarge	24	93	NVIDIA T4 × 1	16GB × 1	7.5	1,200,000	N/A	6	4	10
ecs.gn6i-c40g1.10xlarge	40	155	NVIDIA T4 × 1	16GB × 1	10	1,600,000	N/A	16	10	10
ecs.gn6i-c24g1.12xlarge	48	186	NVIDIA T4 × 2	16GB × 2	15	2,400,000	N/A	12	6	10
ecs.gn6i-c24g1.24xlarge	96	372	NVIDIA T4 × 4	16GB × 4	30	4,800,000	250,000	24	8	10

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance family.

gn6e, GPU-accelerated compute-optimized instance family

Features:

Compute:
- Uses NVIDIA V100 GPUs that each has 32 GB of GPU memory and support NVLink.
- Uses NVIDIA V100 GPUs (SXM2-based) that have the following features:
  - Innovative NVIDIA Volta architecture
  - 32 GB of HBM2 memory (900 GB/s bandwidth) per GPU
  - 5,120 CUDA cores per GPU
  - 640 Tensor cores per GPU
  - Support for up to six NVLink bidirectional connections, which each provide a bandwidth of 25 GB/s in each direction for a total bandwidth of 300 GB/s (6 × 25 × 2 = 300)
- Offers a CPU-to-memory ratio of 1:8.
- Uses 2.5 GHz Intel^® Xeon^® Platinum 8163 (Skylake) processors.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports ESSDs, standard SSDs, and ultra disks.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Deep learning applications such as the training and inference applications of AI algorithms used in image classification, autonomous driving, and speech recognition
- Scientific computing applications, such as computational fluid dynamics, computational finance, molecular dynamics, and environmental analytics.

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.gn6e-c12g1.3xlarge	12	92	NVIDIA V100 × 1	32GB × 1	5	800,000	8	6	10
ecs.gn6e-c12g1.12xlarge	48	368	NVIDIA V100 × 4	32GB × 4	16	2,400,000	8	8	20
ecs.gn6e-c12g1.24xlarge	96	736	NVIDIA V100 × 8	32GB × 8	32	4,800,000	16	8	20

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance family.

gn6v, GPU-accelerated compute-optimized instance family

Features:

Compute:
- Uses NVIDIA V100 GPUs.
- Uses NVIDIA V100 GPUs (SXM2-based) that have the following features:
  - Innovative NVIDIA Volta architecture
  - 16 GB of HBM2 memory (900 GB/s bandwidth) per GPU
  - 5,120 CUDA cores per GPU
  - 640 Tensor cores per GPU
  - Support for up to six NVLink bidirectional connections, which each provide a bandwidth of 25 GB/s in each direction for a total bandwidth of 300 GB/s (6 × 25 × 2 = 300)
- Offers a CPU-to-memory ratio of 1:4.
- Uses 2.5 GHz Intel^® Xeon^® Platinum 8163 (Skylake) processors.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports ESSDs, standard SSDs, and ultra disks.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Deep learning applications such as the training and inference applications of AI algorithms used in image classification, autonomous driving, and speech recognition
- Scientific computing applications, such as computational fluid dynamics, computational finance, molecular dynamics, and environmental analytics.

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	Baseline storage IOPS	NIC queues	ENIs	Private IP addresses per ENI
ecs.gn6v-c8g1.2xlarge	8	32	NVIDIA V100 × 1	16GB × 1	2.5	800,000	N/A	4	4	10
ecs.gn6v-c8g1.8xlarge	32	128	NVIDIA V100 × 4	16GB × 4	10	2,000,000	N/A	8	8	20
ecs.gn6v-c8g1.16xlarge	64	256	NVIDIA V100 × 8	16GB × 8	20	2,500,000	N/A	16	8	20
ecs.gn6v-c10g1.20xlarge	82	336	NVIDIA V100 × 8	16GB × 8	32	4,500,000	250,000	16	8	20

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance family.

ebmgn7e, GPU-accelerated compute-optimized ECS Bare Metal Instance family

Features:

This instance family uses the SHENLONG architecture to provide flexible and powerful software-defined compute.
Compute:
- Uses NVIDIA A100 SXM4 80GB GPUs that support NVSwitches and deliver up to 312 TFLOPS of TensorFloat-32 (TF32) computing power.
- Uses 2.9 GHz Intel^® Xeon^®Scalable processors that deliver an all-core turbo frequency of 3.5 GHz and support PCIe 4.0 interfaces.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports only ESSDs. ESSDs at performance level (PL) 3 can deliver a maximum of 500,000 IOPS and 2,000 MB/s of throughput, which can meet the cache requirements of training.
  Note For more information about the performance of ESSDs, see ESSDs.
Network:
- Supports IPv6.
- Provides ultra-high network performance with a packet forwarding rate of 24,000,000 pps.
Supported scenarios:
- Deep learning training and development
- High-performance computing (HPC) and simulations
Important When you use AI training services that feature a high communication load, such as transformer models, you must enable NVlink for GPU-to-GPU communication. Otherwise, data may be damaged due to unpredictable failures caused by large-scale data transmission over PCIe (Peripheral Component Interconnect Express) links. If you are not sure which communication link topology for training you are using,submit a ticket for Alibaba Cloud technical supports.

Instance types


Instance type	vCPU	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues (primary ENI/secondary ENI)	ENIs
ecs.ebmgn7e.32xlarge	128	1024	NVIDIA A100 * 8	80GB * 8	64	24,000,000	32/12	32

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance type specifications.

You need to manually check the status of the MIG feature and enable or disable the MIG feature after you start an ebmgn7e instance. For more information about MIG, see NVIDIA Multi-Instance GPU User Guide.

The following table describes whether the MIG feature is supported by the instance types in the ebmgn7e instance family.


Instance type	MIG	Description
ecs.ebmgn7e.32xlarge	Supported	The MIG feature is supported by ebmgn7e instances.

ebmgn7i, GPU-accelerated compute optimized ECS Bare Metal Instance family

Features:

This instance family uses the SHENLONG architecture to provide flexible and powerful software-defined compute.
Compute:
- Uses NVIDIA A10 GPUs that have the following features:
  - Innovative NVIDIA Ampere architecture.
  - Support for acceleration features such as vGPU, RTX technology, and TensorRT inference engine
- Uses 2.9 GHz Intel^® Xeon^® Scalable (Ice Lake) processors that deliver an all-core turbo frequency of 3.5 GHz.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports only ESSDs.
Network:
- Supports IPv6.
- Provides ultra-high network performance with a packet forwarding rate of 24,000,000 pps.
Supported scenarios:
- Concurrent AI inference tasks that require high-performance CPUs, memory, and GPUs, such as image recognition, speech recognition, and behavior identification
- Compute-intensive graphics processing tasks that require high-performance 3D graphics virtualization capabilities, such as remote graphic design and cloud gaming
- Scenarios that require high network bandwidth and disk bandwidth, such as the creation of high-performance render farms
- Small-scale deep learning and training applications that require high network bandwidth

Instance types


Instance type	vCPU	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs
ecs.ebmgn7i.32xlarge	128	768	NVIDIA A10 * 4	24GB * 4	64	24,000,000	32	32

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance type specifications.

ebmgn7, GPU-accelerated compute-optimized ECS Bare Metal Instance family

Features:

This instance family uses the SHENLONG architecture to provide flexible and powerful software-defined compute.
Compute:
- Uses NVIDIA A100 GPUs. NVSwitches are used to establish interconnections between NVIDIA A100 GPUs. The GPUs have the following features:
  - Innovative NVIDIA Ampere architecture
  - 40 GB HBM2 memory per GPU
- Uses 2.5 GHz Intel^® Xeon^® Platinum 8269CY (Cascade Lake) processors.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports only ESSDs.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Deep learning applications such as training applications of AI algorithms used in image classification, autonomous vehicles, and speech recognition
- Scientific computing applications that require robust GPU computing capabilities such as computational fluid dynamics, computational finance, molecular dynamics, and environmental analytics

Instance types


Instance type	vCPU	Memory (GiB)	GPU	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.ebmgn7.26xlarge	104	768	NVIDIA A100 * 8	30	18,000,000	16	15	10

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance type specifications.

You need to manually check the status of the MIG feature and enable or disable the MIG feature after you start an ebmgn7 instance. For more information about MIG, see NVIDIA Multi-Instance GPU User Guide.

The following table describes whether the MIG feature is supported by the instance types in the ebmgn7 instance family.


Instance type	MIG	Description
ecs.ebmgn7.26xlarge	Supported	The MIG feature is supported by ebmgn7 instances.

ebmgn6ia, GPU-accelerated compute-optimized ECS Bare Metal Instance family

This instance family is in invitational preview. To use this instance family, Submit a ticket.

Features:

This instance family uses the third-generation SHENLONG architecture and fast path acceleration on chips to provide predictable and consistent ultra-high computing, storage, and network performance.
This instance family uses NVIDIA T4 GPUs to offer GPU acceleration capabilities for graphics and AI applications and adopts container technology to start up to 60 virtual Android devices and provide hardware-accelerated video transcoding.
Compute:
- Offers a CPU-to-memory ratio of 1:3.
- Uses 2.8 GHz Ampere^® Altra^® Arm-based processors that deliver a turbo frequency of 3.0 GHz and provides high performance and high compatibility with applications for Android servers.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports only ESSDs.
Network: supports IPv6.
Supported scenarios: remote application services based on Android, such as always-on cloud-based services, cloud-based mobile games, cloud-based mobile phones, and Android service crawlers.

Instance types


Instance type	vCPU	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.ebmgn6ia.20xlarge	80	256	NVIDIA T4 * 2	16GB * 2	32	24,000,000	32	15	10

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance type specifications.
Ampere^® Altra^® processors have specific requirements on the kernels of operating systems. Instances of this instance type can use Alibaba Cloud Linux 3 images and CentOS 8.4 or later images. We recommend that you use Alibaba Cloud Linux 3 images on the instances. If you want to use another operating system distribution, patch the kernel of an instance that runs an operating system of that distribution, create a custom image from the instance, and then use the custom image to create instances of this instance type. For information about kernel patches, visit Ampere Altra (TM) Linux Kernel Porting Guide.

ebmgn6e, GPU-accelerated compute-optimized ECS Bare Metal Instance family

Features:

This instance family uses the SHENLONG architecture to provide flexible and powerful software-defined compute.
This instance family uses NVIDIA V100 GPUs that each has 32 GB of GPU memory and support NVLink.
This instance family uses NVIDIA V100 GPUs (SXM2-based) that have the following features:
- Innovative NVIDIA Volta architecture
- 32 GB of HBM2 memory (900 GB/s bandwidth) per GPU
- 5,120 CUDA cores per GPU
- 640 Tensor cores per GPU
- Support for up to six NVLink connections. Each NVLink connection provides a bandwidth of 25 GB/s in each direction for a total bandwidth of 300 GB/s (6 × 25 × 2 = 300)
Compute:
- Offers a CPU-to-memory ratio of 1:8.
- Uses 2.5 GHz Intel^® Xeon^® Platinum 8163 (Skylake) processors.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports ESSDs, standard SSDs, and ultra disks.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Deep learning applications, such as training and inference applications of AI algorithms used in image classification, autonomous vehicles, and speech recognition
- Scientific computing applications, such as computational fluid dynamics, computational finance, molecular dynamics, and environmental analytics

Instance types


Instance type	vCPU	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.ebmgn6e.24xlarge	96	768	NVIDIA V100 * 8	32GB * 8	32	4,800,000	16	15	10

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance type specifications.

ebmgn6v, GPU-accelerated compute-optimized ECS Bare Metal Instance family

Features:

This instance family uses the SHENLONG architecture to provide flexible and powerful software-defined compute.
This instance family uses NVIDIA V100 GPUs.
This instance family uses NVIDIA V100 GPUs (SXM2-based) that have the following features:
- Innovative NVIDIA Volta architecture
- 16 GB of HBM2 memory (900 GB/s bandwidth) per GPU
- 5,120 CUDA cores per GPU
- 640 Tensor cores per GPU
- Support for up to six NVLink connections. Each NVLink connection provides a bandwidth of 25 GB/s in each direction for a total bandwidth of 300 GB/s (6 × 25 × 2 = 300)
Compute:
- Offers a CPU-to-memory ratio of 1:4.
- Uses 2.5 GHz Intel^® Xeon^® Platinum 8163 (Skylake) processors.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports ESSDs, standard SSDs, and ultra disks.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Deep learning applications, such as training and inference applications of AI algorithms used in image classification, autonomous vehicles, and speech recognition
- Scientific computing applications, such as computational fluid dynamics, computational finance, molecular dynamics, and environmental analytics

Instance types


Instance type	vCPU	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.ebmgn6v.24xlarge	96	384	NVIDIA V100 * 8	16GB * 8	30	4,500,000	8	32	10

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance type specifications.

ebmgn6i, GPU-accelerated compute-optimized ECS Bare Metal Instance family

Features:

This instance family uses the SHENLONG architecture to provide flexible and powerful software-defined compute.
This instance family uses NVIDIA T4 GPUs that have the following features:
- Innovative NVIDIA Turing architecture
- 16 GB of memory (320 GB/s bandwidth) per GPU
- 2,560 CUDA cores per GPU
- Up to 320 Turing Tensor cores per GPU
- Mixed-precision Tensor cores that support 65 FP16 TFLOPS, 130 INT8 TOPS, and 260 INT4 TOPS
Compute:
- Offers a CPU-to-memory ratio of 1:4.
- Uses 2.5 GHz Intel^® Xeon^® Platinum 8163 (Skylake) processors.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports standard SSDs, ultra disks, and ESSDs that deliver millions of IOPS.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- AI (deep learning and machine learning) inference for computer vision, voice recognition, speech synthesis, natural language processing (NLP), machine translation, and reference systems
- Real-time rendering for cloud games
- Real-time rendering for AR and VR applications
- Graphics workstations or graphics-heavy computing
- GPU-accelerated databases
- High-performance computing

Instance types


Instance type	vCPU	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.ebmgn6i.24xlarge	96	384	NVIDIA T4 * 4	16GB * 4	30	4,500,000	8	32	10

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance type specifications.

sccgn6e, GPU-accelerated compute-optimized SCC instance family

To use this instance family,submit a ticket.

Features:

This instance family provides all features of ECS Bare Metal Instance. For more information, see Overview.
Compute:
- Uses NVIDIA V100 GPUs (SXM2-based) that have the following features:
  - Innovative NVIDIA Volta architecture
  - 32 GB of HBM2 GPU memory
  - CUDA Cores 5120
  - Tensor Cores 640
  - GPU memory bandwidth of up to 900 GB/s
  - Support for up to six bidirectional NVLink connections, which each have a unidirectional bandwidth of 25 GB/s for a total bandwidth of 300 GB/s
- Offers a CPU-to-memory ratio of 1:8.
- Uses 2.5 GHz Intel^® Xeon^® Platinum 8163 (Skylake) processors for consistent computing performance.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports ESSDs, standard SSDs, and ultra disks.
- Supports high-performance Cloud Paralleled File System (CPFS).
Network:
- Supports IPv6.
- Supports VPCs.
- Supports RoCE v2 networks, which are dedicated to low-latency RDMA communication.
Supported scenarios:
- Ultra-large-scale training for machine learning on a distributed GPU cluster
- Large-scale high-performance scientific computing and simulation calculation
- Large-scale data analytics, batch processing, and video encoding

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	GPU memory (GB)	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	RoCE bandwidth (Gbit/s)	NIC queues	ENIs	Private IP addresses per ENI
ecs.sccgn6e.24xlarge	96	768.0	NVIDIA V100 * 8	32GB * 8	32	4,800,000	50	8	32	10

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see Instance type specifications.

sccgn6, GPU-accelerated compute-optimized SCC instance family

Features:

This instance family provides all features of ECS Bare Metal Instance. For more information, see Overview.
Compute:
- Uses NVIDIA V100 GPUs (SXM2-based) that have the following features:
  - Innovative NVIDIA Volta architecture
  - Up to 16 GB of HBM2 GPU memory
  - CUDA Cores 5120
  - Tensor Cores 640
  - GPU memory bandwidth of up to 900 GB/s
  - Support for up to six bidirectional NVLink connections, which each have a unidirectional bandwidth of 25 GB/s for a total bandwidth of 300 GB/s
- Offers a CPU-to-memory ratio of 1:4.
- Uses 2.5 GHz Intel^® Xeon^® Platinum 8163 (Skylake) processors for consistent computing performance.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports ESSDs, standard SSDs, and ultra disks.
- Supports high-performance CPFS.
Network:
- Supports IPv6.
- Supports VPCs.
- Supports RoCE v2 networks, which are dedicated to low-latency RDMA communication.
Supported scenarios:
- Ultra-large-scale training for machine learning on a distributed GPU cluster
- Large-scale high-performance scientific computing and simulation calculation
- Large-scale data analytics, batch processing, and video encoding

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	RoCE bandwidth (Gbit/s)	NIC queues	ENIs	Private IP addresses per ENI
ecs.sccgn6.24xlarge	96	384.0	NVIDIA V100 * 8	30	4,500,000	50	8	32	10

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see Instance type specifications.

gn5, GPU-accelerated compute-optimized instance family

Features:

Compute:
- Uses NVIDIA P100 GPUs.
- Offers multiple CPU-to-memory ratios.
- Uses 2.5 GHz Intel^® Xeon^®E5-2682 v4 (Broadwell) processors.
Storage:
- Supports high-performance local Non-Volatile Memory Express (NVMe) SSDs.
- Is an instance family in which all instances are I/O optimized.
- Supports standard SSDs and ultra disks.
Network:
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Deep learning
- Scientific computing applications, such as computational fluid dynamics, computational finance, genomics, and environmental analytics
- Server-side GPU compute workloads such as high-performance computing, rendering, and multi-media encoding and decoding

Instance types


Instance type	vCPUs	Memory (GiB)	Local storage (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.gn5-c4g1.xlarge	4	30	440	NVIDIA P100 × 1	16GB × 1	3	300,000	1	3	10
ecs.gn5-c8g1.2xlarge	8	60	440	NVIDIA P100 × 1	16GB × 1	3	400,000	1	4	10
ecs.gn5-c4g1.2xlarge	8	60	880	NVIDIA P100 × 2	16GB × 2	5	1,000,000	2	4	10
ecs.gn5-c8g1.4xlarge	16	120	880	NVIDIA P100 × 2	16GB × 2	5	1,000,000	4	8	20
ecs.gn5-c28g1.7xlarge	28	112	440	NVIDIA P100 × 1	16GB × 1	5	1,000,000	8	8	20
ecs.gn5-c8g1.8xlarge	32	240	1760	NVIDIA P100 × 4	16GB × 4	10	2,000,000	8	8	20
ecs.gn5-c28g1.14xlarge	56	224	880	NVIDIA P100 × 2	16GB × 2	10	2,000,000	14	8	20
ecs.gn5-c8g1.14xlarge	54	480	3520	NVIDIA P100 × 8	16GB × 8	25	4,000,000	14	8	20

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance family.

gn5i, GPU-accelerated compute-optimized instance family

Features:

Compute:
- Uses NVIDIA P4 GPUs.
- Offers a CPU-to-memory ratio of 1:4.
- Uses 2.5 GHz Intel^® Xeon^®E5-2682 v4 (Broadwell) processors.
Storage:
- Is an instance family in which all instances are I/O optimized.
- Supports standard SSDs and ultra disks.
Network:
- Supports IPv6.
- Provides high network performance based on large computing capacity.
Supported scenarios:
- Deep learning inference
- Server-side GPU compute workloads such as multi-media encoding and decoding

Instance types


Instance type	vCPUs	Memory (GiB)	GPU	GPU memory	Network bandwidth (Gbit/s)	Packet forwarding rate (pps)	NIC queues	ENIs	Private IP addresses per ENI
ecs.gn5i-c2g1.large	2	8	NVIDIA P4 × 1	8GB × 1	1	100,000	2	2	6
ecs.gn5i-c4g1.xlarge	4	16	NVIDIA P4 × 1	8GB × 1	1.5	200,000	2	3	10
ecs.gn5i-c8g1.2xlarge	8	32	NVIDIA P4 × 1	8GB × 1	2	400,000	4	4	10
ecs.gn5i-c16g1.4xlarge	16	64	NVIDIA P4 × 1	8GB × 1	3	800,000	4	8	20
ecs.gn5i-c16g1.8xlarge	32	128	NVIDIA P4 × 2	8GB × 2	6	1,200,000	8	8	20
ecs.gn5i-c28g1.14xlarge	56	224	NVIDIA P4 × 2	8GB × 2	10	2,000,000	14	8	20

Note

You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.
For more information about these specifications, see the "Instance type specifications" section in Instance family.