Organizations with high performance graphic workloads will soon be able to run an AMD-powered workstation in the cloud as AWS has announced that it is developing new Amazon Elastic Compute Cloud (EC2) instances in the G4 instance family.
These new instances will improve performance while reducing cost for graphics-intensive workloads such as game streaming, animation and video rendering.
AWS’ new G4ad instances will feature AMD’s latest Radeon Pro V520 GPUs as well as the chipmaker’s 2nd generation EPYC processors and they are the first instances in EC2 to feature AMD GPUs.
G4dn vs. G4ad instances
Last year AWS released its G4dn instances which feature Nvidia T4 GPUs and up until now, they were the most cost-effective GPU-based instances in EC2. G4dn instances are ideal for deploying machine learning models in production as well as graphics-intensive applications.
However, when compared to G4dn instances, the new G4ad instances enable up to 45 percent better price performance for graphics-intensive workloads as well as a 40 percent improvement in performance.
While G4dn instances remain the best option for small-scale machine learning (ML) training and GPU-based ML inference due to included hardware optimizations like Tensor Cores and access to Nvidia libraries such as CUDA, CuDNN and NVENC, G4ad instances benefit from the improved price and performance offered by AMD’s CPUs and GPUs.
When G4ad instances become available soon in US East (N. Virginia), US West (Oregon) and Europe (Ireland), customers will be able to choose from three instance sizes depending on the number of GPUs, GPU memory, vCPUs, memory, storage and EBS and network bandwidth they require.