Posted On: Oct 6, 2020
Amazon EC2 has the cloud’s broadest and most capable portfolio of hardware-accelerated instances featuring GPUs, FPGAs, and our own custom ML inference chip, AWS Inferentia. G4dn instances offer the best price/performance for GPU based ML inference, training less-complex ML models, graphics applications others that need access to NVIDIA libraries such as CUDA, CuDNN and NVENC.
Starting today, Amazon EC2 G4dn bare metal instances are now available in 15 additional regions. The Amazon EC2 G4dn bare metal instances are AWS’s first bare metal GPU instances. These instances are ideal for workloads that require access to the hardware feature set (such as Intel® VT-x), for applications that need to run in non-virtualized environments for licensing or support requirements, or for customers who wish to use their own hypervisor.
The G4dn bare metal instances offer eight NVIDIA T4 Tensor Core GPUs, AWS custom second generation Intel® Xeon® Scalable (Cascade Lake) processors, 100 Gbps of networking throughput, and 1.8 TB of local NVMe storage. G4dn bare metal instances also support Elastic Fabric Adapter which can be used to run applications requiring high levels of inter-node communications at scale, with lower and more consistent latency and higher throughput than traditional TCP channels. Managed AWS services such as Elastic Container Service (ECS) and Elastic Kubernetes Service (EKS), that support G4dn instances, also support the new G4dn bare metal instances.
Amazon EC2 G4dn bare metal instances are available in US East (Ohio), US West (N. California), Canada (Central), Europe (Frankfurt, Ireland, London, Milan, Paris, and Stockholm), Asia Pacific (Mumbai, Seoul, Singapore, and Sydney), South America (São Paulo), and Africa (Cape Town). Amazon EC2 G4dn instances are available in 6 sizes, with 4, 8, 16, 32, 48, and 64 vCPUs, in addition to the bare metal instance that has 96 vCPUs, and are purchasable as On-Demand, Reserved or Spot Instances, as well as part of Savings Plans.