Introduction of the Next Generation GPU-Powered EC2 instances the new G3 instance type

The GPU powered computing was launched in 2013 when the G2 instance type was launched. Amazon Web Service Customers used G2 instances for delivering high-performance graphics to Desktop, Mobile devices, and TV sets. AWS has launched G3 instance type which is powered by the NVIDIA Tesla M60 GPUs where all these instances are available in all three sizes.

Model	GPUs	GPU Memory	vCPUs	Main Memory	EBS Bandwidth
g3.4xlarge	1	8 GiB	16	122 GiB	3.5 Gbps
g3.8xlarge	2	16  GiB	32	244  GiB	7  Gbps
g3.16xlarge	4	32  GiB	64	488  GiB	14  Gbps

 Every GPU supports 8 GiB of GPU memory with 2048 parallel processing cores and hardware with a encoder which is capable of supporting up to 10 H.265 (HEVC) 1080p30 streams and also up to 18 H.264 1080p30 streams which makes them preferred fit for 3D rendering and visualization, video encoding, virtual reality, other server-side graphics workload which will need a huge amount of parallel processing power and remote graphics workstation (NVIDIA GRID). The GPUs support OpenCL 1.2, OpenGL 4.5, CUDA 8.0 and DirectX 12.0. You can operate the NVIDIA GRID Virtual Workstation License when you launch the G3 instance and you can utilize the NVIDIA GRID driver without purchasing a license.

The G3 instance uses the Intel Xeon-2686 v4 processor running the speed of 2.7 GHz. Enhanced Networking offers up to 2- Gbps of accumulative network bandwidth which is within the placement group with 14 Gbps of EBS bandwidth.

Amazon Web Service Customer that they are looking forward to visualizing configuration of cars in 3D, huge 3D seismic models and also offers the student an ability to run high-end 2D and 3D applications.