Silicon Updates: Amazon EC2 G3 instances for graphic intensive applications

Amazon is introducing G3 instances specific to accelerate graphic intensive applications, based on Amazon EC2 Accelerated Compute instances. The G3 instances help to achieve the combination of GPU, CPU and host memory for complex workloads. The host memory includes different complex operations like 3D rendering and 3D visualisations.
Read more: Amazon EC2 F1 for custom hardware acceleration.

Amazon EC2 G3 instance

The G3 instances provides twice the capacities of the most powerful GPU instances available today in every aspect - twice the CPU power, twice the power per GPU, and twice the host memory per GPU. The G3 instances provides the flexibility for complex modelling and 3D visualisation analyses. These capabilities of the G3 instances can be applied in digital image processing, seismic visualisations, geographic analysis visualisations and computer-aided designs.
There are three sizes of G3 instances available - g3.4xlarge, g3.8xlarge, and g3.16xlarge. But, these instances are available across different regions of the US.

Different sizes of Amazon EC2 G3 instances

Out of the variants of Amazon EC2 G3 instances, G3.16xlarge is the largest one. It consists of four NVIDIA Tesla M60 GPUs, 64 vCPUs using custom Intel Xeon E5-2686v4 (Broadwell) processors, Amazon Elastic Block Store (EBS) optimized hardware, 488 GiB of host memory, and Enhanced Networking using the Amazon EC2 Elastic Network Adaptor. The G3 instances are the first-of-its-kind, equipped with the capabilities of supporting NVIDIA GRID Virtual Workstation. Also, the G3 instances support streaming for four 4k monitors. In terms of hardware encoding, they support up to 10 High Efficiency Video Coding (HEVC) H.265 1080p30 streams for faster video frame processing. In other words, they support up to 18 H.264 1080p30 streams per GPU for faster video frame processing and improved image fidelity.
Read more: Xelium - Cadence's multi-core parallel simulation engine.