This topic describes the features of GPU-accelerated compute optimized instance families and lists the instance types of each family.

gn7, GPU-accelerated compute optimized instance family

This instance family is in invitational preview and not available for sale.

Features
  • Uses 2.5 GHz Intel ® Xeon ® Platinum 8269CY (Cascade Lake) processors.
  • Uses NVIDIA A100 GPU processors. NVSwitches are used to establish connections between multiple NVIDIA A100 GPU processors.
    • Innovative Ampere architecture
    • 40 GB HBM2 memory per GPU
  • Is an instance family in which all instances are I/O optimized.
  • Supports enhanced SSDs (ESSDs), standard SSDs, and ultra disks.
  • Provides high network performance based on large computing capacity.
  • Suits to the following scenarios:
    • Deep learning applications such as training applications of AI algorithms used in image classification, autonomous vehicles, and speech recognition
    • Scientific computing applications that have high GPU workloads such as computational fluid dynamics, computational finance, molecular dynamics, and environmental analysis
Instance types
Instance type vCPU Memory (GiB) Local storage (GiB) GPU GPU memory (GB) Total inbound and outbound bandwidth (Gbit/s) Total inbound and outbound packet forwarding rate (Kpps) IPv6 support NIC queues ENIs (including one primary ENI)
ecs.gn7-c12g1.3xlarge 12 95.0 None 1*A100 40 4.0 2,500 Yes 4 8
ecs.gn7-c13g1.6xlarge 26 190.0 None 2*A100 80 8.0 4,500 Yes 8 8
ecs.gn7-c13g1.13xlarge 52 380.0 None 4*A100 160 15.0 9,000 Yes 16 8
ecs.gn7-c13g1.26xlarge 104 760.0 None 8*A100 320 30.0 18,000 Yes 16 16
Note

gn6i, GPU-accelerated compute optimized instance family

Features
  • Is an instance family in which all instances are I/O optimized.
  • Offers a CPU-to-memory ratio of 1:4.
  • Uses 2.5 GHz Intel ® Xeon ® Platinum 8163 (Skylake) processors.
  • Supports ESSDs that deliver millions of IOPS, standard SSDs, and ultra disks.
  • Uses NVIDIA T4 GPU accelerators that feature:
    • New 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
  • Provides high network performance based on large computing capacity.
  • Suits to the following 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 gaming
    • Real-time rendering for AR and VR applications
    • Graphics workstations or overloaded graphics computing
    • GPU-accelerated databases
    • High-performance computing
Instance types
Instance type vCPU Memory (GiB) Local storage (GiB) GPU GPU memory (GB) Total inbound and outbound bandwidth (Gbit/s) Total inbound and outbound packet forwarding rate (Kpps) IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.gn6i-c4g1.xlarge 4 15.0 None T4*1 16 4.0 500 Yes 2 2 10
ecs.gn6i-c8g1.2xlarge 8 31.0 None T4*1 16 5.0 800 Yes 2 2 10
ecs.gn6i-c16g1.4xlarge 16 62.0 None T4*1 16 6.0 1,000 Yes 4 3 10
ecs.gn6i-c24g1.6xlarge 24 93.0 None T4*1 16 7.5 1,200 Yes 6 4 10
ecs.gn6i-c24g1.12xlarge 48 186.0 None T4*2 32 15.0 2,400 Yes 12 6 10
ecs.gn6i-c24g1.24xlarge 96 372.0 None T4*4 64 30.0 4,800 Yes 24 8 10
Note

gn6e, GPU-accelerated compute optimized instance family

Features
  • Is an instance family in which all instances are I/O optimized.
  • Supports ESSDs, standard SSDs, and ultra disks.
  • Uses NVIDIA V100 (32 GB NVLink) GPU processors.
  • Offers a CPU-to-memory ratio of 1:8.
  • Uses 2.5 GHz Intel® Xeon® Platinum 8163 (Skylake) processors.
  • Uses NVIDIA V100 GPU computing accelerators (SXM2-based) that feature:
    • New NVIDIA Volta architecture
    • 32 GB 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 and a total bandwidth of 300 GB/s (25 GB/s per connection)
  • Provides high network performance based on large computing capacity.
  • Suits to the following 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 analysis
Instance types
Instance type vCPU Memory (GiB) Local storage (GiB) GPU GPU memory (GB) Total inbound and outbound bandwidth (Gbit/s) Total inbound and outbound packet forwarding rate (Kpps) IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.gn6e-c12g1.3xlarge 12 92.0 None V100*1 32 5.0 800 Yes 8 6 10
ecs.gn6e-c12g1.12xlarge 48 368.0 None V100*4 128 16.0 2,400 Yes 8 8 20
ecs.gn6e-c12g1.24xlarge 96 736.0 None V100*8 256 32.0 4,800 Yes 16 8 20
Note

gn6v, GPU-accelerated compute optimized instance family

Features
  • Is an instance family in which all instances are I/O optimized.
  • Supports ESSDs, standard SSDs, and ultra disks.
  • Uses NVIDIA V100 GPU processors.
  • Offers a CPU-to-memory ratio of 1:4.
  • Uses 2.5 GHz Intel® Xeon® Platinum 8163 (Skylake) processors.
  • Uses NVIDIA V100 GPU computing accelerators (SXM2-based) that feature:
    • New NVIDIA Volta architecture
    • 16 GB 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 and a total bandwidth of 300 GB/s (25 GB/s per connection)
  • Provides high network performance based on large computing capacity.
  • Suits to the following 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 analysis
Instance types
Instance type vCPU Memory (GiB) Local storage (GiB) GPU GPU memory (GB) Total inbound and outbound bandwidth (Gbit/s) Total inbound and outbound packet forwarding rate (Kpps) IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.gn6v-c8g1.2xlarge 8 32.0 None 1 * NVIDIA V100 1 * 16 2.5 800 Yes 4 4 10
ecs.gn6v-c8g1.8xlarge 32 128.0 None 4 * NVIDIA V100 4 * 16 10.0 2,000 Yes 8 8 20
ecs.gn6v-c8g1.16xlarge 64 256.0 None 8 * NVIDIA V100 8 * 16 20.0 2,500 Yes 16 8 20
ecs.gn6v-c10g1.20xlarge 82 336.0 None 8 * NVIDIA V100 8 * 16 32.0 4,500 Yes 16 8 20
Note

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

Features
  • Provides flexible and powerful software-defined compute based on the SHENLONG architecture.
  • Uses NVIDIA A100 GPUs. NVSwitches are used to establish connections between NVIDIA A100 GPUs. The GPUs feature:
    • Innovative Ampere architecture
    • 40 GB HBM2 memory per GPU
  • Uses 2.5 GHz Intel ® Xeon ® Platinum 8269CY (Cascade Lake) processors.
  • Is an instance family in which all instances are I/O optimized.
  • Supports ESSDs, standard SSDs, and ultra disks.
  • Provides high network performance based on large computing capacity.
  • Suits the following scenarios:
    • Deep learning applications such as training applications of AI algorithms used in image classification, autonomous vehicles, and speech recognition
    • Scientific computing applications that have high GPU workloads such as computational fluid dynamics, computational finance, molecular dynamics, and environmental analysis
Instance types
Instance type vCPUs Memory (GiB) Local storage (GiB) GPUs Bandwidth (bidirectional), Gbit/s Packet forwarding rate (bidirectional), Kpps IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.ebmgn7.26xlarge 104 768.0 None A100 × 8 30.0 18,000 Yes 16 15 10
Note

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

Features
  • Provides flexible and powerful software-defined compute based on the SHENLONG architecture.
  • Uses NVIDIA V100 (32 GB NVLink) GPUs.
  • Uses NVIDIA V100 GPUs (SXM2-based) that feature:
    • Innovative Volta architecture
    • 32 GB 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 for a total bandwidth of 300 GB/s per GPU (25 GB/s per connection)
  • Offers a CPU-to-memory ratio of 1:8.
  • Uses 2.5 GHz Intel® Xeon® Platinum 8163 (Skylake) processors.
  • Is an instance family in which all instances are I/O optimized.
  • Supports ESSDs, standard SSDs, and ultra disks.
  • Provides high network performance based on large computing capacity.
  • Suits the following 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 analysis
Instance types
Instance type vCPUs Memory (GiB) Local storage (GiB) GPUs GPU memory (GB) Bandwidth (bidirectional), Gbit/s Packet forwarding rate (bidirectional), Kpps IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.ebmgn6e.24xlarge 96 768.0 None V100 × 8 256 32.0 4,800 Yes 16 15 20
Note

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

Features
  • Provides flexible and powerful software-defined compute based on the SHENLONG architecture.
  • Uses NVIDIA V100 GPUs.
  • Uses NVIDIA V100 GPUs (SXM2-based) that feature:
    • Innovative Volta architecture
    • 16 GB HBM2 memory per GPU (900 Gbit/s bandwidth).
    • 5,120 CUDA cores per GPU
    • 640 Tensor cores per GPU
    • Support for up to six NVLink connections for a total bandwidth of 300 GB/s per GPU (25 GB/s per connection)
  • Offers a CPU-to-memory ratio of 1:4.
  • Uses 2.5 GHz Intel® Xeon® Platinum 8163 (Skylake) processors.
  • Is an instance family in which all instances are I/O optimized.
  • Supports ESSDs, standard SSDs, and ultra disks.
  • Provides high network performance based on large computing capacity.
  • Suits the following 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 analysis
Instance types
Instance type vCPUs Memory (GiB) Local storage (GiB) GPUs GPU memory (GB) Bandwidth (bidirectional), Gbit/s Packet forwarding rate (bidirectional), Kpps IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.ebmgn6v.24xlarge 96 384.0 None V100 × 8 128 30.0 4,500 Yes 8 32 10
Note

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

Features
  • Provides flexible and powerful software-defined compute based on the SHENLONG architecture.
  • Uses NVIDIA T4 GPUs that feature:
    • New 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.
  • Is an instance family in which all instances are I/O optimized.
  • Supports standard SSDs, ultra disks, and ESSDs that deliver millions of IOPS.
  • Provides high network performance based on large computing capacity.
  • Suits the following 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 gaming
    • 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) Local storage (GiB) GPUs GPU memory (GB) Bandwidth (bidirectional), Gbit/s Packet forwarding rate (bidirectional), Kpps IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.ebmgn6i.24xlarge 96 384.0 None T4 × 4 64 30.0 4,500 Yes 8 32 10
Note

sccgn6, GPU-accelerated compute optimized SCC instance family

Features
  • Is an instance family in which all instances are I/O optimized.
  • Offers a CPU-to-memory ratio of 1:4.
  • Uses 2.5 GHz Intel® Xeon® Platinum 8163 (Skylake) processors for consistent computing performance.
  • Provides all features of ECS Bare Metal Instance.
  • Storage:
    • Supports ESSDs, standard SSDs, and ultra disks.
    • Supports high performance CPFSs.
  • Networking:
    • Supports VPCs.
    • Supports RoCE v2 networks, which are dedicated to low-latency RDMA communication.
  • Uses NVIDIA V100 GPUs (SXM2-based) that feature:
    • Innovative Volta architecture
    • Up to 16 GB HBM2 GPU memory
    • 5,120 CUDA cores
    • 640 Tensor cores
    • GPU memory bandwidth of up to 900 Gbit/s
    • Support for up to six NVLink connections and total bandwidth of 300 GB/s (25 GB/s per connection)
  • Suits the following scenarios:
    • Ultra-large-scale training for machine learning on a distributed GPU cluster
    • Large-scale high-performance scientific computing and simulations
    • Large-scale data analysis, batch processing, and video encoding
Instance types
Instance type vCPUs Memory (GiB) Local storage (GiB) GPUs Bandwidth (bidirectional), Gbit/s Packet forwarding rate (bidirectional), Kpps RoCE bandwidth (bidirectional), Gbit/s IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.sccgn6.24xlarge 96 384.0 None V100 × 8 30 4,500 25 × 2 Yes 8 32 10
Note

gn5, GPU-accelerated compute optimized instance family

Features
  • Is an instance family in which all instances are I/O optimized.
  • Supports standard SSDs and ultra disks only.
  • Uses NVIDIA P100 GPU processors.
  • Offers multiple CPU-to-memory ratios.
  • Supports high-performance local NVMe SSDs.
  • Uses 2.5 GHz Intel ® Xeon ® E5-2682 v4 (Broadwell) processors.
  • Provides high network performance based on large computing capacity.
  • Suits to the following scenarios:
    • Deep learning
    • Scientific computing applications, such as fluid dynamics, finance, genomics, and environmental analysis
    • Server-side GPU compute workloads such as high-performance computing, rendering, and multi-media encoding and decoding
Instance types
Instance type vCPU Memory (GiB) Local storage (GiB) GPU GPU memory (GB) Total inbound and outbound bandwidth (Gbit/s) Total inbound and outbound packet forwarding rate (Kpps) IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.gn5-c4g1.xlarge 4 30.0 440 1 * NVIDIA P100 1 * 16 3.0 300 No 1 3 10
ecs.gn5-c8g1.2xlarge 8 60.0 440 1 * NVIDIA P100 1 * 16 3.0 400 No 1 4 10
ecs.gn5-c4g1.2xlarge 8 60.0 880 2 * NVIDIA P100 2 * 16 5.0 1,000 No 2 4 10
ecs.gn5-c8g1.4xlarge 16 120.0 880 2 * NVIDIA P100 2 * 16 5.0 1,000 No 4 8 20
ecs.gn5-c28g1.7xlarge 28 112.0 440 1 * NVIDIA P100 1 * 16 5.0 1,000 No 8 8 20
ecs.gn5-c8g1.8xlarge 32 240.0 1760 4 * NVIDIA P100 4 * 16 10.0 2,000 No 8 8 20
ecs.gn5-c28g1.14xlarge 56 224.0 880 2 * NVIDIA P100 2 * 16 10.0 2,000 No 14 8 20
ecs.gn5-c8g1.14xlarge 54 480.0 3520 8 * NVIDIA P100 8 * 16 25.0 4,000 No 14 8 20
Note

gn5i, GPU-accelerated compute optimized instance family

Features
  • Is an instance family in which all instances are I/O optimized.
  • Supports standard SSDs and ultra disks only.
  • Uses NVIDIA P4 GPU processors.
  • Offers a CPU-to-memory ratio of 1:4
  • Uses 2.5 GHz Intel ® Xeon ® E5-2682 v4 (Broadwell) processors.
  • Provides high network performance based on large computing capacity.
  • Suits to the following scenarios:
    • Deep learning inference
    • Server-side GPU compute workloads such as multi-media encoding and decoding
Instance types
Instance type vCPU Memory (GiB) Local storage (GiB) GPU GPU memory (GB) Total inbound and outbound bandwidth (Gbit/s) Total inbound and outbound packet forwarding rate (Kpps) IPv6 support NIC queues ENIs (including one primary ENI) Private IP addresses per ENI
ecs.gn5i-c2g1.large 2 8.0 None 1 * NVIDIA P4 1 * 8 1.0 100 Yes 2 2 6
ecs.gn5i-c4g1.xlarge 4 16.0 None 1 * NVIDIA P4 1 * 8 1.5 200 Yes 2 3 10
ecs.gn5i-c8g1.2xlarge 8 32.0 None 1 * NVIDIA P4 1 * 8 2.0 400 Yes 4 4 10
ecs.gn5i-c16g1.4xlarge 16 64.0 None 1 * NVIDIA P4 1 * 8 3.0 800 Yes 4 8 20
ecs.gn5i-c16g1.8xlarge 32 128.0 None 2 * NVIDIA P4 2 * 8 6.0 1,200 Yes 8 8 20
ecs.gn5i-c28g1.14xlarge 56 224.0 None 2 * NVIDIA P4 2 * 8 10.0 2,000 Yes 14 8 20
Note