Amazon Web Services (AWS), this week announced a new Elastic Compute Cloud instance family (EC2) for graphics-intensive workloads.
According to Friday’s announcement by AWS, the new G3 instances “make obtaining a powerful combination central processing (CPU), random-access memory(RAM) easy.”
The G3 is available from AWS’ Northern Virginia and Northern California regions, Ohio, Oregon and Ireland, and U.S. GovCloud areas. It joins the F2 and P2 families in AWS’ lineup of instances for “accelerated computing”.
The G3 is optimized for server-side graphics applications such as 3-D rendering, virtual reality, and visual encoding. It is ideal for organizations that do a lot of complex modeling and 3D visualization analyses (such as medical image processing and computer-aided designing and seismic visualization jobs), AWS stated.
These new instances are powered by Intel Xeon E5-2686 v4 processors, and NVIDIA Tesla M60 graphics cards. Jeff Barr, AWS evangelist, wrote that G3 users get an “NVIDIA Grid Virtual Workstation License” and can use the NVIDIAGRID driver without having to purchase a license.
The G3 is available in three sizes currently. The smallest size, g3.4xlarge features a single GPU and 16 virtual CPUs. It also has 122GiB host memory and 8GiB GPU memory.
The g3.8xlarge is twice as big, with two GPUs, 32 CPUs, 244GiB host memory, and 16GiB GPU memory.
The largest is the g3.16xlarge, which has four GPUs and 64 vCPUs. It also includes 488GiB host memory and 32GiB GPU memory.
AWS was the first cloud provider to offer customized instances for graphics-intensive applications. We launched the CG1 type of instance in 2010 to provide a cost-effective and high-performance instance for graphics heavy applications. Today, G3, our third generation GPU powered instance, serves the most demanding graphics workloads like 3D rendering and data visualisation,” stated Matt Garman, vice-president of EC2 at AWS in a prepared statement.
